Post on 01-May-2018
transcript
Research ArticleReverse Auction Based Green Offloading Scheme forSmall Cell Heterogeneous Networks
Xiaodong Xu Chunjing Yuan Jianhui Li Huixin Zhang and Xiaofeng Tao
National Engineering Laboratory for Mobile Network Security Beijing University of Posts and Telecommunications Beijing China
Correspondence should be addressed to Xiaodong Xu xuxiaodongbupteducn
Received 1 February 2016 Accepted 17 July 2016
Academic Editor Gabriel-Miro Muntean
Copyright copy 2016 Xiaodong Xu et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
The small cell is treated as a promising proposition to provide hot spot capacity and higher data rates However even with densesmall cell deployment scenario the heavy traffic load pressure and low energy efficiency in the small cell heterogeneous network(HetNet) still exist Therefore how to make the best use of densely deployed small cells under HetNet environments becomesthe focus of researches Offloading provides a feasible solution to promote cooperation between macrocells and small cells foruser traffic supporting In this paper we propose the reverse auction based Green Offloading (GO) scheme for energy efficiencyimprovements The proposed GO scheme employs the reverse auction theory to handle the offloading decision process aimingat maximizing the system energy efficiency under the constraints of user Quality of Service (QoS) requirements bandwidth andtransmission power limitations Moreover the reverse auction model gives the facility of multicell coordination transmissionswith multiple winners situation The energy efficiency optimization problem with constraints is solved by Dynamic Programmingmethod with Karush-Kuhn-Tucker (KKT) conditionsThe Individual Rationality and Truthfulness of the reverse auctionmodel arealso proved By comparing the energy efficiency performances of the proposedGOschemewith currentworkswithin the LongTermEvolution-Advanced (LTE-A) system downlink scenario simulation results show the improvements of the proposed GO scheme
1 Introduction
With the rapid development of wireless communicationsthere is a growing conflict between increasing mobile trafficdemands and limited radio resources Many projects predictan over 500-fold growth on mobile data traffic in 10 years(2010ndash2020) [1] In order to satisfy this tremendous demandlots of small cells have been deployed in both indoor andoutdoor environments for hot-spot capacity improvements[2]Moreover the corresponding research and standardworkfor the 5th Generation Mobile Communication System (5G)are in a full swing [3 4] For future 5G networks the small cellheterogeneous network (HetNet) is believed as an importantsolution to fulfill the capacity booming requirements
The deployment of the small cell HetNet offers enhancedcapacity and expansile coverage and alleviates the heavytraffic burden in macrocells Relying on above advantagesmobile operators all over the world are now keen on smallcell HetNet deployments According to statistics the totalnumber of already-deployed small cells has exceeded that
of macrocells [5] Despite the fact that lots of small cellshave been deployed densely simply relying on increasingthe spatial density of small cells is not an economicallysustainable manner to fulfill the growing demands Becausethe deployment of a new Base Station (BS) goes along withlarge expenses like energy consumptions laying of dedicatedbackhaul and network replanning processes therefore howto take full advantages of the already-deployedmacrocell andsmall cell BSs becomes the key issue for the mobile industryIn order to solve this problem researches have been focusingon the offloading strategy gradually which is believed as apromising paradigm to improve the network utilizing fordensely deployed small cell HetNets [6 7]
Offloading is a kind of technology that transfers the trafficload from one network to another under certain conditionstypically in case of heavy traffic burden or low efficiencynetworks One typical traditional scenario for offloadingapplication is about the cellular network andWireless Fidelity(WiFi) network Through offloading the cellular traffic toWiFi networks the load pressure of the cellular network
Hindawi Publishing CorporationMobile Information SystemsVolume 2016 Article ID 5087525 10 pageshttpdxdoiorg10115520165087525
2 Mobile Information Systems
could be relieved [8] In recent years some researcheshave been conducted to refine the offloading scheme withincellular networks to alleviate the increasing traffic burdenof macrocells Authors of [9] analyze the proportion oftraffic that can be offloaded from macrocells to femtocellsto increase the system capacity The offloading gain for themacrocell and femtocell colocatedHetNets is explored in [10]Except for enhancing capacity purposes the energy efficiencymetric is also considered during offloading processes In [11]the system energy efficiency is improved through offloadingthe high Quality of Service (QoS) requirement traffic to fem-tocells But the fairness between different users is neglectedIn [12] authors verify the offloading gain in the macrocelland femtocell orthogonal frequency division multiple accesssystem by analytical evaluations Nevertheless the transmis-sion power allocation is not taken into account Benefittingfrom the information exchange between macrocells andpicocells a traffic offloading based on fractional frequencyreuse (TOFFR) algorithm is proposed to improve the systemenergy efficiency in [13] Authors in [14] propose a trafficoffloading scheme based on reverse auction mechanism andgreedy algorithms which incentivizes femtocell owners torent their underutilized spectrum for the enhancement ofnetwork performances in HetNets
From the analyses of above researches there are severalmethods adopted in offloading processes The stochasticgeometry is adopted to analyze the statistical status of theoffloading gain [9] Game theory approaches are explored byoffloading researches in [15 16] Furthermore the auctiontheory is also employed to support the offloading targetdecision and network performance improvements [10 14 1718] The forward auction model is applied in [10] for theoffloading between different operators References [14 18]employ the reverse auction model to establish the offloadingprocedures the properties of the proposed models are alsoanalyzed We also applied the forward auction model toimprove the system energy efficiency of offloading in theHetNet [17]
In economics auction is a common means to determinethe value of an itemwhich has a valuable price Nowadays theauction theory has been widely used in various fields suchas the dynamic spectrummanagement in cognitive networks[19] and hybrid access in HetNets [20] The auction model isparticularly suitable for the processes that need informationexchanges and target choosing with pricing issues Most ofthe auctionmodels employed currently are forward auctionswhich typically involve a single seller and multiple buyersThe buyers send bids to compete for the item sold by theseller The forward auction model is popular in dealing withthe decisions amongmultiple operators to getminimumcosts[10] while the reverse auction model could be used in thesituation that one buyer wants to offer reasonable prices toget service from multiple sellers [21] The offloading processis just adapted to the reverse auction applications that theoffloading user is treated as the buyer while the candidateBSs are acting as the sellersTherefore in this paper a reverseauction process with first price sealed bid mechanism isadopted in the offloading researchThe reverse auctionmodelinvolves a single buyer and multiple sellers where the buyer
makes the decision on whose bid will be accepted accordingto the bids sent by multiple sellers
For the offloading process researched in small cell Het-Nets there are two main concerns adapted to the reverseauction model
(1) The offloading user needs to select the target BSwhich is modeled as one ldquobuyerrdquo with multicandidateldquosellersrdquo The offloading user is acting as a buyer toget service from multicandidate BSs as sellers Thecandidate BSs offer their bids to compete for the serv-ing opportunity of the offloading user which couldimprove their performances such as throughputs orresourceenergy efficiency
(2) The aforementioned ldquobidrdquo offered by the candidateBSs could be their available subcarriers and trans-mission power The ldquobuyerrdquo needs to pay with theprice determined by the reverse auction for gettingthe service where the price is usually defined as theopportunity cost of the offloading target decision
In order to improve the system energy efficiency in thedense small cell HetNet deployment scenario we propose aGreen Offloading (GO) scheme through the Vickrey-Clarke-Grove (VCG) based reverse auction model [22] The smallcells will be incentivized to help offloading the traffic frommacrocells which allows the operator to express diversepreferences for densely deployed HetNets The properties ofthe reverse auction model such as the IndividualRationalityand 119879119903119906119905ℎ119891119906119897119899119890119904119904 are also proved in this paper to guaran-tee the optimality of the proposed reverse auction mecha-nism
In this paper the employed reverse auction model withfirst price sealed bidmechanism is formulated for the offload-ing process The energy efficiency optimization problemwithin the offloading process is modeled with constraints ofuser guaranteed QoS requirements The optimization prob-lem is solved by the Karush-Kuhn-Tucker (KKT) conditionsand Dynamic Programming method The system-level sim-ulation evaluations are conducted for the proposed reverseauction based GO scheme with taking current schemes [1314] as comparisons
The contributions of this paper include three aspects
(1) The reverse auction model is explored in theoffloading scheme design with the first price sealedbid mechanism which matches the mobility basedoffloading process well with limited negotiations anddelays
(2) The multiple winning bidders are also supported forBS coordination transmissions
(3) The optimization of the system energy efficiencyimprovement is solvedwith guaranteed user through-put in the offloading target decision and reallocationof both the resource block and transmission power
The rest of this paper is organized as the following InSection 2 the system model is described and the reverseauction model applied in this paper is clarified The reverse
Mobile Information Systems 3
auction based GO scheme for small cell HetNets is proposedin Section 3 In Section 4 the simulation results are presentedand discussed with comparison schemes Finally concludingremarks are drawn in Section 5
2 System Model
In this section the system model of the proposed reverseauction basedGOscheme is formulatedThenetwork deploy-ment scenario will be introduced at first and the notationdefinitions about the reverse auction model based offloadingprocess are described afterwards
21 Deployment Scenario We focus on the small cell andmacrocell overlaidHetNet deployment scenario in this paperAccording to the 3rd Generation Partnership Project (3GPP)Long Term Evolution-Advanced (LTE-A) standards due tothe scarcity of available spectrum the macrocell tier andsmall cell tier will be inevitably deployed with the sharedspectrum manner In this paper we will focus on theshared spectrum scenario with considerations of cross-tierinterferences which is more realistic Furthermore the coor-dination technologies such as the Coordinated MultipointTransmissionReception (CoMP) defined by 3GPP [23] aresupported in above small cell HetNet deployment scenarioCoMP will bring benefits to the cell edge throughput andenergy efficiency which helps the offloading users usuallylocated in the cell edge area
In order to address the dense deployment scenario forfuture heavy-traffic requirements each macrocell coveragearea is covered by several small cell clusters the numberof which will be 1 4 or 10 based on 3GPP simulationmethodology [24] Each small cell cluster includes severalsmall cell BSs the number of which will be 4 or 10 [25]Both small cell clusters and small cell BSs are considerablyless planed as opposed to the typical planned macrocelldeployments
Based on the standard discussions on the small cell Het-Net deployments in 3GPP [26] themacrocell will take chargeof the control plane for both the macrocell and colocatedsmall cells within its coverage The main responsibility of thesmall cell tier is to offload the high data-rate service fromthe macrocell while the macrocell tier handles the low data-rate traffic and high mobility users This is the typical actualdeployment demand for the HetNet The traffic model of theuser is File Transfer Protocol (FTP) Model 1 as full-buffer[23] The system energy efficiency will be the optimizationobjective during the offloading processes with constraints ofuser QoS guaranteeing
The handover solution for the offloading operation isdifferent from the standard handover criterion due to the usermobility For small cell HetNets 3GPP also discussed newhandover criterion for macro-to-small cells with Cell RangeExtension (CRE) which is defined by 3GPP standard [27]Assisted by the CRE the downlink Reference Signal ReceivedPower (RSRP) of small cells could be added with a bias Bythis means the user will be encouraged to do handover tosmall cells which will help to implement the handover foroffloading users or load balancing operations
Buyer UE
Seller BS1Seller BS2
Seller BS3
Bid
Bid
BidBids collectionresources
AuctionAllocation
Pricing
Response
Resource
Energy efficiency
Figure 1 The framework of reverse auction based GO scheme
22 Reverse Auction One typical implementation scenarioof GO scheme is given in this section Considering the usersrsquoimmense offloading potential in dense small cell HetNet areverse auction model based offloading scheme is suitableto motivate users to conduct traffic offloading that couldachieve higher system energy efficiency Figure 1 illustratesthe typical scenario of offloading process with reverse auctionmodel According to the reverse auction model the UserEquipment (UE) acts as the buyer who ensures highersystem energy efficiency in exchange of bandwidth andtransmission power resources provided by the BS to serve theoffloading UE When the UE requests data transmission itscurrent serving BS is encouraged to broadcast the offloadingrequest to all of its neighboring BSs All the BSs receivingthe request will send their bids along with their availableresources to the UErsquos current serving BS The informationincluded in each bid contains the amount of the availablebandwidth and transmission power that will be providedfor the offloading user Then the UErsquos current serving BScalculates the throughput that each bidder can provide todecide whether it could satisfy the UErsquos QoS requirement
The target BS for offloading will be the winner of thereverse auction process The conditions for judging thewinner bid could be set as the optimal objectives for theoffloading process In this paper the optimal objective of theproposed reverse auction based GO scheme is to maximizethe system energy efficiency subjected to the UErsquos minimumthroughput requirement and bidderrsquos available resourcesduring the offloading process The reverse auction based GOscheme involves two steps 119860119897119897119900119888119886119905119894119900119899 and 119875119903119894119888119894119899119892
In the119860119897119897119900119888119886119905119894119900119899 step theUErsquos current serving BS decideswhich bidders will be the auction winners As describedabove the coordination techniques among different BSs aresupported in the offloading scheme Therefore the winnermay be one BS or several BSs with the same bids
In the 119875119903119894119888119894119899119892 step the improvement of system energyefficiency is evaluated as a payment from the UE Finallythe UErsquos current serving BS sends back the auction resultto the bidders which consist of the required resources andthe expected energy efficiency incrementThen the handoverprocess for the offloading will start As shown in Figure 1
4 Mobile Information Systems
the winning bidders are BS1 and BS3 together who win thereverse auction and are designated as the offloading target BSsto serve the UE coordinately
In order to present the reverse auction based GO schemeclearer the related notation definitions are introduced asfollows
Bid (119887119894) submitted by the 119861119878
119894to convey how much
bandwidth and transmission power it can provide tothe offloading UE which may not always equal allavailable resources the BS can providePrivate Value (119909
119894) available resources in the 119861119878
119894
which is only known by the 119861119878119894itself
Pricing (119901119894) the opportunity cost of the 119894th BS which
makes sure the system energy efficiency could bemaximum
Based on the auction theory when the condition 119887119894= 119909119894
is satisfied the auction process is truthful Moreover 119887119894=
119909119894is a weakly dominant strategy [20] As a result in this
paper we set 119887119894= 119909119894 assuming all the bidders participating
in the auction process will send the bids with all availableresources it can provide which guarantees the truthfulnessof the reverse auction model
Although we have set the 119861119894119889 equal to the PrivateValuethe reverse auction basedGO schemewithmultiple candidatetarget BSs is still NP-hard problem In order to solve thisproblem an approximation algorithm is designed in the nextsection The notations used in the algorithm are introducedin the Notation Definition
23 User Delay Tolerance The reverse auction based GOscheme will start upon the receipt of offloading request at theserving BS periodically For the traditional auction processthere usually exist multiround bidding procedures to achievethe final winner This process would inevitably generate anextra delay for the BSs to wait for the auction consequenceBut for the wireless network the user traffic usually hasdelay tolerance requirements which should be consideredin the offloading scheme Therefore we implement thesingle-round auction in this paper in order to prevent theinformation exchange overhead and corresponding delay forthe offloading users Since the number of BSs surrounding aspecific UE is limited the extra delay caused by the single-round first price sealed bid auction process could be notsignificant This delay avoidance mechanism is also feasiblefor the handover based offloading schemes that usually needmore time for measurements and handover decisions
3 Reverse Auction Based GO Scheme
In this section the proposed reverse auction based GOscheme is described The main steps of the proposed GOscheme are given in Figure 1 Firstly according to theoffloading request the UErsquos current serving BS collects allthe bids from its neighboring BSs Then the serving BScalculates the throughput that all bidders can provide andderives the expected energy efficiency increment for eachcandidate BS Based on the derived throughput and energy
efficiency increment a single-round reverse auction processis performed which includes the 119860119897119897119900119888119886119905119894119900119899 and 119875119903119894119888119894119899119892steps Finally the auction results are sent back to the biddersand the user will be offloaded to the winners accordingly
31 Bidding In order to contribute to the offloading processbidders will append available resources with their bids toreveal the throughput they can provide for the offloadinguser For each bidder the upper bound of bandwidth andtransmission power that a BS can provide is119861bound and119875boundrespectively 119861bound and 119875bound can be divided into multipleunits and classified asmultiple bids b = b1 b2 b119894 b119897to indicate the resources the UE can obtain from each bidderwhere 119897 = max(lfloor119861bound119890119861rfloor lfloor119875bound119890119875rfloor) 119890119861 is the basicbandwidth unit 119890
119875is the basic transmission power unit for
the bidder The b119894 = (119861119894 119875119894) consists of both bandwidth andtransmission power resources After receiving all of the bidsUErsquos current serving BS can know how many resources eachbidder can provide with the value not larger than the sum119897
119894=1119861119894
and sum119897119894=1119875119894 The scale of the bandwidth and transmission
power unit can be flexibly set by the system The smallerunit definition leads to more information in the bid whichimproves the performance of the auction process But it willalso generate more computational costs and increase thecomplexity In this paper one Resource Block (RB) and 01Wtransmission power are chosen as the basic bandwidth unitand transmission power unit respectively We choose 01Was the basic transmission power here because the DynamicProgramming will be adopted in the next subsection to solvethe optimization problem where the integer data are neededby the Dynamic Programming method
32 Reserve Auction Algorithm As mentioned above thereverse auction process includes two steps of the 119860119897119897119900119888119886119905119894119900119899and 119875119903119894119888119894119899119892
321 Allocation In traditional reverse auction processes theallocation result is completely decided by the bids that isthe bidders who offer the largest supply of resources will winthe auction However in this paper besides the resourcesthat the bidders can provide the energy efficiency achievedby the bidder should also be considered Assume that B =B1B2 B
119895 B
119899 and P = P
1P2 P
119895 P
119899
represent the allocation result where B119895= 1198611
119895 1198612
119895 119861
119897119895
119895
is the RB that 119895th BS could provide and 119861119894119895= 0 if the 119894th RB
in the 119895th BS is not needed P119895= 1198751
119895 1198752
119895 119875
119897119895
119895 denotes the
transmission power that the 119895th BS could transmit and 119875119894119895is
the transmission power on 119861119894119895 If B119895or P119895equals zero the 119895th
BS loses in this auction processThe 119860119897119897119900119888119886119905119894119900119899 problem is formulated as
maxB119895 P119895
119862system
119875119905
(1)
st119899
sum
119895=1
119897119895
sum
119894=1
119861119894
119895log(1 +
119875119894
119895119867119895
119868119894
119895+ 1198730119861119894
119895
) ge 120591119903forall119903 (2)
Mobile Information Systems 5
119861119894
119895isin 0 119861 (3)
119875119894
119895isin P119898P119904 (4)
P119898=
119875119894
119895| 0 le 119875
119894
119895le 119875119898
119897119895
sum
119894=1
119875119894
119895le 119875119898
(5)
P119904=
119875119894
119895| 0 le 119875
119894
119895le 119875119904
119897119895
sum
119894=1
119875119894
119895le 119875119904
(6)
In (1) 119862system = sum119899
119895=1119862119895sum119897119895
119894=1119861119894
119895denotes the throughput
119875119905= sum119897119895
119894=1119875119894
119895sum119899
119895=1119897119895denotes the expectation of transmission
power on each RB from all the bidders 119862system119875119905 is thesystem energy efficiency and 119875
119905gt 0
In (2) 119867119895= 119889minus120572119895
119895|ℎ119895|2120594119895is the channel gain between
the 119895th BS and UE where 119889119895is the distance between the
119895th BS and UE 120572119895is the path-loss exponent of the 119895th
BS ℎ119895is the Rayleigh fading component 120594
119895denotes the
log-normally distributed shadow fading Furthermore 119868119894119895=
sum119899
1198951015840=11198951015840=119895119875119894
11989510158401198671198951015840V119894
1198951015840 is the interference experienced by the UE
on the 119894th RB where 1198751198941198951015840 is the transmission power on the 119894th
RB from the 1198951015840th BS1198671198951015840 is the channel gain between the 1198951015840th
BS andUE V1198941198951015840 is the binary variables representing the activity
factor of 1198951015840th BS 1198730is the thermal noise level and 120591
119903is the
guaranteed throughput threshold of the offloading UEConstraint (3) means the RB 119861119894
119895can be occupied or
vacant Constraint (4) denotes the bidder can be a macrocellBS or a small cell BS In (5) and (6) we give the requirementfor the transmission power in the corresponding RB where119875119898and 119875119904are the power limitations for the macrocell BS and
small cell BS respectivelyFor the convenience of solving this problem we trans-
form (1) (2) and (3) into the following form
minB119895P119895
119875119905
119862system
119862system ge 120591119903
0 le 119861119894
119895le 119861
(7)
Assuming the 119862system is an independent variable theKarush-Kuhn-Tucker (KKT) conditions are given as
nablaB119895P119895system119875119905
119862systemminus nablaB119895P119895system (119862system minus 120591119903) 120583
minus nablaB119895 P119895system119861119894
119895120590119894
119895minus nablaB119895P119895system (119861 minus 119861
119894
119895) 120585119894
119895
minus nablaB119895 P119895system119875119894
119895]119894119895minus nablaB119895P119895system (119875119904 (119875119898) minus 119875
119894
119895) 120582119894
119895
minus nablaB119895 P119895system(119875119904 (119875119898) minus
119897119895
sum
119894=1
119875119894
119895)120588119895= 0
(8)
(119862system minus 120591119903) 120583 = 0 (9)
119861119894
119895120590119894
119895= 0 119861
119894
119895ge 0 120590
119894
119895ge 0 (10)
(119861 minus 119861119894
119895) 120585119894
119895= 0 119861 le 119861
119894
119895 120585119894
119895ge 0 (11)
119875119894
119895]119894119895= 0 119875
119894
119895ge 0 ]119894
119895ge 0 (12)
(119875119904(119875119898) minus 119875119894
119895) 120582119894
119895= 0 119875
119894
119895le 119875119904(119875119898) 120582119894
119895ge 0 (13)
(119875119904(119875119898) minus
119897119895
sum
119894=1
119875119894
119895)120588119895= 0
119897119895
sum
119894=1
119875119894
119895le 119875119904(119875119898) 120588119895ge 0 (14)
Equation (8) is changed into (15) by taking the derivativeof 119862system as
minus119875119905
1198622systemminus 120583 = 0 (15)
119875119905
1198622system+ 120583 = 0 (16)
Thismakes sensewhere119875119905and1198622system are nonzero values
Thus 120583 = 0 (9) shows (119862system minus 120591119903)120583 = 0 where
119862system = 120591119903 (17)
When the offloading process is triggered the offloadeduser traffic requirements and the number of BSs participatingin the auction are known The KKT conditions prove that(9) is guaranteed throughput requirement The 119860119897119897119900119888119886119905119894119900119899problem is transformed to solve minB119895P119895119875119905 which can be
simplified into a linear objective functionminsum119897119895119894=1119875119894
119895sum119899
119895=1119897119895
Because the above119860119897119897119900119888119886119905119894119900119899problem is a linear problemit is easy to find that this is a multiple knapsack problemIn order to facilitate the solving process we turn it intoa 0-1 knapsack problem There are 119899 BSs to participate inthe auction the 119894th BS resources are separated into 119872
119894
independent piece of ldquogoodsrdquo that can be loaded into aknapsack This will get a 0-1 knapsack problem where thenumber of items is sum119872
119894 If we solve this problem directly
the computational complexity will be 119874(119862sum119872119894) In order to
reduce the complexity we design another algorithm with thespecific plan as follows
As mentioned above the 119894th BS bandwidth resourceis composed of a number of RB groups Considering thecondition of the binary the guaranteeing of selecting anymultiple resource package strategy still can be achievedafter the transformation of the original multiple knapsackproblem One BS which has119872
119894resource blocks is separated
into several RB groups where these RB groups respectivelyhave 1 2 22 23 119872
119894minus 2119896minus1+ 1 resource blocks The 119894th BS
has Ceiling(log119872119894) different RB groups participating in the
0-1 knapsack problemInitialize the value as follows the weight of knapsack is
119862 which is the offloading user traffic requirements 120591119903 The
starting value 119865[0 119888] = 0 119865[119909 0] = 0 And the 119894th BS hasCeiling(log119872
119894) stages
6 Mobile Information Systems
(1) Let 1205780= 0
(2) for 119895 = 1 to 119899 do(3) if Bidder b119895 is a macrocell BS then(4) for 119875 = 1 to 10119875
119898do
(5) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(6) end for(7) for 119894 = 1 to 119897
119895do
(8) if 119875119894119895gt 0 then
(9) 119861119894
119895= 119861
(10) else(11) 119861
119894
119895= 0
(12) end if(13) end for(14) else if Bidder b119895 is a small cell BS then(15) for 119875 = 1 to 10119875
119904do
(16) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(17) end for(18) for 119894 = 1 to 119897
119895do
(19) if 119875119894119895gt 0 then
(20) 119861119894
119895= 119861
(21) else(22) 119861
119894
119895= 0
(23) end if(24) end for(25) end if(26) end for
Algorithm 1 Reverse auction 119860119897119897119900119888119886119905119894119900119899 (119899 b119899)
Renumber all RB groups 119888119909is the capacity of the 119909
stage which denotes the 119894th RB group of the 119895th BS and thecorresponding value is V
119909
The capacity of the 119909 stage is
119888119909= 119908 [119894 119895] = 119861
119894
119895log(1 +
119875119894
119895119867119895
119868119894
119895+ 1198730119861119894
119895
) (18)
The value of the corresponding V119909is
V119909= V [119894 119895] = 119875119894
119895 (19)
The iterative equation will be
119865 [119909 119888] = min 119865 [119909 minus 1 119888] 119865 [119909 minus 1 119888 minus 119888119909] + V119909 (20)
In (20) 119865[119909 119888] is the minimum value of the transmissionpower in the stage 119909 119888 denotes the remaining space ofthe pack according to the current stage 119888
119909denotes the
provided capacity when choosing the 119909 stage V119909denotes the
provided transmission power when choosing the 119909 stageThecomputational complexity is reduced to 119874(119862sum log119872
119894)
The proposed 119860119897119897119900119888119886119905119894119900119899 algorithm is illustrated inAlgorithm 1 with B = B
1B2 B
119899 and P = P
1P2
P119899In Algorithm 1 the transmission power of each bidder is
chosen firstly and the corresponding bandwidth is decidedbased on the transmission power allocation results As
for 119895 = 1 to 119899 do(2) if 119895th BS is a winning bidder then
Reverse Auction - 119860119897119897119900119888119886119905119894119900119899 (119899 119895B b119894)(4) 119901
119895= 120578B119899b119894 minus (120578B119899 minus 120578b119894 )
else(6) 119901
119895= 0
end if(8) end for
Algorithm 2 Reverse auction 119875119903119894119888119894119899119892 (119899 b119899BP 120578)
mentioned before 01W is chosen as the basic transmissionpower unit and the integer data is needed in the DynamicProgramming Therefore the range of 119875 is defined from 1
to 10119875119898or 10119875
119904 The equation in Line (5) and Line (16) of
Algorithm 1 mean that the 119895th bidderrsquos transmission powershould be chosen to achieve the largest 120578b119895 and guarantee theoffloaded userrsquos throughput threshold 120591
119903at the same time
where 120578b119895 denotes the energy efficiency of the 119895th bidderAfter implementing the Dynamic Programming the resultsof Algorithm 1 P10 and B will be the optimal allocationsolution for the proposed reverse auction process
322 Pricing In traditional 119875119903119894119888119894119899119892 algorithms the biddersare encouraged to set their own bids truthfully as illustratedbefore So in this paper the same energy efficiency that thecorresponding bidder achieves is paid back With regard tothe offloading user throughput threshold 120591
119903 we define 120578
1and
1205782as (21) and (22) as follows
1205781= 120578B119899b119894 = max
B119895B119894P119895P119894
119862system
119864 (119875119905)
(21)
1205782= 120578B119899 minus 120578b119894 = (max
B119895P119895
119862system
119864 (119875119905)) minus
119862b119894
119864 (P119894) (22)
where 1205781denotes the system energy efficiency under the
optimal 119860119897119897119900119888119886119905119894119900119899 solution without the presence of the 119894thBS The 120578
2denotes the system energy efficiency except for
the 119894th BS under current optimal119860119897119897119900119888119886119905119894119900119899 resultsThen theopportunity cost of the 119894th BS is defined as the differencebetween 120578
1and 1205782 just as illustrated in (23) [19] as follows
119901119894= 1205781minus 1205782= 120578B119899b119894 minus (120578B119899 minus 120578b119894) (23)
The 119875119903119894119888119894119899119892 algorithm is given as Algorithm 2
323 Properties In this section the properties of theproposed reverse auction model are analyzed Accordingto the VCG based reverse auction model the IndividualRa-tionality and the 119879119903119906119905ℎ119891119906119897119899119890119904119904 properties need to be proved
IndividualRationality When the utility of each participatingbidder in the119875119903119894119888119894119899119892 stage is greater than zero this algorithmis individual rational for each winning bidder Namely
119901119894= 120578B119899b119894 minus (120578B119899 minus 120578b119894) ge 0 (24)
Mobile Information Systems 7
119879119903119906119905ℎ119891119906119897119899119890119904119904 For each bidder the Truthfulness means thateach bidderrsquos bid price is equal to its private value This isa weakly dominant strategy If BSrsquos bidding is untrue theenergy efficiency will be unlikely the biggest In order toget the maximum energy efficiency the allocation should beformulated as follows
119901119895= 120578B119899b119895 minus (120578B119899 minus 120578b119895)
120575 = 119901119895minus 119901119894
= 120578B119899b119895 minus (120578B119899 minus 120578b119895) minus [120578B119899b119894 minus (120578B119899 minus 120578b119894)]
= 120578B119899b119895 minus 120578B119899 + 120578b119895 minus 120578B119899b119894 + 120578B119899 minus 120578b119894
= 120578B119899b119895 + 120578b119895 minus 120578B119899b119894 minus 120578b119894
= (120578B119899b119895 + 120578b119895) minus (120578B119899b119894 + 120578b119894)
(25)
Based on the proposed model in this paper because 119901119895le
119901119894and120575 le 0 thismeans 120578B119899b119895+120578b119895 le 120578B119899b119894+120578b119894 If and only
if 119895 = 119894 it can take the equal signTherefore each bidder mustbe truthful to obtain the maximum system energy efficiencyThe proof is finished
4 Performance Evaluation
In this section we built the system-level simulation plat-form according to the 3GPP LTE-Advanced simulationmethodology [23] Based on this platform we validate theperformances of the proposed reverse auction based GOscheme with comparison algorithms in the small cell HetNetdownlink scenario
41 Simulation Setting Performance Metrics and ComparisonAlgorithms Theconsidered simulation scenario in this papercomprises 19-hexagonal macrocells with 3 sectors per macro-cell In each sector there is one small cell cluster deployedwith shared spectrum manner The small cell cluster is agroup of densely deployed small cells We deploy 23 usersin the coverage of small cell clusters while the remainingusers are distributed in the coverage area of macro cellsThe users are uniformly distributed Moreover as mentionedabove the bandwidth resource granularity in the simulationis one RB In the initial state each user is served by the BSwhich can provide the highest downlink RSRP Once a newtraffic offloading requirement is requested the reverse auc-tion based GO scheme is triggered The detailed simulationparameters are according to 3GPP LTE-Advanced small cellHetNet evaluation methodology [24] These parameters arelisted in Table 1
The performance metrics include the system energyefficiency offloading gain and throughput In this paper themetric of offloading gain (120574gain) is defined as
120574gain =120591offloading
120591total (26)
where 120591offloading denotes the offloaded throughput and 120591totaldenotes the total system throughput The offloading gain is
Table 1 Simulation parameters
Simulation parameter ValueCarrier frequency 2GHzSystem bandwidth 10MHzTotal transmission power ofmacrocell 46 dBm
Total transmission power of small cell 30 dBmPath-loss of macrocell Pl = 283 + 220log
10(119889)
Path-loss of small cell Pl = 305 + 367log10(119889)
Small cell number per cluster 4sim10Small cell cluster number permacrocell 3
User number per macrocell 60
Antenna gain of macrocells 17 dBiAntenna gain of small cells 5 dBiTraffic model FTP Model 1Power spectrum density of thermalnoise minus174 dBmHz
User throughput threshold 5Mbps
a more straightforward notation about howmuch traffic loadcould be offloaded to improve the energy efficiency
In order to evaluate the performances of the proposedGOscheme we compare it with the TOFFR algorithm proposedin [13] and the incentivized scheme proposed in [14] whichhave been introduced in the related works The simulationresults are given as below
42 Impacts of Small Cell Numbers on Energy EfficiencyAccording to 3GPP simulation assumptions there are 23UEs distributed in the coverage of small cell cluster whilethe remainingUEs are uniformly distributed in the remainingarea of macrocells In this section the impacts of deploymentdensity of small cells in a cluster on the system energyefficiency are investigated According to 3GPP simulationmethodology the number of small cells in a cluster variesfrom 4 to 10
As shown in Figure 2 the system energy efficiency ofdifferent algorithms versus small cell numbers per cluster isdemonstrated We can observe that the system energy effi-ciency is improved with the increasing of small cell numbersThe reason lies that small cells usually can provide higherenergy efficiency than macrocells due to lower transmissionpower attenuations in hot spot deployment scenarios Whenthere exist more small cells inside one macrocell more usertraffic could be offloaded potentially to small cellsThereforehigher system energy efficiency could be further achieved
But the system energy efficiency increases slightly whenthe number of small cells is large This is because when thesmall cells are deployed more densely the intercell interfer-enceswill bemore severe indicating the requirement of largertransmission power to ensure the same user throughputMoreover from results in Figure 2 we have proved that theproposedGO scheme outperforms the TOFFR algorithm andincentivized scheme regardless of small cell density deployed
8 Mobile Information Systems
5 6 7 8 9 104Number of small cells
TOFFRwo offloading
GO schemeIncentivized scheme
5
6
7
8
9
10
11
12
13Sy
stem
ener
gy effi
cien
cy (k
bps
W)
Figure 2 System energy efficiency versus small cell numbers percluster
per cluster In the TOFFR algorithm the fractional frequencyreuse scheme is adopted to improve UE performances betterin the cell edge The UE located in the central area cannotcontribute to the energy efficiency improvement The incen-tivized scheme focuses on maximizing the offloading utilitythat purchases the available unused bandwidth in femtocellsSo the proposed GO scheme has better energy efficiencyperformance than both of them
We also compare these three offloading algorithms withthe situation of no offloading (denoted as wo offloading insimulations) as the baseline It is obvious thatwhen offloadingschemes are adopted more user traffic originally served bymacrocell will be actively offloaded to small cells There-fore all of the offloading schemes including the proposedGO scheme TOFFR algorithm and incentivized algorithmachieve higher system energy efficiency
43 Impacts of Small Cell Numbers on Offloading GainAs shown in Figure 3 the offloading gain versus differentsmall cell numbers per cluster is dipicted We can observethat the offloading gain increases with small cell numbersdue to the capacity growth with the increase of small cellnumbers per cluster Besides in order to maximize thesystem energy efficiency there is limitation on the amountof offloaded throughputs as demonstrated in Figure 3Whenthe offloading gain reaches 62 the rise against the small cellnumbers becomes rather slowThe simulation results furthersuggest that the proposed GO scheme outperforms theTOFFR algorithm and incentivized scheme in terms of notonly energy efficiency but also the offloading gain becausein TOFFR algorithm and incentivized scheme the offloadingis mainly focused on the cell edge users which limits theperformance improvements Moreover Figure 3 shows thatthe increasing of the small cell numbers of all these threeschemes will reach a plateau The reason lies that there arealways several specific users out of the coverage of the small
TOFFRGO schemeIncentivized scheme
4 5 6 7 8 9 10 113Number of small cells per cluster
0
10
20
30
40
50
60
70
Offl
oadi
ng g
ain
()
Figure 3 Offloading gain versus small cell numbers per cluster
91011121314151617181920
Smal
l cel
l clu
ster t
hrou
ghpu
t (M
bps)
5 6 7 8 9 104Number of small cells per cluster
TOFFRwo offloading
GO schemeIncentivized scheme
Figure 4 Small cell throughput versus small cell numbers percluster
cell clusters in the simulations Just as mentioned before23 UEs are deployed in the coverage of small cell clusterand the simulation results also show that the offloading gainlimitation of all three algorithms can only reach to near23 which in turn indicates the offloading limitation will bedecided by the distribution and location of users
44 Impact of Small Cell Numbers on Small Cell ThroughputIn this section the impacts on the throughput of small cellcluster versus different small cell numbers are investigatedFrom Figure 4 we can see the throughput increases with theincrease of small cell numbers per cluster The reason is thatmore users will be served by the small cells after offloadingprocesses Moreover the throughput increments grow slowly
Mobile Information Systems 9
when small cell numbers per cluster are relatively large Thisis because of the increase of intercell interferences caused bydenser small cell deployments Finally the results in Figure 4prove that the proposed GO scheme outperforms TOFFRalgorithm and incentivized algorithm also in terms of smallcell cluster throughput
5 Conclusion
This paper aims to solve the problem regarding how toperformoffloading in the small cell HetNet deployments withoptimization on maximizing the system energy efficiencyThe reverse auction theory has been implemented with theproposed GO scheme design to decide the offloading targetBS or BSs with coordination transmission enabled technol-ogy The reverse auction model is formulated by multiplesellers (BSs) and a single buyer (offloading user) with thefirst price sealed bid mechanism The BS coordination trans-missions are also supported for multiple winning biddersscenarios According to the proposed reverse auction basedGO scheme the energy efficiency optimization problemwith constraints of user guaranteed throughput thresholdbandwidth occupation and transmission power limitationis solved by Dynamic Programming method with KKTconditions The Individual Rationality and Truthfulness ofthe VCG based reverse auction model are also proved in thepaper System-level simulations have been conducted to ver-ify the effectiveness of the proposed GO scheme according to3GPP LTE-Advanced evaluation methodologies The perfor-mances when applying theGO scheme comparison schemesand the baseline without offloading situation are evaluatedwith performance metrics of energy efficiency offloadinggain and throughput The simulation results prove that theproposed GO scheme can achieve supreme performances
Notation Definition
b = b1 b2 b119894 b119897 BidsB = B
1B2 B
119895 B
119899 Bandwidth allocation results
P = P1P2 P
119895 P
119899 Power allocation results
B119895= 1198611
119895 1198612
119895 119861
119897119895
119895 119895th BS bandwidth allocation
resultsP119895= 1198751
119895 1198752
119895 119875
119897119895
119895 119895th BS power allocation
results119861119894
119895 119894th subcarrier in the 119895th BS
119875119894
119895 Transmission power on 119861119894
119895
b119899= b1 b2 b119899 Bids sent by first 119899 BSs
120578 Energy efficiency
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper is supported by the National High Technol-ogy Research and Development Program of China no2014AA01A701 Nature and Science Foundation of China
under Grants nos 61471068 and 61421061 InternationalCollaboration Project no 2015DFT10160 andNationalMajorProject no 2016ZX03001009-003
References
[1] T Nakamura S Nagata A Benjebbour et al ldquoTrends in smallcell enhancements in LTE advancedrdquo IEEE CommunicationsMagazine vol 51 no 2 pp 98ndash105 2013
[2] W Ni and I B Collings ldquoA new adaptive small-cell architec-turerdquo IEEE Journal on Selected Areas in Communications vol31 no 5 pp 829ndash839 2013
[3] L Hanzo H Haas S Imre D OrsquoBrien M Rupp and LGyongyosi ldquoWireless myths realities and futures from 3G4Gto optical and quantum wirelessrdquo Proceedings of the IEEE vol100 pp 1853ndash1888 2012
[4] TheMETIS 2020 ProjectmdashLaying the Foundation of 5G httpswwwmetis2020com
[5] ldquoSmall cell market statusrdquo White Paper Informa and Small CellForum 1 2013
[6] D Calin H Claussen and H Uzunalioglu ldquoOn femto deploy-ment architectures and macrocell offloading benefits in jointmacro-femto deploymentsrdquo IEEE Communications Magazinevol 48 no 1 pp 26ndash32 2010
[7] S-I Sou ldquoMobile data offloading with policy and chargingcontrol in 3GPP core networkrdquo IEEE Transactions on VehicularTechnology vol 62 no 7 pp 3481ndash3486 2013
[8] J Korhonen T Savolainen A Y Ding and M Kojo ldquoTowardnetwork controlled IP traffic offloadingrdquo IEEE CommunicationsMagazine vol 51 no 3 pp 96ndash102 2013
[9] H Elsawy E Hossain and S Camorlinga ldquoTraffic offloadingtechniques in two-tier femtocell networksrdquo in Proceedings of theIEEE International Conference onCommunications (ICC rsquo13) pp6086ndash6090 IEEE Budapest Hungary June 2013
[10] Z Lu P Sinha and R Srikant ldquoEasyBid enabling cellularoffloading via small playersrdquo in Proceedings of the 33rd IEEEConference on Computer Communications (IEEE INFOCOMrsquo14) pp 691ndash699 IEEE Toronto Canada May 2014
[11] M Usman A Vastberg and T Edler ldquoEnergy efficient highcapacityHETNETby offloading highQoSusers through femtordquoin Proceedings of the 17th IEEE International Conference onNetworks (ICON rsquo11) pp 19ndash24 Singapore December 2011
[12] P Chandhar and S S Das ldquoAnalytical evaluation of offloadinggain in macrocell-femtocell OFDMA networksrdquo in Proceedingsof the IEEE 77th Vehicular Technology Conference (VTC Springrsquo13) pp 1ndash6 June 2013
[13] Q Liu G Feng and S Qin ldquoEnergy-efficient traffic offloadingin Macro-Pico networksrdquo in Proceedings of the 22nd Wirelessand Optical Communications Conference (WOCC rsquo13) pp 236ndash241 IEEE Chongqing China May 2013
[14] Y Jia M Zhao K Wang and W Zhou ldquoAn incentivizedoffloading mechanism via truthful auction in heterogeneousnetworksrdquo in Proceedings of the 6th International Conference onWireless Communications and Signal Processing (WCSP rsquo14) pp1ndash6 Hefei China October 2014
[15] L Gao G Iosifidis J Huang L Tassiulas and D Li ldquoBargain-ing-based mobile data offloadingrdquo IEEE Journal on SelectedAreas in Communications vol 32 no 6 pp 1114ndash1125 2014
[16] F Zhang W Zhang and Q Ling ldquoNon-cooperative game forcapacity offloadrdquo IEEE Transactions on Wireless Communica-tions vol 11 no 4 pp 1565ndash1575 2012
10 Mobile Information Systems
[17] X Xu H Zhang X Dai and X Tao ldquoOptimal Energy EfficientOffloading in small cell HetNet with auctionrdquo in Proceedingsof the 9th International Conference on Communications andNetworking in China (CHINACOM rsquo14) pp 335ndash340MaomingChina August 2014
[18] D P Bertsekas D A Castanon and H Tsaknakis ldquoReverseauction and the solution of inequality constrained assignmentproblemsrdquo SIAM Journal on Optimization vol 3 no 2 pp 268ndash297 1993
[19] M Khaledi and A A Abouzeid ldquoDynamic spectrum sharingauction with time-evolving channel qualitiesrdquo IEEE Transac-tions onWireless Communications vol 14 no 11 pp 5900ndash59122015
[20] F Shen D Li P-H Lin and E Jorswieck ldquoAuction basedspectrum sharing for hybrid access in macro-femtocell net-works under QoS requirementsrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC rsquo15) pp3335ndash3340 IEEE London UK June 2015
[21] X Zhuo W Gao G Cao and S Hua ldquoAn incentive frameworkfor cellular traffic offloadingrdquo IEEE Transactions on MobileComputing vol 13 no 3 pp 541ndash555 2014
[22] W Vickrey ldquoCounterspeculation auctions and competitivesealed tendersrdquo The Journal of Finance vol 16 no 1 pp 8ndash371961
[23] 3GPP-TR36814 (v1110) ldquoEvolved Universal Terrestrial RadioAccess (E-UTRA) Further advancements for E-UTRA physicallayer aspectsrdquo 2013
[24] 3GPP ldquoSmall cell enhancements for E-UTRA and EUTRANphysical layer aspects (release 12)rdquo 3GPP TR 36872 2013
[25] 3GPP R1-130744 ldquoWF on evaluation assumptions for SCEphysical layerrdquo Huawei HiSilicon CATR CMCC 2013
[26] 3GPP ldquoStudy on small cell enhancements for EUTRA and E-UTRAN higher layer aspectsrdquo 3GPP TR 36842 (v1200) 2013
[27] 3GPP TS 36300 ldquoTechnical Specification Group Radio AccessNetwork Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Net-work (EUTRAN) Overall description Stage 2 (Release 12)rdquo2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
2 Mobile Information Systems
could be relieved [8] In recent years some researcheshave been conducted to refine the offloading scheme withincellular networks to alleviate the increasing traffic burdenof macrocells Authors of [9] analyze the proportion oftraffic that can be offloaded from macrocells to femtocellsto increase the system capacity The offloading gain for themacrocell and femtocell colocatedHetNets is explored in [10]Except for enhancing capacity purposes the energy efficiencymetric is also considered during offloading processes In [11]the system energy efficiency is improved through offloadingthe high Quality of Service (QoS) requirement traffic to fem-tocells But the fairness between different users is neglectedIn [12] authors verify the offloading gain in the macrocelland femtocell orthogonal frequency division multiple accesssystem by analytical evaluations Nevertheless the transmis-sion power allocation is not taken into account Benefittingfrom the information exchange between macrocells andpicocells a traffic offloading based on fractional frequencyreuse (TOFFR) algorithm is proposed to improve the systemenergy efficiency in [13] Authors in [14] propose a trafficoffloading scheme based on reverse auction mechanism andgreedy algorithms which incentivizes femtocell owners torent their underutilized spectrum for the enhancement ofnetwork performances in HetNets
From the analyses of above researches there are severalmethods adopted in offloading processes The stochasticgeometry is adopted to analyze the statistical status of theoffloading gain [9] Game theory approaches are explored byoffloading researches in [15 16] Furthermore the auctiontheory is also employed to support the offloading targetdecision and network performance improvements [10 14 1718] The forward auction model is applied in [10] for theoffloading between different operators References [14 18]employ the reverse auction model to establish the offloadingprocedures the properties of the proposed models are alsoanalyzed We also applied the forward auction model toimprove the system energy efficiency of offloading in theHetNet [17]
In economics auction is a common means to determinethe value of an itemwhich has a valuable price Nowadays theauction theory has been widely used in various fields suchas the dynamic spectrummanagement in cognitive networks[19] and hybrid access in HetNets [20] The auction model isparticularly suitable for the processes that need informationexchanges and target choosing with pricing issues Most ofthe auctionmodels employed currently are forward auctionswhich typically involve a single seller and multiple buyersThe buyers send bids to compete for the item sold by theseller The forward auction model is popular in dealing withthe decisions amongmultiple operators to getminimumcosts[10] while the reverse auction model could be used in thesituation that one buyer wants to offer reasonable prices toget service from multiple sellers [21] The offloading processis just adapted to the reverse auction applications that theoffloading user is treated as the buyer while the candidateBSs are acting as the sellersTherefore in this paper a reverseauction process with first price sealed bid mechanism isadopted in the offloading researchThe reverse auctionmodelinvolves a single buyer and multiple sellers where the buyer
makes the decision on whose bid will be accepted accordingto the bids sent by multiple sellers
For the offloading process researched in small cell Het-Nets there are two main concerns adapted to the reverseauction model
(1) The offloading user needs to select the target BSwhich is modeled as one ldquobuyerrdquo with multicandidateldquosellersrdquo The offloading user is acting as a buyer toget service from multicandidate BSs as sellers Thecandidate BSs offer their bids to compete for the serv-ing opportunity of the offloading user which couldimprove their performances such as throughputs orresourceenergy efficiency
(2) The aforementioned ldquobidrdquo offered by the candidateBSs could be their available subcarriers and trans-mission power The ldquobuyerrdquo needs to pay with theprice determined by the reverse auction for gettingthe service where the price is usually defined as theopportunity cost of the offloading target decision
In order to improve the system energy efficiency in thedense small cell HetNet deployment scenario we propose aGreen Offloading (GO) scheme through the Vickrey-Clarke-Grove (VCG) based reverse auction model [22] The smallcells will be incentivized to help offloading the traffic frommacrocells which allows the operator to express diversepreferences for densely deployed HetNets The properties ofthe reverse auction model such as the IndividualRationalityand 119879119903119906119905ℎ119891119906119897119899119890119904119904 are also proved in this paper to guaran-tee the optimality of the proposed reverse auction mecha-nism
In this paper the employed reverse auction model withfirst price sealed bidmechanism is formulated for the offload-ing process The energy efficiency optimization problemwithin the offloading process is modeled with constraints ofuser guaranteed QoS requirements The optimization prob-lem is solved by the Karush-Kuhn-Tucker (KKT) conditionsand Dynamic Programming method The system-level sim-ulation evaluations are conducted for the proposed reverseauction based GO scheme with taking current schemes [1314] as comparisons
The contributions of this paper include three aspects
(1) The reverse auction model is explored in theoffloading scheme design with the first price sealedbid mechanism which matches the mobility basedoffloading process well with limited negotiations anddelays
(2) The multiple winning bidders are also supported forBS coordination transmissions
(3) The optimization of the system energy efficiencyimprovement is solvedwith guaranteed user through-put in the offloading target decision and reallocationof both the resource block and transmission power
The rest of this paper is organized as the following InSection 2 the system model is described and the reverseauction model applied in this paper is clarified The reverse
Mobile Information Systems 3
auction based GO scheme for small cell HetNets is proposedin Section 3 In Section 4 the simulation results are presentedand discussed with comparison schemes Finally concludingremarks are drawn in Section 5
2 System Model
In this section the system model of the proposed reverseauction basedGOscheme is formulatedThenetwork deploy-ment scenario will be introduced at first and the notationdefinitions about the reverse auction model based offloadingprocess are described afterwards
21 Deployment Scenario We focus on the small cell andmacrocell overlaidHetNet deployment scenario in this paperAccording to the 3rd Generation Partnership Project (3GPP)Long Term Evolution-Advanced (LTE-A) standards due tothe scarcity of available spectrum the macrocell tier andsmall cell tier will be inevitably deployed with the sharedspectrum manner In this paper we will focus on theshared spectrum scenario with considerations of cross-tierinterferences which is more realistic Furthermore the coor-dination technologies such as the Coordinated MultipointTransmissionReception (CoMP) defined by 3GPP [23] aresupported in above small cell HetNet deployment scenarioCoMP will bring benefits to the cell edge throughput andenergy efficiency which helps the offloading users usuallylocated in the cell edge area
In order to address the dense deployment scenario forfuture heavy-traffic requirements each macrocell coveragearea is covered by several small cell clusters the numberof which will be 1 4 or 10 based on 3GPP simulationmethodology [24] Each small cell cluster includes severalsmall cell BSs the number of which will be 4 or 10 [25]Both small cell clusters and small cell BSs are considerablyless planed as opposed to the typical planned macrocelldeployments
Based on the standard discussions on the small cell Het-Net deployments in 3GPP [26] themacrocell will take chargeof the control plane for both the macrocell and colocatedsmall cells within its coverage The main responsibility of thesmall cell tier is to offload the high data-rate service fromthe macrocell while the macrocell tier handles the low data-rate traffic and high mobility users This is the typical actualdeployment demand for the HetNet The traffic model of theuser is File Transfer Protocol (FTP) Model 1 as full-buffer[23] The system energy efficiency will be the optimizationobjective during the offloading processes with constraints ofuser QoS guaranteeing
The handover solution for the offloading operation isdifferent from the standard handover criterion due to the usermobility For small cell HetNets 3GPP also discussed newhandover criterion for macro-to-small cells with Cell RangeExtension (CRE) which is defined by 3GPP standard [27]Assisted by the CRE the downlink Reference Signal ReceivedPower (RSRP) of small cells could be added with a bias Bythis means the user will be encouraged to do handover tosmall cells which will help to implement the handover foroffloading users or load balancing operations
Buyer UE
Seller BS1Seller BS2
Seller BS3
Bid
Bid
BidBids collectionresources
AuctionAllocation
Pricing
Response
Resource
Energy efficiency
Figure 1 The framework of reverse auction based GO scheme
22 Reverse Auction One typical implementation scenarioof GO scheme is given in this section Considering the usersrsquoimmense offloading potential in dense small cell HetNet areverse auction model based offloading scheme is suitableto motivate users to conduct traffic offloading that couldachieve higher system energy efficiency Figure 1 illustratesthe typical scenario of offloading process with reverse auctionmodel According to the reverse auction model the UserEquipment (UE) acts as the buyer who ensures highersystem energy efficiency in exchange of bandwidth andtransmission power resources provided by the BS to serve theoffloading UE When the UE requests data transmission itscurrent serving BS is encouraged to broadcast the offloadingrequest to all of its neighboring BSs All the BSs receivingthe request will send their bids along with their availableresources to the UErsquos current serving BS The informationincluded in each bid contains the amount of the availablebandwidth and transmission power that will be providedfor the offloading user Then the UErsquos current serving BScalculates the throughput that each bidder can provide todecide whether it could satisfy the UErsquos QoS requirement
The target BS for offloading will be the winner of thereverse auction process The conditions for judging thewinner bid could be set as the optimal objectives for theoffloading process In this paper the optimal objective of theproposed reverse auction based GO scheme is to maximizethe system energy efficiency subjected to the UErsquos minimumthroughput requirement and bidderrsquos available resourcesduring the offloading process The reverse auction based GOscheme involves two steps 119860119897119897119900119888119886119905119894119900119899 and 119875119903119894119888119894119899119892
In the119860119897119897119900119888119886119905119894119900119899 step theUErsquos current serving BS decideswhich bidders will be the auction winners As describedabove the coordination techniques among different BSs aresupported in the offloading scheme Therefore the winnermay be one BS or several BSs with the same bids
In the 119875119903119894119888119894119899119892 step the improvement of system energyefficiency is evaluated as a payment from the UE Finallythe UErsquos current serving BS sends back the auction resultto the bidders which consist of the required resources andthe expected energy efficiency incrementThen the handoverprocess for the offloading will start As shown in Figure 1
4 Mobile Information Systems
the winning bidders are BS1 and BS3 together who win thereverse auction and are designated as the offloading target BSsto serve the UE coordinately
In order to present the reverse auction based GO schemeclearer the related notation definitions are introduced asfollows
Bid (119887119894) submitted by the 119861119878
119894to convey how much
bandwidth and transmission power it can provide tothe offloading UE which may not always equal allavailable resources the BS can providePrivate Value (119909
119894) available resources in the 119861119878
119894
which is only known by the 119861119878119894itself
Pricing (119901119894) the opportunity cost of the 119894th BS which
makes sure the system energy efficiency could bemaximum
Based on the auction theory when the condition 119887119894= 119909119894
is satisfied the auction process is truthful Moreover 119887119894=
119909119894is a weakly dominant strategy [20] As a result in this
paper we set 119887119894= 119909119894 assuming all the bidders participating
in the auction process will send the bids with all availableresources it can provide which guarantees the truthfulnessof the reverse auction model
Although we have set the 119861119894119889 equal to the PrivateValuethe reverse auction basedGO schemewithmultiple candidatetarget BSs is still NP-hard problem In order to solve thisproblem an approximation algorithm is designed in the nextsection The notations used in the algorithm are introducedin the Notation Definition
23 User Delay Tolerance The reverse auction based GOscheme will start upon the receipt of offloading request at theserving BS periodically For the traditional auction processthere usually exist multiround bidding procedures to achievethe final winner This process would inevitably generate anextra delay for the BSs to wait for the auction consequenceBut for the wireless network the user traffic usually hasdelay tolerance requirements which should be consideredin the offloading scheme Therefore we implement thesingle-round auction in this paper in order to prevent theinformation exchange overhead and corresponding delay forthe offloading users Since the number of BSs surrounding aspecific UE is limited the extra delay caused by the single-round first price sealed bid auction process could be notsignificant This delay avoidance mechanism is also feasiblefor the handover based offloading schemes that usually needmore time for measurements and handover decisions
3 Reverse Auction Based GO Scheme
In this section the proposed reverse auction based GOscheme is described The main steps of the proposed GOscheme are given in Figure 1 Firstly according to theoffloading request the UErsquos current serving BS collects allthe bids from its neighboring BSs Then the serving BScalculates the throughput that all bidders can provide andderives the expected energy efficiency increment for eachcandidate BS Based on the derived throughput and energy
efficiency increment a single-round reverse auction processis performed which includes the 119860119897119897119900119888119886119905119894119900119899 and 119875119903119894119888119894119899119892steps Finally the auction results are sent back to the biddersand the user will be offloaded to the winners accordingly
31 Bidding In order to contribute to the offloading processbidders will append available resources with their bids toreveal the throughput they can provide for the offloadinguser For each bidder the upper bound of bandwidth andtransmission power that a BS can provide is119861bound and119875boundrespectively 119861bound and 119875bound can be divided into multipleunits and classified asmultiple bids b = b1 b2 b119894 b119897to indicate the resources the UE can obtain from each bidderwhere 119897 = max(lfloor119861bound119890119861rfloor lfloor119875bound119890119875rfloor) 119890119861 is the basicbandwidth unit 119890
119875is the basic transmission power unit for
the bidder The b119894 = (119861119894 119875119894) consists of both bandwidth andtransmission power resources After receiving all of the bidsUErsquos current serving BS can know how many resources eachbidder can provide with the value not larger than the sum119897
119894=1119861119894
and sum119897119894=1119875119894 The scale of the bandwidth and transmission
power unit can be flexibly set by the system The smallerunit definition leads to more information in the bid whichimproves the performance of the auction process But it willalso generate more computational costs and increase thecomplexity In this paper one Resource Block (RB) and 01Wtransmission power are chosen as the basic bandwidth unitand transmission power unit respectively We choose 01Was the basic transmission power here because the DynamicProgramming will be adopted in the next subsection to solvethe optimization problem where the integer data are neededby the Dynamic Programming method
32 Reserve Auction Algorithm As mentioned above thereverse auction process includes two steps of the 119860119897119897119900119888119886119905119894119900119899and 119875119903119894119888119894119899119892
321 Allocation In traditional reverse auction processes theallocation result is completely decided by the bids that isthe bidders who offer the largest supply of resources will winthe auction However in this paper besides the resourcesthat the bidders can provide the energy efficiency achievedby the bidder should also be considered Assume that B =B1B2 B
119895 B
119899 and P = P
1P2 P
119895 P
119899
represent the allocation result where B119895= 1198611
119895 1198612
119895 119861
119897119895
119895
is the RB that 119895th BS could provide and 119861119894119895= 0 if the 119894th RB
in the 119895th BS is not needed P119895= 1198751
119895 1198752
119895 119875
119897119895
119895 denotes the
transmission power that the 119895th BS could transmit and 119875119894119895is
the transmission power on 119861119894119895 If B119895or P119895equals zero the 119895th
BS loses in this auction processThe 119860119897119897119900119888119886119905119894119900119899 problem is formulated as
maxB119895 P119895
119862system
119875119905
(1)
st119899
sum
119895=1
119897119895
sum
119894=1
119861119894
119895log(1 +
119875119894
119895119867119895
119868119894
119895+ 1198730119861119894
119895
) ge 120591119903forall119903 (2)
Mobile Information Systems 5
119861119894
119895isin 0 119861 (3)
119875119894
119895isin P119898P119904 (4)
P119898=
119875119894
119895| 0 le 119875
119894
119895le 119875119898
119897119895
sum
119894=1
119875119894
119895le 119875119898
(5)
P119904=
119875119894
119895| 0 le 119875
119894
119895le 119875119904
119897119895
sum
119894=1
119875119894
119895le 119875119904
(6)
In (1) 119862system = sum119899
119895=1119862119895sum119897119895
119894=1119861119894
119895denotes the throughput
119875119905= sum119897119895
119894=1119875119894
119895sum119899
119895=1119897119895denotes the expectation of transmission
power on each RB from all the bidders 119862system119875119905 is thesystem energy efficiency and 119875
119905gt 0
In (2) 119867119895= 119889minus120572119895
119895|ℎ119895|2120594119895is the channel gain between
the 119895th BS and UE where 119889119895is the distance between the
119895th BS and UE 120572119895is the path-loss exponent of the 119895th
BS ℎ119895is the Rayleigh fading component 120594
119895denotes the
log-normally distributed shadow fading Furthermore 119868119894119895=
sum119899
1198951015840=11198951015840=119895119875119894
11989510158401198671198951015840V119894
1198951015840 is the interference experienced by the UE
on the 119894th RB where 1198751198941198951015840 is the transmission power on the 119894th
RB from the 1198951015840th BS1198671198951015840 is the channel gain between the 1198951015840th
BS andUE V1198941198951015840 is the binary variables representing the activity
factor of 1198951015840th BS 1198730is the thermal noise level and 120591
119903is the
guaranteed throughput threshold of the offloading UEConstraint (3) means the RB 119861119894
119895can be occupied or
vacant Constraint (4) denotes the bidder can be a macrocellBS or a small cell BS In (5) and (6) we give the requirementfor the transmission power in the corresponding RB where119875119898and 119875119904are the power limitations for the macrocell BS and
small cell BS respectivelyFor the convenience of solving this problem we trans-
form (1) (2) and (3) into the following form
minB119895P119895
119875119905
119862system
119862system ge 120591119903
0 le 119861119894
119895le 119861
(7)
Assuming the 119862system is an independent variable theKarush-Kuhn-Tucker (KKT) conditions are given as
nablaB119895P119895system119875119905
119862systemminus nablaB119895P119895system (119862system minus 120591119903) 120583
minus nablaB119895 P119895system119861119894
119895120590119894
119895minus nablaB119895P119895system (119861 minus 119861
119894
119895) 120585119894
119895
minus nablaB119895 P119895system119875119894
119895]119894119895minus nablaB119895P119895system (119875119904 (119875119898) minus 119875
119894
119895) 120582119894
119895
minus nablaB119895 P119895system(119875119904 (119875119898) minus
119897119895
sum
119894=1
119875119894
119895)120588119895= 0
(8)
(119862system minus 120591119903) 120583 = 0 (9)
119861119894
119895120590119894
119895= 0 119861
119894
119895ge 0 120590
119894
119895ge 0 (10)
(119861 minus 119861119894
119895) 120585119894
119895= 0 119861 le 119861
119894
119895 120585119894
119895ge 0 (11)
119875119894
119895]119894119895= 0 119875
119894
119895ge 0 ]119894
119895ge 0 (12)
(119875119904(119875119898) minus 119875119894
119895) 120582119894
119895= 0 119875
119894
119895le 119875119904(119875119898) 120582119894
119895ge 0 (13)
(119875119904(119875119898) minus
119897119895
sum
119894=1
119875119894
119895)120588119895= 0
119897119895
sum
119894=1
119875119894
119895le 119875119904(119875119898) 120588119895ge 0 (14)
Equation (8) is changed into (15) by taking the derivativeof 119862system as
minus119875119905
1198622systemminus 120583 = 0 (15)
119875119905
1198622system+ 120583 = 0 (16)
Thismakes sensewhere119875119905and1198622system are nonzero values
Thus 120583 = 0 (9) shows (119862system minus 120591119903)120583 = 0 where
119862system = 120591119903 (17)
When the offloading process is triggered the offloadeduser traffic requirements and the number of BSs participatingin the auction are known The KKT conditions prove that(9) is guaranteed throughput requirement The 119860119897119897119900119888119886119905119894119900119899problem is transformed to solve minB119895P119895119875119905 which can be
simplified into a linear objective functionminsum119897119895119894=1119875119894
119895sum119899
119895=1119897119895
Because the above119860119897119897119900119888119886119905119894119900119899problem is a linear problemit is easy to find that this is a multiple knapsack problemIn order to facilitate the solving process we turn it intoa 0-1 knapsack problem There are 119899 BSs to participate inthe auction the 119894th BS resources are separated into 119872
119894
independent piece of ldquogoodsrdquo that can be loaded into aknapsack This will get a 0-1 knapsack problem where thenumber of items is sum119872
119894 If we solve this problem directly
the computational complexity will be 119874(119862sum119872119894) In order to
reduce the complexity we design another algorithm with thespecific plan as follows
As mentioned above the 119894th BS bandwidth resourceis composed of a number of RB groups Considering thecondition of the binary the guaranteeing of selecting anymultiple resource package strategy still can be achievedafter the transformation of the original multiple knapsackproblem One BS which has119872
119894resource blocks is separated
into several RB groups where these RB groups respectivelyhave 1 2 22 23 119872
119894minus 2119896minus1+ 1 resource blocks The 119894th BS
has Ceiling(log119872119894) different RB groups participating in the
0-1 knapsack problemInitialize the value as follows the weight of knapsack is
119862 which is the offloading user traffic requirements 120591119903 The
starting value 119865[0 119888] = 0 119865[119909 0] = 0 And the 119894th BS hasCeiling(log119872
119894) stages
6 Mobile Information Systems
(1) Let 1205780= 0
(2) for 119895 = 1 to 119899 do(3) if Bidder b119895 is a macrocell BS then(4) for 119875 = 1 to 10119875
119898do
(5) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(6) end for(7) for 119894 = 1 to 119897
119895do
(8) if 119875119894119895gt 0 then
(9) 119861119894
119895= 119861
(10) else(11) 119861
119894
119895= 0
(12) end if(13) end for(14) else if Bidder b119895 is a small cell BS then(15) for 119875 = 1 to 10119875
119904do
(16) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(17) end for(18) for 119894 = 1 to 119897
119895do
(19) if 119875119894119895gt 0 then
(20) 119861119894
119895= 119861
(21) else(22) 119861
119894
119895= 0
(23) end if(24) end for(25) end if(26) end for
Algorithm 1 Reverse auction 119860119897119897119900119888119886119905119894119900119899 (119899 b119899)
Renumber all RB groups 119888119909is the capacity of the 119909
stage which denotes the 119894th RB group of the 119895th BS and thecorresponding value is V
119909
The capacity of the 119909 stage is
119888119909= 119908 [119894 119895] = 119861
119894
119895log(1 +
119875119894
119895119867119895
119868119894
119895+ 1198730119861119894
119895
) (18)
The value of the corresponding V119909is
V119909= V [119894 119895] = 119875119894
119895 (19)
The iterative equation will be
119865 [119909 119888] = min 119865 [119909 minus 1 119888] 119865 [119909 minus 1 119888 minus 119888119909] + V119909 (20)
In (20) 119865[119909 119888] is the minimum value of the transmissionpower in the stage 119909 119888 denotes the remaining space ofthe pack according to the current stage 119888
119909denotes the
provided capacity when choosing the 119909 stage V119909denotes the
provided transmission power when choosing the 119909 stageThecomputational complexity is reduced to 119874(119862sum log119872
119894)
The proposed 119860119897119897119900119888119886119905119894119900119899 algorithm is illustrated inAlgorithm 1 with B = B
1B2 B
119899 and P = P
1P2
P119899In Algorithm 1 the transmission power of each bidder is
chosen firstly and the corresponding bandwidth is decidedbased on the transmission power allocation results As
for 119895 = 1 to 119899 do(2) if 119895th BS is a winning bidder then
Reverse Auction - 119860119897119897119900119888119886119905119894119900119899 (119899 119895B b119894)(4) 119901
119895= 120578B119899b119894 minus (120578B119899 minus 120578b119894 )
else(6) 119901
119895= 0
end if(8) end for
Algorithm 2 Reverse auction 119875119903119894119888119894119899119892 (119899 b119899BP 120578)
mentioned before 01W is chosen as the basic transmissionpower unit and the integer data is needed in the DynamicProgramming Therefore the range of 119875 is defined from 1
to 10119875119898or 10119875
119904 The equation in Line (5) and Line (16) of
Algorithm 1 mean that the 119895th bidderrsquos transmission powershould be chosen to achieve the largest 120578b119895 and guarantee theoffloaded userrsquos throughput threshold 120591
119903at the same time
where 120578b119895 denotes the energy efficiency of the 119895th bidderAfter implementing the Dynamic Programming the resultsof Algorithm 1 P10 and B will be the optimal allocationsolution for the proposed reverse auction process
322 Pricing In traditional 119875119903119894119888119894119899119892 algorithms the biddersare encouraged to set their own bids truthfully as illustratedbefore So in this paper the same energy efficiency that thecorresponding bidder achieves is paid back With regard tothe offloading user throughput threshold 120591
119903 we define 120578
1and
1205782as (21) and (22) as follows
1205781= 120578B119899b119894 = max
B119895B119894P119895P119894
119862system
119864 (119875119905)
(21)
1205782= 120578B119899 minus 120578b119894 = (max
B119895P119895
119862system
119864 (119875119905)) minus
119862b119894
119864 (P119894) (22)
where 1205781denotes the system energy efficiency under the
optimal 119860119897119897119900119888119886119905119894119900119899 solution without the presence of the 119894thBS The 120578
2denotes the system energy efficiency except for
the 119894th BS under current optimal119860119897119897119900119888119886119905119894119900119899 resultsThen theopportunity cost of the 119894th BS is defined as the differencebetween 120578
1and 1205782 just as illustrated in (23) [19] as follows
119901119894= 1205781minus 1205782= 120578B119899b119894 minus (120578B119899 minus 120578b119894) (23)
The 119875119903119894119888119894119899119892 algorithm is given as Algorithm 2
323 Properties In this section the properties of theproposed reverse auction model are analyzed Accordingto the VCG based reverse auction model the IndividualRa-tionality and the 119879119903119906119905ℎ119891119906119897119899119890119904119904 properties need to be proved
IndividualRationality When the utility of each participatingbidder in the119875119903119894119888119894119899119892 stage is greater than zero this algorithmis individual rational for each winning bidder Namely
119901119894= 120578B119899b119894 minus (120578B119899 minus 120578b119894) ge 0 (24)
Mobile Information Systems 7
119879119903119906119905ℎ119891119906119897119899119890119904119904 For each bidder the Truthfulness means thateach bidderrsquos bid price is equal to its private value This isa weakly dominant strategy If BSrsquos bidding is untrue theenergy efficiency will be unlikely the biggest In order toget the maximum energy efficiency the allocation should beformulated as follows
119901119895= 120578B119899b119895 minus (120578B119899 minus 120578b119895)
120575 = 119901119895minus 119901119894
= 120578B119899b119895 minus (120578B119899 minus 120578b119895) minus [120578B119899b119894 minus (120578B119899 minus 120578b119894)]
= 120578B119899b119895 minus 120578B119899 + 120578b119895 minus 120578B119899b119894 + 120578B119899 minus 120578b119894
= 120578B119899b119895 + 120578b119895 minus 120578B119899b119894 minus 120578b119894
= (120578B119899b119895 + 120578b119895) minus (120578B119899b119894 + 120578b119894)
(25)
Based on the proposed model in this paper because 119901119895le
119901119894and120575 le 0 thismeans 120578B119899b119895+120578b119895 le 120578B119899b119894+120578b119894 If and only
if 119895 = 119894 it can take the equal signTherefore each bidder mustbe truthful to obtain the maximum system energy efficiencyThe proof is finished
4 Performance Evaluation
In this section we built the system-level simulation plat-form according to the 3GPP LTE-Advanced simulationmethodology [23] Based on this platform we validate theperformances of the proposed reverse auction based GOscheme with comparison algorithms in the small cell HetNetdownlink scenario
41 Simulation Setting Performance Metrics and ComparisonAlgorithms Theconsidered simulation scenario in this papercomprises 19-hexagonal macrocells with 3 sectors per macro-cell In each sector there is one small cell cluster deployedwith shared spectrum manner The small cell cluster is agroup of densely deployed small cells We deploy 23 usersin the coverage of small cell clusters while the remainingusers are distributed in the coverage area of macro cellsThe users are uniformly distributed Moreover as mentionedabove the bandwidth resource granularity in the simulationis one RB In the initial state each user is served by the BSwhich can provide the highest downlink RSRP Once a newtraffic offloading requirement is requested the reverse auc-tion based GO scheme is triggered The detailed simulationparameters are according to 3GPP LTE-Advanced small cellHetNet evaluation methodology [24] These parameters arelisted in Table 1
The performance metrics include the system energyefficiency offloading gain and throughput In this paper themetric of offloading gain (120574gain) is defined as
120574gain =120591offloading
120591total (26)
where 120591offloading denotes the offloaded throughput and 120591totaldenotes the total system throughput The offloading gain is
Table 1 Simulation parameters
Simulation parameter ValueCarrier frequency 2GHzSystem bandwidth 10MHzTotal transmission power ofmacrocell 46 dBm
Total transmission power of small cell 30 dBmPath-loss of macrocell Pl = 283 + 220log
10(119889)
Path-loss of small cell Pl = 305 + 367log10(119889)
Small cell number per cluster 4sim10Small cell cluster number permacrocell 3
User number per macrocell 60
Antenna gain of macrocells 17 dBiAntenna gain of small cells 5 dBiTraffic model FTP Model 1Power spectrum density of thermalnoise minus174 dBmHz
User throughput threshold 5Mbps
a more straightforward notation about howmuch traffic loadcould be offloaded to improve the energy efficiency
In order to evaluate the performances of the proposedGOscheme we compare it with the TOFFR algorithm proposedin [13] and the incentivized scheme proposed in [14] whichhave been introduced in the related works The simulationresults are given as below
42 Impacts of Small Cell Numbers on Energy EfficiencyAccording to 3GPP simulation assumptions there are 23UEs distributed in the coverage of small cell cluster whilethe remainingUEs are uniformly distributed in the remainingarea of macrocells In this section the impacts of deploymentdensity of small cells in a cluster on the system energyefficiency are investigated According to 3GPP simulationmethodology the number of small cells in a cluster variesfrom 4 to 10
As shown in Figure 2 the system energy efficiency ofdifferent algorithms versus small cell numbers per cluster isdemonstrated We can observe that the system energy effi-ciency is improved with the increasing of small cell numbersThe reason lies that small cells usually can provide higherenergy efficiency than macrocells due to lower transmissionpower attenuations in hot spot deployment scenarios Whenthere exist more small cells inside one macrocell more usertraffic could be offloaded potentially to small cellsThereforehigher system energy efficiency could be further achieved
But the system energy efficiency increases slightly whenthe number of small cells is large This is because when thesmall cells are deployed more densely the intercell interfer-enceswill bemore severe indicating the requirement of largertransmission power to ensure the same user throughputMoreover from results in Figure 2 we have proved that theproposedGO scheme outperforms the TOFFR algorithm andincentivized scheme regardless of small cell density deployed
8 Mobile Information Systems
5 6 7 8 9 104Number of small cells
TOFFRwo offloading
GO schemeIncentivized scheme
5
6
7
8
9
10
11
12
13Sy
stem
ener
gy effi
cien
cy (k
bps
W)
Figure 2 System energy efficiency versus small cell numbers percluster
per cluster In the TOFFR algorithm the fractional frequencyreuse scheme is adopted to improve UE performances betterin the cell edge The UE located in the central area cannotcontribute to the energy efficiency improvement The incen-tivized scheme focuses on maximizing the offloading utilitythat purchases the available unused bandwidth in femtocellsSo the proposed GO scheme has better energy efficiencyperformance than both of them
We also compare these three offloading algorithms withthe situation of no offloading (denoted as wo offloading insimulations) as the baseline It is obvious thatwhen offloadingschemes are adopted more user traffic originally served bymacrocell will be actively offloaded to small cells There-fore all of the offloading schemes including the proposedGO scheme TOFFR algorithm and incentivized algorithmachieve higher system energy efficiency
43 Impacts of Small Cell Numbers on Offloading GainAs shown in Figure 3 the offloading gain versus differentsmall cell numbers per cluster is dipicted We can observethat the offloading gain increases with small cell numbersdue to the capacity growth with the increase of small cellnumbers per cluster Besides in order to maximize thesystem energy efficiency there is limitation on the amountof offloaded throughputs as demonstrated in Figure 3Whenthe offloading gain reaches 62 the rise against the small cellnumbers becomes rather slowThe simulation results furthersuggest that the proposed GO scheme outperforms theTOFFR algorithm and incentivized scheme in terms of notonly energy efficiency but also the offloading gain becausein TOFFR algorithm and incentivized scheme the offloadingis mainly focused on the cell edge users which limits theperformance improvements Moreover Figure 3 shows thatthe increasing of the small cell numbers of all these threeschemes will reach a plateau The reason lies that there arealways several specific users out of the coverage of the small
TOFFRGO schemeIncentivized scheme
4 5 6 7 8 9 10 113Number of small cells per cluster
0
10
20
30
40
50
60
70
Offl
oadi
ng g
ain
()
Figure 3 Offloading gain versus small cell numbers per cluster
91011121314151617181920
Smal
l cel
l clu
ster t
hrou
ghpu
t (M
bps)
5 6 7 8 9 104Number of small cells per cluster
TOFFRwo offloading
GO schemeIncentivized scheme
Figure 4 Small cell throughput versus small cell numbers percluster
cell clusters in the simulations Just as mentioned before23 UEs are deployed in the coverage of small cell clusterand the simulation results also show that the offloading gainlimitation of all three algorithms can only reach to near23 which in turn indicates the offloading limitation will bedecided by the distribution and location of users
44 Impact of Small Cell Numbers on Small Cell ThroughputIn this section the impacts on the throughput of small cellcluster versus different small cell numbers are investigatedFrom Figure 4 we can see the throughput increases with theincrease of small cell numbers per cluster The reason is thatmore users will be served by the small cells after offloadingprocesses Moreover the throughput increments grow slowly
Mobile Information Systems 9
when small cell numbers per cluster are relatively large Thisis because of the increase of intercell interferences caused bydenser small cell deployments Finally the results in Figure 4prove that the proposed GO scheme outperforms TOFFRalgorithm and incentivized algorithm also in terms of smallcell cluster throughput
5 Conclusion
This paper aims to solve the problem regarding how toperformoffloading in the small cell HetNet deployments withoptimization on maximizing the system energy efficiencyThe reverse auction theory has been implemented with theproposed GO scheme design to decide the offloading targetBS or BSs with coordination transmission enabled technol-ogy The reverse auction model is formulated by multiplesellers (BSs) and a single buyer (offloading user) with thefirst price sealed bid mechanism The BS coordination trans-missions are also supported for multiple winning biddersscenarios According to the proposed reverse auction basedGO scheme the energy efficiency optimization problemwith constraints of user guaranteed throughput thresholdbandwidth occupation and transmission power limitationis solved by Dynamic Programming method with KKTconditions The Individual Rationality and Truthfulness ofthe VCG based reverse auction model are also proved in thepaper System-level simulations have been conducted to ver-ify the effectiveness of the proposed GO scheme according to3GPP LTE-Advanced evaluation methodologies The perfor-mances when applying theGO scheme comparison schemesand the baseline without offloading situation are evaluatedwith performance metrics of energy efficiency offloadinggain and throughput The simulation results prove that theproposed GO scheme can achieve supreme performances
Notation Definition
b = b1 b2 b119894 b119897 BidsB = B
1B2 B
119895 B
119899 Bandwidth allocation results
P = P1P2 P
119895 P
119899 Power allocation results
B119895= 1198611
119895 1198612
119895 119861
119897119895
119895 119895th BS bandwidth allocation
resultsP119895= 1198751
119895 1198752
119895 119875
119897119895
119895 119895th BS power allocation
results119861119894
119895 119894th subcarrier in the 119895th BS
119875119894
119895 Transmission power on 119861119894
119895
b119899= b1 b2 b119899 Bids sent by first 119899 BSs
120578 Energy efficiency
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper is supported by the National High Technol-ogy Research and Development Program of China no2014AA01A701 Nature and Science Foundation of China
under Grants nos 61471068 and 61421061 InternationalCollaboration Project no 2015DFT10160 andNationalMajorProject no 2016ZX03001009-003
References
[1] T Nakamura S Nagata A Benjebbour et al ldquoTrends in smallcell enhancements in LTE advancedrdquo IEEE CommunicationsMagazine vol 51 no 2 pp 98ndash105 2013
[2] W Ni and I B Collings ldquoA new adaptive small-cell architec-turerdquo IEEE Journal on Selected Areas in Communications vol31 no 5 pp 829ndash839 2013
[3] L Hanzo H Haas S Imre D OrsquoBrien M Rupp and LGyongyosi ldquoWireless myths realities and futures from 3G4Gto optical and quantum wirelessrdquo Proceedings of the IEEE vol100 pp 1853ndash1888 2012
[4] TheMETIS 2020 ProjectmdashLaying the Foundation of 5G httpswwwmetis2020com
[5] ldquoSmall cell market statusrdquo White Paper Informa and Small CellForum 1 2013
[6] D Calin H Claussen and H Uzunalioglu ldquoOn femto deploy-ment architectures and macrocell offloading benefits in jointmacro-femto deploymentsrdquo IEEE Communications Magazinevol 48 no 1 pp 26ndash32 2010
[7] S-I Sou ldquoMobile data offloading with policy and chargingcontrol in 3GPP core networkrdquo IEEE Transactions on VehicularTechnology vol 62 no 7 pp 3481ndash3486 2013
[8] J Korhonen T Savolainen A Y Ding and M Kojo ldquoTowardnetwork controlled IP traffic offloadingrdquo IEEE CommunicationsMagazine vol 51 no 3 pp 96ndash102 2013
[9] H Elsawy E Hossain and S Camorlinga ldquoTraffic offloadingtechniques in two-tier femtocell networksrdquo in Proceedings of theIEEE International Conference onCommunications (ICC rsquo13) pp6086ndash6090 IEEE Budapest Hungary June 2013
[10] Z Lu P Sinha and R Srikant ldquoEasyBid enabling cellularoffloading via small playersrdquo in Proceedings of the 33rd IEEEConference on Computer Communications (IEEE INFOCOMrsquo14) pp 691ndash699 IEEE Toronto Canada May 2014
[11] M Usman A Vastberg and T Edler ldquoEnergy efficient highcapacityHETNETby offloading highQoSusers through femtordquoin Proceedings of the 17th IEEE International Conference onNetworks (ICON rsquo11) pp 19ndash24 Singapore December 2011
[12] P Chandhar and S S Das ldquoAnalytical evaluation of offloadinggain in macrocell-femtocell OFDMA networksrdquo in Proceedingsof the IEEE 77th Vehicular Technology Conference (VTC Springrsquo13) pp 1ndash6 June 2013
[13] Q Liu G Feng and S Qin ldquoEnergy-efficient traffic offloadingin Macro-Pico networksrdquo in Proceedings of the 22nd Wirelessand Optical Communications Conference (WOCC rsquo13) pp 236ndash241 IEEE Chongqing China May 2013
[14] Y Jia M Zhao K Wang and W Zhou ldquoAn incentivizedoffloading mechanism via truthful auction in heterogeneousnetworksrdquo in Proceedings of the 6th International Conference onWireless Communications and Signal Processing (WCSP rsquo14) pp1ndash6 Hefei China October 2014
[15] L Gao G Iosifidis J Huang L Tassiulas and D Li ldquoBargain-ing-based mobile data offloadingrdquo IEEE Journal on SelectedAreas in Communications vol 32 no 6 pp 1114ndash1125 2014
[16] F Zhang W Zhang and Q Ling ldquoNon-cooperative game forcapacity offloadrdquo IEEE Transactions on Wireless Communica-tions vol 11 no 4 pp 1565ndash1575 2012
10 Mobile Information Systems
[17] X Xu H Zhang X Dai and X Tao ldquoOptimal Energy EfficientOffloading in small cell HetNet with auctionrdquo in Proceedingsof the 9th International Conference on Communications andNetworking in China (CHINACOM rsquo14) pp 335ndash340MaomingChina August 2014
[18] D P Bertsekas D A Castanon and H Tsaknakis ldquoReverseauction and the solution of inequality constrained assignmentproblemsrdquo SIAM Journal on Optimization vol 3 no 2 pp 268ndash297 1993
[19] M Khaledi and A A Abouzeid ldquoDynamic spectrum sharingauction with time-evolving channel qualitiesrdquo IEEE Transac-tions onWireless Communications vol 14 no 11 pp 5900ndash59122015
[20] F Shen D Li P-H Lin and E Jorswieck ldquoAuction basedspectrum sharing for hybrid access in macro-femtocell net-works under QoS requirementsrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC rsquo15) pp3335ndash3340 IEEE London UK June 2015
[21] X Zhuo W Gao G Cao and S Hua ldquoAn incentive frameworkfor cellular traffic offloadingrdquo IEEE Transactions on MobileComputing vol 13 no 3 pp 541ndash555 2014
[22] W Vickrey ldquoCounterspeculation auctions and competitivesealed tendersrdquo The Journal of Finance vol 16 no 1 pp 8ndash371961
[23] 3GPP-TR36814 (v1110) ldquoEvolved Universal Terrestrial RadioAccess (E-UTRA) Further advancements for E-UTRA physicallayer aspectsrdquo 2013
[24] 3GPP ldquoSmall cell enhancements for E-UTRA and EUTRANphysical layer aspects (release 12)rdquo 3GPP TR 36872 2013
[25] 3GPP R1-130744 ldquoWF on evaluation assumptions for SCEphysical layerrdquo Huawei HiSilicon CATR CMCC 2013
[26] 3GPP ldquoStudy on small cell enhancements for EUTRA and E-UTRAN higher layer aspectsrdquo 3GPP TR 36842 (v1200) 2013
[27] 3GPP TS 36300 ldquoTechnical Specification Group Radio AccessNetwork Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Net-work (EUTRAN) Overall description Stage 2 (Release 12)rdquo2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mobile Information Systems 3
auction based GO scheme for small cell HetNets is proposedin Section 3 In Section 4 the simulation results are presentedand discussed with comparison schemes Finally concludingremarks are drawn in Section 5
2 System Model
In this section the system model of the proposed reverseauction basedGOscheme is formulatedThenetwork deploy-ment scenario will be introduced at first and the notationdefinitions about the reverse auction model based offloadingprocess are described afterwards
21 Deployment Scenario We focus on the small cell andmacrocell overlaidHetNet deployment scenario in this paperAccording to the 3rd Generation Partnership Project (3GPP)Long Term Evolution-Advanced (LTE-A) standards due tothe scarcity of available spectrum the macrocell tier andsmall cell tier will be inevitably deployed with the sharedspectrum manner In this paper we will focus on theshared spectrum scenario with considerations of cross-tierinterferences which is more realistic Furthermore the coor-dination technologies such as the Coordinated MultipointTransmissionReception (CoMP) defined by 3GPP [23] aresupported in above small cell HetNet deployment scenarioCoMP will bring benefits to the cell edge throughput andenergy efficiency which helps the offloading users usuallylocated in the cell edge area
In order to address the dense deployment scenario forfuture heavy-traffic requirements each macrocell coveragearea is covered by several small cell clusters the numberof which will be 1 4 or 10 based on 3GPP simulationmethodology [24] Each small cell cluster includes severalsmall cell BSs the number of which will be 4 or 10 [25]Both small cell clusters and small cell BSs are considerablyless planed as opposed to the typical planned macrocelldeployments
Based on the standard discussions on the small cell Het-Net deployments in 3GPP [26] themacrocell will take chargeof the control plane for both the macrocell and colocatedsmall cells within its coverage The main responsibility of thesmall cell tier is to offload the high data-rate service fromthe macrocell while the macrocell tier handles the low data-rate traffic and high mobility users This is the typical actualdeployment demand for the HetNet The traffic model of theuser is File Transfer Protocol (FTP) Model 1 as full-buffer[23] The system energy efficiency will be the optimizationobjective during the offloading processes with constraints ofuser QoS guaranteeing
The handover solution for the offloading operation isdifferent from the standard handover criterion due to the usermobility For small cell HetNets 3GPP also discussed newhandover criterion for macro-to-small cells with Cell RangeExtension (CRE) which is defined by 3GPP standard [27]Assisted by the CRE the downlink Reference Signal ReceivedPower (RSRP) of small cells could be added with a bias Bythis means the user will be encouraged to do handover tosmall cells which will help to implement the handover foroffloading users or load balancing operations
Buyer UE
Seller BS1Seller BS2
Seller BS3
Bid
Bid
BidBids collectionresources
AuctionAllocation
Pricing
Response
Resource
Energy efficiency
Figure 1 The framework of reverse auction based GO scheme
22 Reverse Auction One typical implementation scenarioof GO scheme is given in this section Considering the usersrsquoimmense offloading potential in dense small cell HetNet areverse auction model based offloading scheme is suitableto motivate users to conduct traffic offloading that couldachieve higher system energy efficiency Figure 1 illustratesthe typical scenario of offloading process with reverse auctionmodel According to the reverse auction model the UserEquipment (UE) acts as the buyer who ensures highersystem energy efficiency in exchange of bandwidth andtransmission power resources provided by the BS to serve theoffloading UE When the UE requests data transmission itscurrent serving BS is encouraged to broadcast the offloadingrequest to all of its neighboring BSs All the BSs receivingthe request will send their bids along with their availableresources to the UErsquos current serving BS The informationincluded in each bid contains the amount of the availablebandwidth and transmission power that will be providedfor the offloading user Then the UErsquos current serving BScalculates the throughput that each bidder can provide todecide whether it could satisfy the UErsquos QoS requirement
The target BS for offloading will be the winner of thereverse auction process The conditions for judging thewinner bid could be set as the optimal objectives for theoffloading process In this paper the optimal objective of theproposed reverse auction based GO scheme is to maximizethe system energy efficiency subjected to the UErsquos minimumthroughput requirement and bidderrsquos available resourcesduring the offloading process The reverse auction based GOscheme involves two steps 119860119897119897119900119888119886119905119894119900119899 and 119875119903119894119888119894119899119892
In the119860119897119897119900119888119886119905119894119900119899 step theUErsquos current serving BS decideswhich bidders will be the auction winners As describedabove the coordination techniques among different BSs aresupported in the offloading scheme Therefore the winnermay be one BS or several BSs with the same bids
In the 119875119903119894119888119894119899119892 step the improvement of system energyefficiency is evaluated as a payment from the UE Finallythe UErsquos current serving BS sends back the auction resultto the bidders which consist of the required resources andthe expected energy efficiency incrementThen the handoverprocess for the offloading will start As shown in Figure 1
4 Mobile Information Systems
the winning bidders are BS1 and BS3 together who win thereverse auction and are designated as the offloading target BSsto serve the UE coordinately
In order to present the reverse auction based GO schemeclearer the related notation definitions are introduced asfollows
Bid (119887119894) submitted by the 119861119878
119894to convey how much
bandwidth and transmission power it can provide tothe offloading UE which may not always equal allavailable resources the BS can providePrivate Value (119909
119894) available resources in the 119861119878
119894
which is only known by the 119861119878119894itself
Pricing (119901119894) the opportunity cost of the 119894th BS which
makes sure the system energy efficiency could bemaximum
Based on the auction theory when the condition 119887119894= 119909119894
is satisfied the auction process is truthful Moreover 119887119894=
119909119894is a weakly dominant strategy [20] As a result in this
paper we set 119887119894= 119909119894 assuming all the bidders participating
in the auction process will send the bids with all availableresources it can provide which guarantees the truthfulnessof the reverse auction model
Although we have set the 119861119894119889 equal to the PrivateValuethe reverse auction basedGO schemewithmultiple candidatetarget BSs is still NP-hard problem In order to solve thisproblem an approximation algorithm is designed in the nextsection The notations used in the algorithm are introducedin the Notation Definition
23 User Delay Tolerance The reverse auction based GOscheme will start upon the receipt of offloading request at theserving BS periodically For the traditional auction processthere usually exist multiround bidding procedures to achievethe final winner This process would inevitably generate anextra delay for the BSs to wait for the auction consequenceBut for the wireless network the user traffic usually hasdelay tolerance requirements which should be consideredin the offloading scheme Therefore we implement thesingle-round auction in this paper in order to prevent theinformation exchange overhead and corresponding delay forthe offloading users Since the number of BSs surrounding aspecific UE is limited the extra delay caused by the single-round first price sealed bid auction process could be notsignificant This delay avoidance mechanism is also feasiblefor the handover based offloading schemes that usually needmore time for measurements and handover decisions
3 Reverse Auction Based GO Scheme
In this section the proposed reverse auction based GOscheme is described The main steps of the proposed GOscheme are given in Figure 1 Firstly according to theoffloading request the UErsquos current serving BS collects allthe bids from its neighboring BSs Then the serving BScalculates the throughput that all bidders can provide andderives the expected energy efficiency increment for eachcandidate BS Based on the derived throughput and energy
efficiency increment a single-round reverse auction processis performed which includes the 119860119897119897119900119888119886119905119894119900119899 and 119875119903119894119888119894119899119892steps Finally the auction results are sent back to the biddersand the user will be offloaded to the winners accordingly
31 Bidding In order to contribute to the offloading processbidders will append available resources with their bids toreveal the throughput they can provide for the offloadinguser For each bidder the upper bound of bandwidth andtransmission power that a BS can provide is119861bound and119875boundrespectively 119861bound and 119875bound can be divided into multipleunits and classified asmultiple bids b = b1 b2 b119894 b119897to indicate the resources the UE can obtain from each bidderwhere 119897 = max(lfloor119861bound119890119861rfloor lfloor119875bound119890119875rfloor) 119890119861 is the basicbandwidth unit 119890
119875is the basic transmission power unit for
the bidder The b119894 = (119861119894 119875119894) consists of both bandwidth andtransmission power resources After receiving all of the bidsUErsquos current serving BS can know how many resources eachbidder can provide with the value not larger than the sum119897
119894=1119861119894
and sum119897119894=1119875119894 The scale of the bandwidth and transmission
power unit can be flexibly set by the system The smallerunit definition leads to more information in the bid whichimproves the performance of the auction process But it willalso generate more computational costs and increase thecomplexity In this paper one Resource Block (RB) and 01Wtransmission power are chosen as the basic bandwidth unitand transmission power unit respectively We choose 01Was the basic transmission power here because the DynamicProgramming will be adopted in the next subsection to solvethe optimization problem where the integer data are neededby the Dynamic Programming method
32 Reserve Auction Algorithm As mentioned above thereverse auction process includes two steps of the 119860119897119897119900119888119886119905119894119900119899and 119875119903119894119888119894119899119892
321 Allocation In traditional reverse auction processes theallocation result is completely decided by the bids that isthe bidders who offer the largest supply of resources will winthe auction However in this paper besides the resourcesthat the bidders can provide the energy efficiency achievedby the bidder should also be considered Assume that B =B1B2 B
119895 B
119899 and P = P
1P2 P
119895 P
119899
represent the allocation result where B119895= 1198611
119895 1198612
119895 119861
119897119895
119895
is the RB that 119895th BS could provide and 119861119894119895= 0 if the 119894th RB
in the 119895th BS is not needed P119895= 1198751
119895 1198752
119895 119875
119897119895
119895 denotes the
transmission power that the 119895th BS could transmit and 119875119894119895is
the transmission power on 119861119894119895 If B119895or P119895equals zero the 119895th
BS loses in this auction processThe 119860119897119897119900119888119886119905119894119900119899 problem is formulated as
maxB119895 P119895
119862system
119875119905
(1)
st119899
sum
119895=1
119897119895
sum
119894=1
119861119894
119895log(1 +
119875119894
119895119867119895
119868119894
119895+ 1198730119861119894
119895
) ge 120591119903forall119903 (2)
Mobile Information Systems 5
119861119894
119895isin 0 119861 (3)
119875119894
119895isin P119898P119904 (4)
P119898=
119875119894
119895| 0 le 119875
119894
119895le 119875119898
119897119895
sum
119894=1
119875119894
119895le 119875119898
(5)
P119904=
119875119894
119895| 0 le 119875
119894
119895le 119875119904
119897119895
sum
119894=1
119875119894
119895le 119875119904
(6)
In (1) 119862system = sum119899
119895=1119862119895sum119897119895
119894=1119861119894
119895denotes the throughput
119875119905= sum119897119895
119894=1119875119894
119895sum119899
119895=1119897119895denotes the expectation of transmission
power on each RB from all the bidders 119862system119875119905 is thesystem energy efficiency and 119875
119905gt 0
In (2) 119867119895= 119889minus120572119895
119895|ℎ119895|2120594119895is the channel gain between
the 119895th BS and UE where 119889119895is the distance between the
119895th BS and UE 120572119895is the path-loss exponent of the 119895th
BS ℎ119895is the Rayleigh fading component 120594
119895denotes the
log-normally distributed shadow fading Furthermore 119868119894119895=
sum119899
1198951015840=11198951015840=119895119875119894
11989510158401198671198951015840V119894
1198951015840 is the interference experienced by the UE
on the 119894th RB where 1198751198941198951015840 is the transmission power on the 119894th
RB from the 1198951015840th BS1198671198951015840 is the channel gain between the 1198951015840th
BS andUE V1198941198951015840 is the binary variables representing the activity
factor of 1198951015840th BS 1198730is the thermal noise level and 120591
119903is the
guaranteed throughput threshold of the offloading UEConstraint (3) means the RB 119861119894
119895can be occupied or
vacant Constraint (4) denotes the bidder can be a macrocellBS or a small cell BS In (5) and (6) we give the requirementfor the transmission power in the corresponding RB where119875119898and 119875119904are the power limitations for the macrocell BS and
small cell BS respectivelyFor the convenience of solving this problem we trans-
form (1) (2) and (3) into the following form
minB119895P119895
119875119905
119862system
119862system ge 120591119903
0 le 119861119894
119895le 119861
(7)
Assuming the 119862system is an independent variable theKarush-Kuhn-Tucker (KKT) conditions are given as
nablaB119895P119895system119875119905
119862systemminus nablaB119895P119895system (119862system minus 120591119903) 120583
minus nablaB119895 P119895system119861119894
119895120590119894
119895minus nablaB119895P119895system (119861 minus 119861
119894
119895) 120585119894
119895
minus nablaB119895 P119895system119875119894
119895]119894119895minus nablaB119895P119895system (119875119904 (119875119898) minus 119875
119894
119895) 120582119894
119895
minus nablaB119895 P119895system(119875119904 (119875119898) minus
119897119895
sum
119894=1
119875119894
119895)120588119895= 0
(8)
(119862system minus 120591119903) 120583 = 0 (9)
119861119894
119895120590119894
119895= 0 119861
119894
119895ge 0 120590
119894
119895ge 0 (10)
(119861 minus 119861119894
119895) 120585119894
119895= 0 119861 le 119861
119894
119895 120585119894
119895ge 0 (11)
119875119894
119895]119894119895= 0 119875
119894
119895ge 0 ]119894
119895ge 0 (12)
(119875119904(119875119898) minus 119875119894
119895) 120582119894
119895= 0 119875
119894
119895le 119875119904(119875119898) 120582119894
119895ge 0 (13)
(119875119904(119875119898) minus
119897119895
sum
119894=1
119875119894
119895)120588119895= 0
119897119895
sum
119894=1
119875119894
119895le 119875119904(119875119898) 120588119895ge 0 (14)
Equation (8) is changed into (15) by taking the derivativeof 119862system as
minus119875119905
1198622systemminus 120583 = 0 (15)
119875119905
1198622system+ 120583 = 0 (16)
Thismakes sensewhere119875119905and1198622system are nonzero values
Thus 120583 = 0 (9) shows (119862system minus 120591119903)120583 = 0 where
119862system = 120591119903 (17)
When the offloading process is triggered the offloadeduser traffic requirements and the number of BSs participatingin the auction are known The KKT conditions prove that(9) is guaranteed throughput requirement The 119860119897119897119900119888119886119905119894119900119899problem is transformed to solve minB119895P119895119875119905 which can be
simplified into a linear objective functionminsum119897119895119894=1119875119894
119895sum119899
119895=1119897119895
Because the above119860119897119897119900119888119886119905119894119900119899problem is a linear problemit is easy to find that this is a multiple knapsack problemIn order to facilitate the solving process we turn it intoa 0-1 knapsack problem There are 119899 BSs to participate inthe auction the 119894th BS resources are separated into 119872
119894
independent piece of ldquogoodsrdquo that can be loaded into aknapsack This will get a 0-1 knapsack problem where thenumber of items is sum119872
119894 If we solve this problem directly
the computational complexity will be 119874(119862sum119872119894) In order to
reduce the complexity we design another algorithm with thespecific plan as follows
As mentioned above the 119894th BS bandwidth resourceis composed of a number of RB groups Considering thecondition of the binary the guaranteeing of selecting anymultiple resource package strategy still can be achievedafter the transformation of the original multiple knapsackproblem One BS which has119872
119894resource blocks is separated
into several RB groups where these RB groups respectivelyhave 1 2 22 23 119872
119894minus 2119896minus1+ 1 resource blocks The 119894th BS
has Ceiling(log119872119894) different RB groups participating in the
0-1 knapsack problemInitialize the value as follows the weight of knapsack is
119862 which is the offloading user traffic requirements 120591119903 The
starting value 119865[0 119888] = 0 119865[119909 0] = 0 And the 119894th BS hasCeiling(log119872
119894) stages
6 Mobile Information Systems
(1) Let 1205780= 0
(2) for 119895 = 1 to 119899 do(3) if Bidder b119895 is a macrocell BS then(4) for 119875 = 1 to 10119875
119898do
(5) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(6) end for(7) for 119894 = 1 to 119897
119895do
(8) if 119875119894119895gt 0 then
(9) 119861119894
119895= 119861
(10) else(11) 119861
119894
119895= 0
(12) end if(13) end for(14) else if Bidder b119895 is a small cell BS then(15) for 119875 = 1 to 10119875
119904do
(16) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(17) end for(18) for 119894 = 1 to 119897
119895do
(19) if 119875119894119895gt 0 then
(20) 119861119894
119895= 119861
(21) else(22) 119861
119894
119895= 0
(23) end if(24) end for(25) end if(26) end for
Algorithm 1 Reverse auction 119860119897119897119900119888119886119905119894119900119899 (119899 b119899)
Renumber all RB groups 119888119909is the capacity of the 119909
stage which denotes the 119894th RB group of the 119895th BS and thecorresponding value is V
119909
The capacity of the 119909 stage is
119888119909= 119908 [119894 119895] = 119861
119894
119895log(1 +
119875119894
119895119867119895
119868119894
119895+ 1198730119861119894
119895
) (18)
The value of the corresponding V119909is
V119909= V [119894 119895] = 119875119894
119895 (19)
The iterative equation will be
119865 [119909 119888] = min 119865 [119909 minus 1 119888] 119865 [119909 minus 1 119888 minus 119888119909] + V119909 (20)
In (20) 119865[119909 119888] is the minimum value of the transmissionpower in the stage 119909 119888 denotes the remaining space ofthe pack according to the current stage 119888
119909denotes the
provided capacity when choosing the 119909 stage V119909denotes the
provided transmission power when choosing the 119909 stageThecomputational complexity is reduced to 119874(119862sum log119872
119894)
The proposed 119860119897119897119900119888119886119905119894119900119899 algorithm is illustrated inAlgorithm 1 with B = B
1B2 B
119899 and P = P
1P2
P119899In Algorithm 1 the transmission power of each bidder is
chosen firstly and the corresponding bandwidth is decidedbased on the transmission power allocation results As
for 119895 = 1 to 119899 do(2) if 119895th BS is a winning bidder then
Reverse Auction - 119860119897119897119900119888119886119905119894119900119899 (119899 119895B b119894)(4) 119901
119895= 120578B119899b119894 minus (120578B119899 minus 120578b119894 )
else(6) 119901
119895= 0
end if(8) end for
Algorithm 2 Reverse auction 119875119903119894119888119894119899119892 (119899 b119899BP 120578)
mentioned before 01W is chosen as the basic transmissionpower unit and the integer data is needed in the DynamicProgramming Therefore the range of 119875 is defined from 1
to 10119875119898or 10119875
119904 The equation in Line (5) and Line (16) of
Algorithm 1 mean that the 119895th bidderrsquos transmission powershould be chosen to achieve the largest 120578b119895 and guarantee theoffloaded userrsquos throughput threshold 120591
119903at the same time
where 120578b119895 denotes the energy efficiency of the 119895th bidderAfter implementing the Dynamic Programming the resultsof Algorithm 1 P10 and B will be the optimal allocationsolution for the proposed reverse auction process
322 Pricing In traditional 119875119903119894119888119894119899119892 algorithms the biddersare encouraged to set their own bids truthfully as illustratedbefore So in this paper the same energy efficiency that thecorresponding bidder achieves is paid back With regard tothe offloading user throughput threshold 120591
119903 we define 120578
1and
1205782as (21) and (22) as follows
1205781= 120578B119899b119894 = max
B119895B119894P119895P119894
119862system
119864 (119875119905)
(21)
1205782= 120578B119899 minus 120578b119894 = (max
B119895P119895
119862system
119864 (119875119905)) minus
119862b119894
119864 (P119894) (22)
where 1205781denotes the system energy efficiency under the
optimal 119860119897119897119900119888119886119905119894119900119899 solution without the presence of the 119894thBS The 120578
2denotes the system energy efficiency except for
the 119894th BS under current optimal119860119897119897119900119888119886119905119894119900119899 resultsThen theopportunity cost of the 119894th BS is defined as the differencebetween 120578
1and 1205782 just as illustrated in (23) [19] as follows
119901119894= 1205781minus 1205782= 120578B119899b119894 minus (120578B119899 minus 120578b119894) (23)
The 119875119903119894119888119894119899119892 algorithm is given as Algorithm 2
323 Properties In this section the properties of theproposed reverse auction model are analyzed Accordingto the VCG based reverse auction model the IndividualRa-tionality and the 119879119903119906119905ℎ119891119906119897119899119890119904119904 properties need to be proved
IndividualRationality When the utility of each participatingbidder in the119875119903119894119888119894119899119892 stage is greater than zero this algorithmis individual rational for each winning bidder Namely
119901119894= 120578B119899b119894 minus (120578B119899 minus 120578b119894) ge 0 (24)
Mobile Information Systems 7
119879119903119906119905ℎ119891119906119897119899119890119904119904 For each bidder the Truthfulness means thateach bidderrsquos bid price is equal to its private value This isa weakly dominant strategy If BSrsquos bidding is untrue theenergy efficiency will be unlikely the biggest In order toget the maximum energy efficiency the allocation should beformulated as follows
119901119895= 120578B119899b119895 minus (120578B119899 minus 120578b119895)
120575 = 119901119895minus 119901119894
= 120578B119899b119895 minus (120578B119899 minus 120578b119895) minus [120578B119899b119894 minus (120578B119899 minus 120578b119894)]
= 120578B119899b119895 minus 120578B119899 + 120578b119895 minus 120578B119899b119894 + 120578B119899 minus 120578b119894
= 120578B119899b119895 + 120578b119895 minus 120578B119899b119894 minus 120578b119894
= (120578B119899b119895 + 120578b119895) minus (120578B119899b119894 + 120578b119894)
(25)
Based on the proposed model in this paper because 119901119895le
119901119894and120575 le 0 thismeans 120578B119899b119895+120578b119895 le 120578B119899b119894+120578b119894 If and only
if 119895 = 119894 it can take the equal signTherefore each bidder mustbe truthful to obtain the maximum system energy efficiencyThe proof is finished
4 Performance Evaluation
In this section we built the system-level simulation plat-form according to the 3GPP LTE-Advanced simulationmethodology [23] Based on this platform we validate theperformances of the proposed reverse auction based GOscheme with comparison algorithms in the small cell HetNetdownlink scenario
41 Simulation Setting Performance Metrics and ComparisonAlgorithms Theconsidered simulation scenario in this papercomprises 19-hexagonal macrocells with 3 sectors per macro-cell In each sector there is one small cell cluster deployedwith shared spectrum manner The small cell cluster is agroup of densely deployed small cells We deploy 23 usersin the coverage of small cell clusters while the remainingusers are distributed in the coverage area of macro cellsThe users are uniformly distributed Moreover as mentionedabove the bandwidth resource granularity in the simulationis one RB In the initial state each user is served by the BSwhich can provide the highest downlink RSRP Once a newtraffic offloading requirement is requested the reverse auc-tion based GO scheme is triggered The detailed simulationparameters are according to 3GPP LTE-Advanced small cellHetNet evaluation methodology [24] These parameters arelisted in Table 1
The performance metrics include the system energyefficiency offloading gain and throughput In this paper themetric of offloading gain (120574gain) is defined as
120574gain =120591offloading
120591total (26)
where 120591offloading denotes the offloaded throughput and 120591totaldenotes the total system throughput The offloading gain is
Table 1 Simulation parameters
Simulation parameter ValueCarrier frequency 2GHzSystem bandwidth 10MHzTotal transmission power ofmacrocell 46 dBm
Total transmission power of small cell 30 dBmPath-loss of macrocell Pl = 283 + 220log
10(119889)
Path-loss of small cell Pl = 305 + 367log10(119889)
Small cell number per cluster 4sim10Small cell cluster number permacrocell 3
User number per macrocell 60
Antenna gain of macrocells 17 dBiAntenna gain of small cells 5 dBiTraffic model FTP Model 1Power spectrum density of thermalnoise minus174 dBmHz
User throughput threshold 5Mbps
a more straightforward notation about howmuch traffic loadcould be offloaded to improve the energy efficiency
In order to evaluate the performances of the proposedGOscheme we compare it with the TOFFR algorithm proposedin [13] and the incentivized scheme proposed in [14] whichhave been introduced in the related works The simulationresults are given as below
42 Impacts of Small Cell Numbers on Energy EfficiencyAccording to 3GPP simulation assumptions there are 23UEs distributed in the coverage of small cell cluster whilethe remainingUEs are uniformly distributed in the remainingarea of macrocells In this section the impacts of deploymentdensity of small cells in a cluster on the system energyefficiency are investigated According to 3GPP simulationmethodology the number of small cells in a cluster variesfrom 4 to 10
As shown in Figure 2 the system energy efficiency ofdifferent algorithms versus small cell numbers per cluster isdemonstrated We can observe that the system energy effi-ciency is improved with the increasing of small cell numbersThe reason lies that small cells usually can provide higherenergy efficiency than macrocells due to lower transmissionpower attenuations in hot spot deployment scenarios Whenthere exist more small cells inside one macrocell more usertraffic could be offloaded potentially to small cellsThereforehigher system energy efficiency could be further achieved
But the system energy efficiency increases slightly whenthe number of small cells is large This is because when thesmall cells are deployed more densely the intercell interfer-enceswill bemore severe indicating the requirement of largertransmission power to ensure the same user throughputMoreover from results in Figure 2 we have proved that theproposedGO scheme outperforms the TOFFR algorithm andincentivized scheme regardless of small cell density deployed
8 Mobile Information Systems
5 6 7 8 9 104Number of small cells
TOFFRwo offloading
GO schemeIncentivized scheme
5
6
7
8
9
10
11
12
13Sy
stem
ener
gy effi
cien
cy (k
bps
W)
Figure 2 System energy efficiency versus small cell numbers percluster
per cluster In the TOFFR algorithm the fractional frequencyreuse scheme is adopted to improve UE performances betterin the cell edge The UE located in the central area cannotcontribute to the energy efficiency improvement The incen-tivized scheme focuses on maximizing the offloading utilitythat purchases the available unused bandwidth in femtocellsSo the proposed GO scheme has better energy efficiencyperformance than both of them
We also compare these three offloading algorithms withthe situation of no offloading (denoted as wo offloading insimulations) as the baseline It is obvious thatwhen offloadingschemes are adopted more user traffic originally served bymacrocell will be actively offloaded to small cells There-fore all of the offloading schemes including the proposedGO scheme TOFFR algorithm and incentivized algorithmachieve higher system energy efficiency
43 Impacts of Small Cell Numbers on Offloading GainAs shown in Figure 3 the offloading gain versus differentsmall cell numbers per cluster is dipicted We can observethat the offloading gain increases with small cell numbersdue to the capacity growth with the increase of small cellnumbers per cluster Besides in order to maximize thesystem energy efficiency there is limitation on the amountof offloaded throughputs as demonstrated in Figure 3Whenthe offloading gain reaches 62 the rise against the small cellnumbers becomes rather slowThe simulation results furthersuggest that the proposed GO scheme outperforms theTOFFR algorithm and incentivized scheme in terms of notonly energy efficiency but also the offloading gain becausein TOFFR algorithm and incentivized scheme the offloadingis mainly focused on the cell edge users which limits theperformance improvements Moreover Figure 3 shows thatthe increasing of the small cell numbers of all these threeschemes will reach a plateau The reason lies that there arealways several specific users out of the coverage of the small
TOFFRGO schemeIncentivized scheme
4 5 6 7 8 9 10 113Number of small cells per cluster
0
10
20
30
40
50
60
70
Offl
oadi
ng g
ain
()
Figure 3 Offloading gain versus small cell numbers per cluster
91011121314151617181920
Smal
l cel
l clu
ster t
hrou
ghpu
t (M
bps)
5 6 7 8 9 104Number of small cells per cluster
TOFFRwo offloading
GO schemeIncentivized scheme
Figure 4 Small cell throughput versus small cell numbers percluster
cell clusters in the simulations Just as mentioned before23 UEs are deployed in the coverage of small cell clusterand the simulation results also show that the offloading gainlimitation of all three algorithms can only reach to near23 which in turn indicates the offloading limitation will bedecided by the distribution and location of users
44 Impact of Small Cell Numbers on Small Cell ThroughputIn this section the impacts on the throughput of small cellcluster versus different small cell numbers are investigatedFrom Figure 4 we can see the throughput increases with theincrease of small cell numbers per cluster The reason is thatmore users will be served by the small cells after offloadingprocesses Moreover the throughput increments grow slowly
Mobile Information Systems 9
when small cell numbers per cluster are relatively large Thisis because of the increase of intercell interferences caused bydenser small cell deployments Finally the results in Figure 4prove that the proposed GO scheme outperforms TOFFRalgorithm and incentivized algorithm also in terms of smallcell cluster throughput
5 Conclusion
This paper aims to solve the problem regarding how toperformoffloading in the small cell HetNet deployments withoptimization on maximizing the system energy efficiencyThe reverse auction theory has been implemented with theproposed GO scheme design to decide the offloading targetBS or BSs with coordination transmission enabled technol-ogy The reverse auction model is formulated by multiplesellers (BSs) and a single buyer (offloading user) with thefirst price sealed bid mechanism The BS coordination trans-missions are also supported for multiple winning biddersscenarios According to the proposed reverse auction basedGO scheme the energy efficiency optimization problemwith constraints of user guaranteed throughput thresholdbandwidth occupation and transmission power limitationis solved by Dynamic Programming method with KKTconditions The Individual Rationality and Truthfulness ofthe VCG based reverse auction model are also proved in thepaper System-level simulations have been conducted to ver-ify the effectiveness of the proposed GO scheme according to3GPP LTE-Advanced evaluation methodologies The perfor-mances when applying theGO scheme comparison schemesand the baseline without offloading situation are evaluatedwith performance metrics of energy efficiency offloadinggain and throughput The simulation results prove that theproposed GO scheme can achieve supreme performances
Notation Definition
b = b1 b2 b119894 b119897 BidsB = B
1B2 B
119895 B
119899 Bandwidth allocation results
P = P1P2 P
119895 P
119899 Power allocation results
B119895= 1198611
119895 1198612
119895 119861
119897119895
119895 119895th BS bandwidth allocation
resultsP119895= 1198751
119895 1198752
119895 119875
119897119895
119895 119895th BS power allocation
results119861119894
119895 119894th subcarrier in the 119895th BS
119875119894
119895 Transmission power on 119861119894
119895
b119899= b1 b2 b119899 Bids sent by first 119899 BSs
120578 Energy efficiency
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper is supported by the National High Technol-ogy Research and Development Program of China no2014AA01A701 Nature and Science Foundation of China
under Grants nos 61471068 and 61421061 InternationalCollaboration Project no 2015DFT10160 andNationalMajorProject no 2016ZX03001009-003
References
[1] T Nakamura S Nagata A Benjebbour et al ldquoTrends in smallcell enhancements in LTE advancedrdquo IEEE CommunicationsMagazine vol 51 no 2 pp 98ndash105 2013
[2] W Ni and I B Collings ldquoA new adaptive small-cell architec-turerdquo IEEE Journal on Selected Areas in Communications vol31 no 5 pp 829ndash839 2013
[3] L Hanzo H Haas S Imre D OrsquoBrien M Rupp and LGyongyosi ldquoWireless myths realities and futures from 3G4Gto optical and quantum wirelessrdquo Proceedings of the IEEE vol100 pp 1853ndash1888 2012
[4] TheMETIS 2020 ProjectmdashLaying the Foundation of 5G httpswwwmetis2020com
[5] ldquoSmall cell market statusrdquo White Paper Informa and Small CellForum 1 2013
[6] D Calin H Claussen and H Uzunalioglu ldquoOn femto deploy-ment architectures and macrocell offloading benefits in jointmacro-femto deploymentsrdquo IEEE Communications Magazinevol 48 no 1 pp 26ndash32 2010
[7] S-I Sou ldquoMobile data offloading with policy and chargingcontrol in 3GPP core networkrdquo IEEE Transactions on VehicularTechnology vol 62 no 7 pp 3481ndash3486 2013
[8] J Korhonen T Savolainen A Y Ding and M Kojo ldquoTowardnetwork controlled IP traffic offloadingrdquo IEEE CommunicationsMagazine vol 51 no 3 pp 96ndash102 2013
[9] H Elsawy E Hossain and S Camorlinga ldquoTraffic offloadingtechniques in two-tier femtocell networksrdquo in Proceedings of theIEEE International Conference onCommunications (ICC rsquo13) pp6086ndash6090 IEEE Budapest Hungary June 2013
[10] Z Lu P Sinha and R Srikant ldquoEasyBid enabling cellularoffloading via small playersrdquo in Proceedings of the 33rd IEEEConference on Computer Communications (IEEE INFOCOMrsquo14) pp 691ndash699 IEEE Toronto Canada May 2014
[11] M Usman A Vastberg and T Edler ldquoEnergy efficient highcapacityHETNETby offloading highQoSusers through femtordquoin Proceedings of the 17th IEEE International Conference onNetworks (ICON rsquo11) pp 19ndash24 Singapore December 2011
[12] P Chandhar and S S Das ldquoAnalytical evaluation of offloadinggain in macrocell-femtocell OFDMA networksrdquo in Proceedingsof the IEEE 77th Vehicular Technology Conference (VTC Springrsquo13) pp 1ndash6 June 2013
[13] Q Liu G Feng and S Qin ldquoEnergy-efficient traffic offloadingin Macro-Pico networksrdquo in Proceedings of the 22nd Wirelessand Optical Communications Conference (WOCC rsquo13) pp 236ndash241 IEEE Chongqing China May 2013
[14] Y Jia M Zhao K Wang and W Zhou ldquoAn incentivizedoffloading mechanism via truthful auction in heterogeneousnetworksrdquo in Proceedings of the 6th International Conference onWireless Communications and Signal Processing (WCSP rsquo14) pp1ndash6 Hefei China October 2014
[15] L Gao G Iosifidis J Huang L Tassiulas and D Li ldquoBargain-ing-based mobile data offloadingrdquo IEEE Journal on SelectedAreas in Communications vol 32 no 6 pp 1114ndash1125 2014
[16] F Zhang W Zhang and Q Ling ldquoNon-cooperative game forcapacity offloadrdquo IEEE Transactions on Wireless Communica-tions vol 11 no 4 pp 1565ndash1575 2012
10 Mobile Information Systems
[17] X Xu H Zhang X Dai and X Tao ldquoOptimal Energy EfficientOffloading in small cell HetNet with auctionrdquo in Proceedingsof the 9th International Conference on Communications andNetworking in China (CHINACOM rsquo14) pp 335ndash340MaomingChina August 2014
[18] D P Bertsekas D A Castanon and H Tsaknakis ldquoReverseauction and the solution of inequality constrained assignmentproblemsrdquo SIAM Journal on Optimization vol 3 no 2 pp 268ndash297 1993
[19] M Khaledi and A A Abouzeid ldquoDynamic spectrum sharingauction with time-evolving channel qualitiesrdquo IEEE Transac-tions onWireless Communications vol 14 no 11 pp 5900ndash59122015
[20] F Shen D Li P-H Lin and E Jorswieck ldquoAuction basedspectrum sharing for hybrid access in macro-femtocell net-works under QoS requirementsrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC rsquo15) pp3335ndash3340 IEEE London UK June 2015
[21] X Zhuo W Gao G Cao and S Hua ldquoAn incentive frameworkfor cellular traffic offloadingrdquo IEEE Transactions on MobileComputing vol 13 no 3 pp 541ndash555 2014
[22] W Vickrey ldquoCounterspeculation auctions and competitivesealed tendersrdquo The Journal of Finance vol 16 no 1 pp 8ndash371961
[23] 3GPP-TR36814 (v1110) ldquoEvolved Universal Terrestrial RadioAccess (E-UTRA) Further advancements for E-UTRA physicallayer aspectsrdquo 2013
[24] 3GPP ldquoSmall cell enhancements for E-UTRA and EUTRANphysical layer aspects (release 12)rdquo 3GPP TR 36872 2013
[25] 3GPP R1-130744 ldquoWF on evaluation assumptions for SCEphysical layerrdquo Huawei HiSilicon CATR CMCC 2013
[26] 3GPP ldquoStudy on small cell enhancements for EUTRA and E-UTRAN higher layer aspectsrdquo 3GPP TR 36842 (v1200) 2013
[27] 3GPP TS 36300 ldquoTechnical Specification Group Radio AccessNetwork Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Net-work (EUTRAN) Overall description Stage 2 (Release 12)rdquo2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
4 Mobile Information Systems
the winning bidders are BS1 and BS3 together who win thereverse auction and are designated as the offloading target BSsto serve the UE coordinately
In order to present the reverse auction based GO schemeclearer the related notation definitions are introduced asfollows
Bid (119887119894) submitted by the 119861119878
119894to convey how much
bandwidth and transmission power it can provide tothe offloading UE which may not always equal allavailable resources the BS can providePrivate Value (119909
119894) available resources in the 119861119878
119894
which is only known by the 119861119878119894itself
Pricing (119901119894) the opportunity cost of the 119894th BS which
makes sure the system energy efficiency could bemaximum
Based on the auction theory when the condition 119887119894= 119909119894
is satisfied the auction process is truthful Moreover 119887119894=
119909119894is a weakly dominant strategy [20] As a result in this
paper we set 119887119894= 119909119894 assuming all the bidders participating
in the auction process will send the bids with all availableresources it can provide which guarantees the truthfulnessof the reverse auction model
Although we have set the 119861119894119889 equal to the PrivateValuethe reverse auction basedGO schemewithmultiple candidatetarget BSs is still NP-hard problem In order to solve thisproblem an approximation algorithm is designed in the nextsection The notations used in the algorithm are introducedin the Notation Definition
23 User Delay Tolerance The reverse auction based GOscheme will start upon the receipt of offloading request at theserving BS periodically For the traditional auction processthere usually exist multiround bidding procedures to achievethe final winner This process would inevitably generate anextra delay for the BSs to wait for the auction consequenceBut for the wireless network the user traffic usually hasdelay tolerance requirements which should be consideredin the offloading scheme Therefore we implement thesingle-round auction in this paper in order to prevent theinformation exchange overhead and corresponding delay forthe offloading users Since the number of BSs surrounding aspecific UE is limited the extra delay caused by the single-round first price sealed bid auction process could be notsignificant This delay avoidance mechanism is also feasiblefor the handover based offloading schemes that usually needmore time for measurements and handover decisions
3 Reverse Auction Based GO Scheme
In this section the proposed reverse auction based GOscheme is described The main steps of the proposed GOscheme are given in Figure 1 Firstly according to theoffloading request the UErsquos current serving BS collects allthe bids from its neighboring BSs Then the serving BScalculates the throughput that all bidders can provide andderives the expected energy efficiency increment for eachcandidate BS Based on the derived throughput and energy
efficiency increment a single-round reverse auction processis performed which includes the 119860119897119897119900119888119886119905119894119900119899 and 119875119903119894119888119894119899119892steps Finally the auction results are sent back to the biddersand the user will be offloaded to the winners accordingly
31 Bidding In order to contribute to the offloading processbidders will append available resources with their bids toreveal the throughput they can provide for the offloadinguser For each bidder the upper bound of bandwidth andtransmission power that a BS can provide is119861bound and119875boundrespectively 119861bound and 119875bound can be divided into multipleunits and classified asmultiple bids b = b1 b2 b119894 b119897to indicate the resources the UE can obtain from each bidderwhere 119897 = max(lfloor119861bound119890119861rfloor lfloor119875bound119890119875rfloor) 119890119861 is the basicbandwidth unit 119890
119875is the basic transmission power unit for
the bidder The b119894 = (119861119894 119875119894) consists of both bandwidth andtransmission power resources After receiving all of the bidsUErsquos current serving BS can know how many resources eachbidder can provide with the value not larger than the sum119897
119894=1119861119894
and sum119897119894=1119875119894 The scale of the bandwidth and transmission
power unit can be flexibly set by the system The smallerunit definition leads to more information in the bid whichimproves the performance of the auction process But it willalso generate more computational costs and increase thecomplexity In this paper one Resource Block (RB) and 01Wtransmission power are chosen as the basic bandwidth unitand transmission power unit respectively We choose 01Was the basic transmission power here because the DynamicProgramming will be adopted in the next subsection to solvethe optimization problem where the integer data are neededby the Dynamic Programming method
32 Reserve Auction Algorithm As mentioned above thereverse auction process includes two steps of the 119860119897119897119900119888119886119905119894119900119899and 119875119903119894119888119894119899119892
321 Allocation In traditional reverse auction processes theallocation result is completely decided by the bids that isthe bidders who offer the largest supply of resources will winthe auction However in this paper besides the resourcesthat the bidders can provide the energy efficiency achievedby the bidder should also be considered Assume that B =B1B2 B
119895 B
119899 and P = P
1P2 P
119895 P
119899
represent the allocation result where B119895= 1198611
119895 1198612
119895 119861
119897119895
119895
is the RB that 119895th BS could provide and 119861119894119895= 0 if the 119894th RB
in the 119895th BS is not needed P119895= 1198751
119895 1198752
119895 119875
119897119895
119895 denotes the
transmission power that the 119895th BS could transmit and 119875119894119895is
the transmission power on 119861119894119895 If B119895or P119895equals zero the 119895th
BS loses in this auction processThe 119860119897119897119900119888119886119905119894119900119899 problem is formulated as
maxB119895 P119895
119862system
119875119905
(1)
st119899
sum
119895=1
119897119895
sum
119894=1
119861119894
119895log(1 +
119875119894
119895119867119895
119868119894
119895+ 1198730119861119894
119895
) ge 120591119903forall119903 (2)
Mobile Information Systems 5
119861119894
119895isin 0 119861 (3)
119875119894
119895isin P119898P119904 (4)
P119898=
119875119894
119895| 0 le 119875
119894
119895le 119875119898
119897119895
sum
119894=1
119875119894
119895le 119875119898
(5)
P119904=
119875119894
119895| 0 le 119875
119894
119895le 119875119904
119897119895
sum
119894=1
119875119894
119895le 119875119904
(6)
In (1) 119862system = sum119899
119895=1119862119895sum119897119895
119894=1119861119894
119895denotes the throughput
119875119905= sum119897119895
119894=1119875119894
119895sum119899
119895=1119897119895denotes the expectation of transmission
power on each RB from all the bidders 119862system119875119905 is thesystem energy efficiency and 119875
119905gt 0
In (2) 119867119895= 119889minus120572119895
119895|ℎ119895|2120594119895is the channel gain between
the 119895th BS and UE where 119889119895is the distance between the
119895th BS and UE 120572119895is the path-loss exponent of the 119895th
BS ℎ119895is the Rayleigh fading component 120594
119895denotes the
log-normally distributed shadow fading Furthermore 119868119894119895=
sum119899
1198951015840=11198951015840=119895119875119894
11989510158401198671198951015840V119894
1198951015840 is the interference experienced by the UE
on the 119894th RB where 1198751198941198951015840 is the transmission power on the 119894th
RB from the 1198951015840th BS1198671198951015840 is the channel gain between the 1198951015840th
BS andUE V1198941198951015840 is the binary variables representing the activity
factor of 1198951015840th BS 1198730is the thermal noise level and 120591
119903is the
guaranteed throughput threshold of the offloading UEConstraint (3) means the RB 119861119894
119895can be occupied or
vacant Constraint (4) denotes the bidder can be a macrocellBS or a small cell BS In (5) and (6) we give the requirementfor the transmission power in the corresponding RB where119875119898and 119875119904are the power limitations for the macrocell BS and
small cell BS respectivelyFor the convenience of solving this problem we trans-
form (1) (2) and (3) into the following form
minB119895P119895
119875119905
119862system
119862system ge 120591119903
0 le 119861119894
119895le 119861
(7)
Assuming the 119862system is an independent variable theKarush-Kuhn-Tucker (KKT) conditions are given as
nablaB119895P119895system119875119905
119862systemminus nablaB119895P119895system (119862system minus 120591119903) 120583
minus nablaB119895 P119895system119861119894
119895120590119894
119895minus nablaB119895P119895system (119861 minus 119861
119894
119895) 120585119894
119895
minus nablaB119895 P119895system119875119894
119895]119894119895minus nablaB119895P119895system (119875119904 (119875119898) minus 119875
119894
119895) 120582119894
119895
minus nablaB119895 P119895system(119875119904 (119875119898) minus
119897119895
sum
119894=1
119875119894
119895)120588119895= 0
(8)
(119862system minus 120591119903) 120583 = 0 (9)
119861119894
119895120590119894
119895= 0 119861
119894
119895ge 0 120590
119894
119895ge 0 (10)
(119861 minus 119861119894
119895) 120585119894
119895= 0 119861 le 119861
119894
119895 120585119894
119895ge 0 (11)
119875119894
119895]119894119895= 0 119875
119894
119895ge 0 ]119894
119895ge 0 (12)
(119875119904(119875119898) minus 119875119894
119895) 120582119894
119895= 0 119875
119894
119895le 119875119904(119875119898) 120582119894
119895ge 0 (13)
(119875119904(119875119898) minus
119897119895
sum
119894=1
119875119894
119895)120588119895= 0
119897119895
sum
119894=1
119875119894
119895le 119875119904(119875119898) 120588119895ge 0 (14)
Equation (8) is changed into (15) by taking the derivativeof 119862system as
minus119875119905
1198622systemminus 120583 = 0 (15)
119875119905
1198622system+ 120583 = 0 (16)
Thismakes sensewhere119875119905and1198622system are nonzero values
Thus 120583 = 0 (9) shows (119862system minus 120591119903)120583 = 0 where
119862system = 120591119903 (17)
When the offloading process is triggered the offloadeduser traffic requirements and the number of BSs participatingin the auction are known The KKT conditions prove that(9) is guaranteed throughput requirement The 119860119897119897119900119888119886119905119894119900119899problem is transformed to solve minB119895P119895119875119905 which can be
simplified into a linear objective functionminsum119897119895119894=1119875119894
119895sum119899
119895=1119897119895
Because the above119860119897119897119900119888119886119905119894119900119899problem is a linear problemit is easy to find that this is a multiple knapsack problemIn order to facilitate the solving process we turn it intoa 0-1 knapsack problem There are 119899 BSs to participate inthe auction the 119894th BS resources are separated into 119872
119894
independent piece of ldquogoodsrdquo that can be loaded into aknapsack This will get a 0-1 knapsack problem where thenumber of items is sum119872
119894 If we solve this problem directly
the computational complexity will be 119874(119862sum119872119894) In order to
reduce the complexity we design another algorithm with thespecific plan as follows
As mentioned above the 119894th BS bandwidth resourceis composed of a number of RB groups Considering thecondition of the binary the guaranteeing of selecting anymultiple resource package strategy still can be achievedafter the transformation of the original multiple knapsackproblem One BS which has119872
119894resource blocks is separated
into several RB groups where these RB groups respectivelyhave 1 2 22 23 119872
119894minus 2119896minus1+ 1 resource blocks The 119894th BS
has Ceiling(log119872119894) different RB groups participating in the
0-1 knapsack problemInitialize the value as follows the weight of knapsack is
119862 which is the offloading user traffic requirements 120591119903 The
starting value 119865[0 119888] = 0 119865[119909 0] = 0 And the 119894th BS hasCeiling(log119872
119894) stages
6 Mobile Information Systems
(1) Let 1205780= 0
(2) for 119895 = 1 to 119899 do(3) if Bidder b119895 is a macrocell BS then(4) for 119875 = 1 to 10119875
119898do
(5) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(6) end for(7) for 119894 = 1 to 119897
119895do
(8) if 119875119894119895gt 0 then
(9) 119861119894
119895= 119861
(10) else(11) 119861
119894
119895= 0
(12) end if(13) end for(14) else if Bidder b119895 is a small cell BS then(15) for 119875 = 1 to 10119875
119904do
(16) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(17) end for(18) for 119894 = 1 to 119897
119895do
(19) if 119875119894119895gt 0 then
(20) 119861119894
119895= 119861
(21) else(22) 119861
119894
119895= 0
(23) end if(24) end for(25) end if(26) end for
Algorithm 1 Reverse auction 119860119897119897119900119888119886119905119894119900119899 (119899 b119899)
Renumber all RB groups 119888119909is the capacity of the 119909
stage which denotes the 119894th RB group of the 119895th BS and thecorresponding value is V
119909
The capacity of the 119909 stage is
119888119909= 119908 [119894 119895] = 119861
119894
119895log(1 +
119875119894
119895119867119895
119868119894
119895+ 1198730119861119894
119895
) (18)
The value of the corresponding V119909is
V119909= V [119894 119895] = 119875119894
119895 (19)
The iterative equation will be
119865 [119909 119888] = min 119865 [119909 minus 1 119888] 119865 [119909 minus 1 119888 minus 119888119909] + V119909 (20)
In (20) 119865[119909 119888] is the minimum value of the transmissionpower in the stage 119909 119888 denotes the remaining space ofthe pack according to the current stage 119888
119909denotes the
provided capacity when choosing the 119909 stage V119909denotes the
provided transmission power when choosing the 119909 stageThecomputational complexity is reduced to 119874(119862sum log119872
119894)
The proposed 119860119897119897119900119888119886119905119894119900119899 algorithm is illustrated inAlgorithm 1 with B = B
1B2 B
119899 and P = P
1P2
P119899In Algorithm 1 the transmission power of each bidder is
chosen firstly and the corresponding bandwidth is decidedbased on the transmission power allocation results As
for 119895 = 1 to 119899 do(2) if 119895th BS is a winning bidder then
Reverse Auction - 119860119897119897119900119888119886119905119894119900119899 (119899 119895B b119894)(4) 119901
119895= 120578B119899b119894 minus (120578B119899 minus 120578b119894 )
else(6) 119901
119895= 0
end if(8) end for
Algorithm 2 Reverse auction 119875119903119894119888119894119899119892 (119899 b119899BP 120578)
mentioned before 01W is chosen as the basic transmissionpower unit and the integer data is needed in the DynamicProgramming Therefore the range of 119875 is defined from 1
to 10119875119898or 10119875
119904 The equation in Line (5) and Line (16) of
Algorithm 1 mean that the 119895th bidderrsquos transmission powershould be chosen to achieve the largest 120578b119895 and guarantee theoffloaded userrsquos throughput threshold 120591
119903at the same time
where 120578b119895 denotes the energy efficiency of the 119895th bidderAfter implementing the Dynamic Programming the resultsof Algorithm 1 P10 and B will be the optimal allocationsolution for the proposed reverse auction process
322 Pricing In traditional 119875119903119894119888119894119899119892 algorithms the biddersare encouraged to set their own bids truthfully as illustratedbefore So in this paper the same energy efficiency that thecorresponding bidder achieves is paid back With regard tothe offloading user throughput threshold 120591
119903 we define 120578
1and
1205782as (21) and (22) as follows
1205781= 120578B119899b119894 = max
B119895B119894P119895P119894
119862system
119864 (119875119905)
(21)
1205782= 120578B119899 minus 120578b119894 = (max
B119895P119895
119862system
119864 (119875119905)) minus
119862b119894
119864 (P119894) (22)
where 1205781denotes the system energy efficiency under the
optimal 119860119897119897119900119888119886119905119894119900119899 solution without the presence of the 119894thBS The 120578
2denotes the system energy efficiency except for
the 119894th BS under current optimal119860119897119897119900119888119886119905119894119900119899 resultsThen theopportunity cost of the 119894th BS is defined as the differencebetween 120578
1and 1205782 just as illustrated in (23) [19] as follows
119901119894= 1205781minus 1205782= 120578B119899b119894 minus (120578B119899 minus 120578b119894) (23)
The 119875119903119894119888119894119899119892 algorithm is given as Algorithm 2
323 Properties In this section the properties of theproposed reverse auction model are analyzed Accordingto the VCG based reverse auction model the IndividualRa-tionality and the 119879119903119906119905ℎ119891119906119897119899119890119904119904 properties need to be proved
IndividualRationality When the utility of each participatingbidder in the119875119903119894119888119894119899119892 stage is greater than zero this algorithmis individual rational for each winning bidder Namely
119901119894= 120578B119899b119894 minus (120578B119899 minus 120578b119894) ge 0 (24)
Mobile Information Systems 7
119879119903119906119905ℎ119891119906119897119899119890119904119904 For each bidder the Truthfulness means thateach bidderrsquos bid price is equal to its private value This isa weakly dominant strategy If BSrsquos bidding is untrue theenergy efficiency will be unlikely the biggest In order toget the maximum energy efficiency the allocation should beformulated as follows
119901119895= 120578B119899b119895 minus (120578B119899 minus 120578b119895)
120575 = 119901119895minus 119901119894
= 120578B119899b119895 minus (120578B119899 minus 120578b119895) minus [120578B119899b119894 minus (120578B119899 minus 120578b119894)]
= 120578B119899b119895 minus 120578B119899 + 120578b119895 minus 120578B119899b119894 + 120578B119899 minus 120578b119894
= 120578B119899b119895 + 120578b119895 minus 120578B119899b119894 minus 120578b119894
= (120578B119899b119895 + 120578b119895) minus (120578B119899b119894 + 120578b119894)
(25)
Based on the proposed model in this paper because 119901119895le
119901119894and120575 le 0 thismeans 120578B119899b119895+120578b119895 le 120578B119899b119894+120578b119894 If and only
if 119895 = 119894 it can take the equal signTherefore each bidder mustbe truthful to obtain the maximum system energy efficiencyThe proof is finished
4 Performance Evaluation
In this section we built the system-level simulation plat-form according to the 3GPP LTE-Advanced simulationmethodology [23] Based on this platform we validate theperformances of the proposed reverse auction based GOscheme with comparison algorithms in the small cell HetNetdownlink scenario
41 Simulation Setting Performance Metrics and ComparisonAlgorithms Theconsidered simulation scenario in this papercomprises 19-hexagonal macrocells with 3 sectors per macro-cell In each sector there is one small cell cluster deployedwith shared spectrum manner The small cell cluster is agroup of densely deployed small cells We deploy 23 usersin the coverage of small cell clusters while the remainingusers are distributed in the coverage area of macro cellsThe users are uniformly distributed Moreover as mentionedabove the bandwidth resource granularity in the simulationis one RB In the initial state each user is served by the BSwhich can provide the highest downlink RSRP Once a newtraffic offloading requirement is requested the reverse auc-tion based GO scheme is triggered The detailed simulationparameters are according to 3GPP LTE-Advanced small cellHetNet evaluation methodology [24] These parameters arelisted in Table 1
The performance metrics include the system energyefficiency offloading gain and throughput In this paper themetric of offloading gain (120574gain) is defined as
120574gain =120591offloading
120591total (26)
where 120591offloading denotes the offloaded throughput and 120591totaldenotes the total system throughput The offloading gain is
Table 1 Simulation parameters
Simulation parameter ValueCarrier frequency 2GHzSystem bandwidth 10MHzTotal transmission power ofmacrocell 46 dBm
Total transmission power of small cell 30 dBmPath-loss of macrocell Pl = 283 + 220log
10(119889)
Path-loss of small cell Pl = 305 + 367log10(119889)
Small cell number per cluster 4sim10Small cell cluster number permacrocell 3
User number per macrocell 60
Antenna gain of macrocells 17 dBiAntenna gain of small cells 5 dBiTraffic model FTP Model 1Power spectrum density of thermalnoise minus174 dBmHz
User throughput threshold 5Mbps
a more straightforward notation about howmuch traffic loadcould be offloaded to improve the energy efficiency
In order to evaluate the performances of the proposedGOscheme we compare it with the TOFFR algorithm proposedin [13] and the incentivized scheme proposed in [14] whichhave been introduced in the related works The simulationresults are given as below
42 Impacts of Small Cell Numbers on Energy EfficiencyAccording to 3GPP simulation assumptions there are 23UEs distributed in the coverage of small cell cluster whilethe remainingUEs are uniformly distributed in the remainingarea of macrocells In this section the impacts of deploymentdensity of small cells in a cluster on the system energyefficiency are investigated According to 3GPP simulationmethodology the number of small cells in a cluster variesfrom 4 to 10
As shown in Figure 2 the system energy efficiency ofdifferent algorithms versus small cell numbers per cluster isdemonstrated We can observe that the system energy effi-ciency is improved with the increasing of small cell numbersThe reason lies that small cells usually can provide higherenergy efficiency than macrocells due to lower transmissionpower attenuations in hot spot deployment scenarios Whenthere exist more small cells inside one macrocell more usertraffic could be offloaded potentially to small cellsThereforehigher system energy efficiency could be further achieved
But the system energy efficiency increases slightly whenthe number of small cells is large This is because when thesmall cells are deployed more densely the intercell interfer-enceswill bemore severe indicating the requirement of largertransmission power to ensure the same user throughputMoreover from results in Figure 2 we have proved that theproposedGO scheme outperforms the TOFFR algorithm andincentivized scheme regardless of small cell density deployed
8 Mobile Information Systems
5 6 7 8 9 104Number of small cells
TOFFRwo offloading
GO schemeIncentivized scheme
5
6
7
8
9
10
11
12
13Sy
stem
ener
gy effi
cien
cy (k
bps
W)
Figure 2 System energy efficiency versus small cell numbers percluster
per cluster In the TOFFR algorithm the fractional frequencyreuse scheme is adopted to improve UE performances betterin the cell edge The UE located in the central area cannotcontribute to the energy efficiency improvement The incen-tivized scheme focuses on maximizing the offloading utilitythat purchases the available unused bandwidth in femtocellsSo the proposed GO scheme has better energy efficiencyperformance than both of them
We also compare these three offloading algorithms withthe situation of no offloading (denoted as wo offloading insimulations) as the baseline It is obvious thatwhen offloadingschemes are adopted more user traffic originally served bymacrocell will be actively offloaded to small cells There-fore all of the offloading schemes including the proposedGO scheme TOFFR algorithm and incentivized algorithmachieve higher system energy efficiency
43 Impacts of Small Cell Numbers on Offloading GainAs shown in Figure 3 the offloading gain versus differentsmall cell numbers per cluster is dipicted We can observethat the offloading gain increases with small cell numbersdue to the capacity growth with the increase of small cellnumbers per cluster Besides in order to maximize thesystem energy efficiency there is limitation on the amountof offloaded throughputs as demonstrated in Figure 3Whenthe offloading gain reaches 62 the rise against the small cellnumbers becomes rather slowThe simulation results furthersuggest that the proposed GO scheme outperforms theTOFFR algorithm and incentivized scheme in terms of notonly energy efficiency but also the offloading gain becausein TOFFR algorithm and incentivized scheme the offloadingis mainly focused on the cell edge users which limits theperformance improvements Moreover Figure 3 shows thatthe increasing of the small cell numbers of all these threeschemes will reach a plateau The reason lies that there arealways several specific users out of the coverage of the small
TOFFRGO schemeIncentivized scheme
4 5 6 7 8 9 10 113Number of small cells per cluster
0
10
20
30
40
50
60
70
Offl
oadi
ng g
ain
()
Figure 3 Offloading gain versus small cell numbers per cluster
91011121314151617181920
Smal
l cel
l clu
ster t
hrou
ghpu
t (M
bps)
5 6 7 8 9 104Number of small cells per cluster
TOFFRwo offloading
GO schemeIncentivized scheme
Figure 4 Small cell throughput versus small cell numbers percluster
cell clusters in the simulations Just as mentioned before23 UEs are deployed in the coverage of small cell clusterand the simulation results also show that the offloading gainlimitation of all three algorithms can only reach to near23 which in turn indicates the offloading limitation will bedecided by the distribution and location of users
44 Impact of Small Cell Numbers on Small Cell ThroughputIn this section the impacts on the throughput of small cellcluster versus different small cell numbers are investigatedFrom Figure 4 we can see the throughput increases with theincrease of small cell numbers per cluster The reason is thatmore users will be served by the small cells after offloadingprocesses Moreover the throughput increments grow slowly
Mobile Information Systems 9
when small cell numbers per cluster are relatively large Thisis because of the increase of intercell interferences caused bydenser small cell deployments Finally the results in Figure 4prove that the proposed GO scheme outperforms TOFFRalgorithm and incentivized algorithm also in terms of smallcell cluster throughput
5 Conclusion
This paper aims to solve the problem regarding how toperformoffloading in the small cell HetNet deployments withoptimization on maximizing the system energy efficiencyThe reverse auction theory has been implemented with theproposed GO scheme design to decide the offloading targetBS or BSs with coordination transmission enabled technol-ogy The reverse auction model is formulated by multiplesellers (BSs) and a single buyer (offloading user) with thefirst price sealed bid mechanism The BS coordination trans-missions are also supported for multiple winning biddersscenarios According to the proposed reverse auction basedGO scheme the energy efficiency optimization problemwith constraints of user guaranteed throughput thresholdbandwidth occupation and transmission power limitationis solved by Dynamic Programming method with KKTconditions The Individual Rationality and Truthfulness ofthe VCG based reverse auction model are also proved in thepaper System-level simulations have been conducted to ver-ify the effectiveness of the proposed GO scheme according to3GPP LTE-Advanced evaluation methodologies The perfor-mances when applying theGO scheme comparison schemesand the baseline without offloading situation are evaluatedwith performance metrics of energy efficiency offloadinggain and throughput The simulation results prove that theproposed GO scheme can achieve supreme performances
Notation Definition
b = b1 b2 b119894 b119897 BidsB = B
1B2 B
119895 B
119899 Bandwidth allocation results
P = P1P2 P
119895 P
119899 Power allocation results
B119895= 1198611
119895 1198612
119895 119861
119897119895
119895 119895th BS bandwidth allocation
resultsP119895= 1198751
119895 1198752
119895 119875
119897119895
119895 119895th BS power allocation
results119861119894
119895 119894th subcarrier in the 119895th BS
119875119894
119895 Transmission power on 119861119894
119895
b119899= b1 b2 b119899 Bids sent by first 119899 BSs
120578 Energy efficiency
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper is supported by the National High Technol-ogy Research and Development Program of China no2014AA01A701 Nature and Science Foundation of China
under Grants nos 61471068 and 61421061 InternationalCollaboration Project no 2015DFT10160 andNationalMajorProject no 2016ZX03001009-003
References
[1] T Nakamura S Nagata A Benjebbour et al ldquoTrends in smallcell enhancements in LTE advancedrdquo IEEE CommunicationsMagazine vol 51 no 2 pp 98ndash105 2013
[2] W Ni and I B Collings ldquoA new adaptive small-cell architec-turerdquo IEEE Journal on Selected Areas in Communications vol31 no 5 pp 829ndash839 2013
[3] L Hanzo H Haas S Imre D OrsquoBrien M Rupp and LGyongyosi ldquoWireless myths realities and futures from 3G4Gto optical and quantum wirelessrdquo Proceedings of the IEEE vol100 pp 1853ndash1888 2012
[4] TheMETIS 2020 ProjectmdashLaying the Foundation of 5G httpswwwmetis2020com
[5] ldquoSmall cell market statusrdquo White Paper Informa and Small CellForum 1 2013
[6] D Calin H Claussen and H Uzunalioglu ldquoOn femto deploy-ment architectures and macrocell offloading benefits in jointmacro-femto deploymentsrdquo IEEE Communications Magazinevol 48 no 1 pp 26ndash32 2010
[7] S-I Sou ldquoMobile data offloading with policy and chargingcontrol in 3GPP core networkrdquo IEEE Transactions on VehicularTechnology vol 62 no 7 pp 3481ndash3486 2013
[8] J Korhonen T Savolainen A Y Ding and M Kojo ldquoTowardnetwork controlled IP traffic offloadingrdquo IEEE CommunicationsMagazine vol 51 no 3 pp 96ndash102 2013
[9] H Elsawy E Hossain and S Camorlinga ldquoTraffic offloadingtechniques in two-tier femtocell networksrdquo in Proceedings of theIEEE International Conference onCommunications (ICC rsquo13) pp6086ndash6090 IEEE Budapest Hungary June 2013
[10] Z Lu P Sinha and R Srikant ldquoEasyBid enabling cellularoffloading via small playersrdquo in Proceedings of the 33rd IEEEConference on Computer Communications (IEEE INFOCOMrsquo14) pp 691ndash699 IEEE Toronto Canada May 2014
[11] M Usman A Vastberg and T Edler ldquoEnergy efficient highcapacityHETNETby offloading highQoSusers through femtordquoin Proceedings of the 17th IEEE International Conference onNetworks (ICON rsquo11) pp 19ndash24 Singapore December 2011
[12] P Chandhar and S S Das ldquoAnalytical evaluation of offloadinggain in macrocell-femtocell OFDMA networksrdquo in Proceedingsof the IEEE 77th Vehicular Technology Conference (VTC Springrsquo13) pp 1ndash6 June 2013
[13] Q Liu G Feng and S Qin ldquoEnergy-efficient traffic offloadingin Macro-Pico networksrdquo in Proceedings of the 22nd Wirelessand Optical Communications Conference (WOCC rsquo13) pp 236ndash241 IEEE Chongqing China May 2013
[14] Y Jia M Zhao K Wang and W Zhou ldquoAn incentivizedoffloading mechanism via truthful auction in heterogeneousnetworksrdquo in Proceedings of the 6th International Conference onWireless Communications and Signal Processing (WCSP rsquo14) pp1ndash6 Hefei China October 2014
[15] L Gao G Iosifidis J Huang L Tassiulas and D Li ldquoBargain-ing-based mobile data offloadingrdquo IEEE Journal on SelectedAreas in Communications vol 32 no 6 pp 1114ndash1125 2014
[16] F Zhang W Zhang and Q Ling ldquoNon-cooperative game forcapacity offloadrdquo IEEE Transactions on Wireless Communica-tions vol 11 no 4 pp 1565ndash1575 2012
10 Mobile Information Systems
[17] X Xu H Zhang X Dai and X Tao ldquoOptimal Energy EfficientOffloading in small cell HetNet with auctionrdquo in Proceedingsof the 9th International Conference on Communications andNetworking in China (CHINACOM rsquo14) pp 335ndash340MaomingChina August 2014
[18] D P Bertsekas D A Castanon and H Tsaknakis ldquoReverseauction and the solution of inequality constrained assignmentproblemsrdquo SIAM Journal on Optimization vol 3 no 2 pp 268ndash297 1993
[19] M Khaledi and A A Abouzeid ldquoDynamic spectrum sharingauction with time-evolving channel qualitiesrdquo IEEE Transac-tions onWireless Communications vol 14 no 11 pp 5900ndash59122015
[20] F Shen D Li P-H Lin and E Jorswieck ldquoAuction basedspectrum sharing for hybrid access in macro-femtocell net-works under QoS requirementsrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC rsquo15) pp3335ndash3340 IEEE London UK June 2015
[21] X Zhuo W Gao G Cao and S Hua ldquoAn incentive frameworkfor cellular traffic offloadingrdquo IEEE Transactions on MobileComputing vol 13 no 3 pp 541ndash555 2014
[22] W Vickrey ldquoCounterspeculation auctions and competitivesealed tendersrdquo The Journal of Finance vol 16 no 1 pp 8ndash371961
[23] 3GPP-TR36814 (v1110) ldquoEvolved Universal Terrestrial RadioAccess (E-UTRA) Further advancements for E-UTRA physicallayer aspectsrdquo 2013
[24] 3GPP ldquoSmall cell enhancements for E-UTRA and EUTRANphysical layer aspects (release 12)rdquo 3GPP TR 36872 2013
[25] 3GPP R1-130744 ldquoWF on evaluation assumptions for SCEphysical layerrdquo Huawei HiSilicon CATR CMCC 2013
[26] 3GPP ldquoStudy on small cell enhancements for EUTRA and E-UTRAN higher layer aspectsrdquo 3GPP TR 36842 (v1200) 2013
[27] 3GPP TS 36300 ldquoTechnical Specification Group Radio AccessNetwork Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Net-work (EUTRAN) Overall description Stage 2 (Release 12)rdquo2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mobile Information Systems 5
119861119894
119895isin 0 119861 (3)
119875119894
119895isin P119898P119904 (4)
P119898=
119875119894
119895| 0 le 119875
119894
119895le 119875119898
119897119895
sum
119894=1
119875119894
119895le 119875119898
(5)
P119904=
119875119894
119895| 0 le 119875
119894
119895le 119875119904
119897119895
sum
119894=1
119875119894
119895le 119875119904
(6)
In (1) 119862system = sum119899
119895=1119862119895sum119897119895
119894=1119861119894
119895denotes the throughput
119875119905= sum119897119895
119894=1119875119894
119895sum119899
119895=1119897119895denotes the expectation of transmission
power on each RB from all the bidders 119862system119875119905 is thesystem energy efficiency and 119875
119905gt 0
In (2) 119867119895= 119889minus120572119895
119895|ℎ119895|2120594119895is the channel gain between
the 119895th BS and UE where 119889119895is the distance between the
119895th BS and UE 120572119895is the path-loss exponent of the 119895th
BS ℎ119895is the Rayleigh fading component 120594
119895denotes the
log-normally distributed shadow fading Furthermore 119868119894119895=
sum119899
1198951015840=11198951015840=119895119875119894
11989510158401198671198951015840V119894
1198951015840 is the interference experienced by the UE
on the 119894th RB where 1198751198941198951015840 is the transmission power on the 119894th
RB from the 1198951015840th BS1198671198951015840 is the channel gain between the 1198951015840th
BS andUE V1198941198951015840 is the binary variables representing the activity
factor of 1198951015840th BS 1198730is the thermal noise level and 120591
119903is the
guaranteed throughput threshold of the offloading UEConstraint (3) means the RB 119861119894
119895can be occupied or
vacant Constraint (4) denotes the bidder can be a macrocellBS or a small cell BS In (5) and (6) we give the requirementfor the transmission power in the corresponding RB where119875119898and 119875119904are the power limitations for the macrocell BS and
small cell BS respectivelyFor the convenience of solving this problem we trans-
form (1) (2) and (3) into the following form
minB119895P119895
119875119905
119862system
119862system ge 120591119903
0 le 119861119894
119895le 119861
(7)
Assuming the 119862system is an independent variable theKarush-Kuhn-Tucker (KKT) conditions are given as
nablaB119895P119895system119875119905
119862systemminus nablaB119895P119895system (119862system minus 120591119903) 120583
minus nablaB119895 P119895system119861119894
119895120590119894
119895minus nablaB119895P119895system (119861 minus 119861
119894
119895) 120585119894
119895
minus nablaB119895 P119895system119875119894
119895]119894119895minus nablaB119895P119895system (119875119904 (119875119898) minus 119875
119894
119895) 120582119894
119895
minus nablaB119895 P119895system(119875119904 (119875119898) minus
119897119895
sum
119894=1
119875119894
119895)120588119895= 0
(8)
(119862system minus 120591119903) 120583 = 0 (9)
119861119894
119895120590119894
119895= 0 119861
119894
119895ge 0 120590
119894
119895ge 0 (10)
(119861 minus 119861119894
119895) 120585119894
119895= 0 119861 le 119861
119894
119895 120585119894
119895ge 0 (11)
119875119894
119895]119894119895= 0 119875
119894
119895ge 0 ]119894
119895ge 0 (12)
(119875119904(119875119898) minus 119875119894
119895) 120582119894
119895= 0 119875
119894
119895le 119875119904(119875119898) 120582119894
119895ge 0 (13)
(119875119904(119875119898) minus
119897119895
sum
119894=1
119875119894
119895)120588119895= 0
119897119895
sum
119894=1
119875119894
119895le 119875119904(119875119898) 120588119895ge 0 (14)
Equation (8) is changed into (15) by taking the derivativeof 119862system as
minus119875119905
1198622systemminus 120583 = 0 (15)
119875119905
1198622system+ 120583 = 0 (16)
Thismakes sensewhere119875119905and1198622system are nonzero values
Thus 120583 = 0 (9) shows (119862system minus 120591119903)120583 = 0 where
119862system = 120591119903 (17)
When the offloading process is triggered the offloadeduser traffic requirements and the number of BSs participatingin the auction are known The KKT conditions prove that(9) is guaranteed throughput requirement The 119860119897119897119900119888119886119905119894119900119899problem is transformed to solve minB119895P119895119875119905 which can be
simplified into a linear objective functionminsum119897119895119894=1119875119894
119895sum119899
119895=1119897119895
Because the above119860119897119897119900119888119886119905119894119900119899problem is a linear problemit is easy to find that this is a multiple knapsack problemIn order to facilitate the solving process we turn it intoa 0-1 knapsack problem There are 119899 BSs to participate inthe auction the 119894th BS resources are separated into 119872
119894
independent piece of ldquogoodsrdquo that can be loaded into aknapsack This will get a 0-1 knapsack problem where thenumber of items is sum119872
119894 If we solve this problem directly
the computational complexity will be 119874(119862sum119872119894) In order to
reduce the complexity we design another algorithm with thespecific plan as follows
As mentioned above the 119894th BS bandwidth resourceis composed of a number of RB groups Considering thecondition of the binary the guaranteeing of selecting anymultiple resource package strategy still can be achievedafter the transformation of the original multiple knapsackproblem One BS which has119872
119894resource blocks is separated
into several RB groups where these RB groups respectivelyhave 1 2 22 23 119872
119894minus 2119896minus1+ 1 resource blocks The 119894th BS
has Ceiling(log119872119894) different RB groups participating in the
0-1 knapsack problemInitialize the value as follows the weight of knapsack is
119862 which is the offloading user traffic requirements 120591119903 The
starting value 119865[0 119888] = 0 119865[119909 0] = 0 And the 119894th BS hasCeiling(log119872
119894) stages
6 Mobile Information Systems
(1) Let 1205780= 0
(2) for 119895 = 1 to 119899 do(3) if Bidder b119895 is a macrocell BS then(4) for 119875 = 1 to 10119875
119898do
(5) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(6) end for(7) for 119894 = 1 to 119897
119895do
(8) if 119875119894119895gt 0 then
(9) 119861119894
119895= 119861
(10) else(11) 119861
119894
119895= 0
(12) end if(13) end for(14) else if Bidder b119895 is a small cell BS then(15) for 119875 = 1 to 10119875
119904do
(16) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(17) end for(18) for 119894 = 1 to 119897
119895do
(19) if 119875119894119895gt 0 then
(20) 119861119894
119895= 119861
(21) else(22) 119861
119894
119895= 0
(23) end if(24) end for(25) end if(26) end for
Algorithm 1 Reverse auction 119860119897119897119900119888119886119905119894119900119899 (119899 b119899)
Renumber all RB groups 119888119909is the capacity of the 119909
stage which denotes the 119894th RB group of the 119895th BS and thecorresponding value is V
119909
The capacity of the 119909 stage is
119888119909= 119908 [119894 119895] = 119861
119894
119895log(1 +
119875119894
119895119867119895
119868119894
119895+ 1198730119861119894
119895
) (18)
The value of the corresponding V119909is
V119909= V [119894 119895] = 119875119894
119895 (19)
The iterative equation will be
119865 [119909 119888] = min 119865 [119909 minus 1 119888] 119865 [119909 minus 1 119888 minus 119888119909] + V119909 (20)
In (20) 119865[119909 119888] is the minimum value of the transmissionpower in the stage 119909 119888 denotes the remaining space ofthe pack according to the current stage 119888
119909denotes the
provided capacity when choosing the 119909 stage V119909denotes the
provided transmission power when choosing the 119909 stageThecomputational complexity is reduced to 119874(119862sum log119872
119894)
The proposed 119860119897119897119900119888119886119905119894119900119899 algorithm is illustrated inAlgorithm 1 with B = B
1B2 B
119899 and P = P
1P2
P119899In Algorithm 1 the transmission power of each bidder is
chosen firstly and the corresponding bandwidth is decidedbased on the transmission power allocation results As
for 119895 = 1 to 119899 do(2) if 119895th BS is a winning bidder then
Reverse Auction - 119860119897119897119900119888119886119905119894119900119899 (119899 119895B b119894)(4) 119901
119895= 120578B119899b119894 minus (120578B119899 minus 120578b119894 )
else(6) 119901
119895= 0
end if(8) end for
Algorithm 2 Reverse auction 119875119903119894119888119894119899119892 (119899 b119899BP 120578)
mentioned before 01W is chosen as the basic transmissionpower unit and the integer data is needed in the DynamicProgramming Therefore the range of 119875 is defined from 1
to 10119875119898or 10119875
119904 The equation in Line (5) and Line (16) of
Algorithm 1 mean that the 119895th bidderrsquos transmission powershould be chosen to achieve the largest 120578b119895 and guarantee theoffloaded userrsquos throughput threshold 120591
119903at the same time
where 120578b119895 denotes the energy efficiency of the 119895th bidderAfter implementing the Dynamic Programming the resultsof Algorithm 1 P10 and B will be the optimal allocationsolution for the proposed reverse auction process
322 Pricing In traditional 119875119903119894119888119894119899119892 algorithms the biddersare encouraged to set their own bids truthfully as illustratedbefore So in this paper the same energy efficiency that thecorresponding bidder achieves is paid back With regard tothe offloading user throughput threshold 120591
119903 we define 120578
1and
1205782as (21) and (22) as follows
1205781= 120578B119899b119894 = max
B119895B119894P119895P119894
119862system
119864 (119875119905)
(21)
1205782= 120578B119899 minus 120578b119894 = (max
B119895P119895
119862system
119864 (119875119905)) minus
119862b119894
119864 (P119894) (22)
where 1205781denotes the system energy efficiency under the
optimal 119860119897119897119900119888119886119905119894119900119899 solution without the presence of the 119894thBS The 120578
2denotes the system energy efficiency except for
the 119894th BS under current optimal119860119897119897119900119888119886119905119894119900119899 resultsThen theopportunity cost of the 119894th BS is defined as the differencebetween 120578
1and 1205782 just as illustrated in (23) [19] as follows
119901119894= 1205781minus 1205782= 120578B119899b119894 minus (120578B119899 minus 120578b119894) (23)
The 119875119903119894119888119894119899119892 algorithm is given as Algorithm 2
323 Properties In this section the properties of theproposed reverse auction model are analyzed Accordingto the VCG based reverse auction model the IndividualRa-tionality and the 119879119903119906119905ℎ119891119906119897119899119890119904119904 properties need to be proved
IndividualRationality When the utility of each participatingbidder in the119875119903119894119888119894119899119892 stage is greater than zero this algorithmis individual rational for each winning bidder Namely
119901119894= 120578B119899b119894 minus (120578B119899 minus 120578b119894) ge 0 (24)
Mobile Information Systems 7
119879119903119906119905ℎ119891119906119897119899119890119904119904 For each bidder the Truthfulness means thateach bidderrsquos bid price is equal to its private value This isa weakly dominant strategy If BSrsquos bidding is untrue theenergy efficiency will be unlikely the biggest In order toget the maximum energy efficiency the allocation should beformulated as follows
119901119895= 120578B119899b119895 minus (120578B119899 minus 120578b119895)
120575 = 119901119895minus 119901119894
= 120578B119899b119895 minus (120578B119899 minus 120578b119895) minus [120578B119899b119894 minus (120578B119899 minus 120578b119894)]
= 120578B119899b119895 minus 120578B119899 + 120578b119895 minus 120578B119899b119894 + 120578B119899 minus 120578b119894
= 120578B119899b119895 + 120578b119895 minus 120578B119899b119894 minus 120578b119894
= (120578B119899b119895 + 120578b119895) minus (120578B119899b119894 + 120578b119894)
(25)
Based on the proposed model in this paper because 119901119895le
119901119894and120575 le 0 thismeans 120578B119899b119895+120578b119895 le 120578B119899b119894+120578b119894 If and only
if 119895 = 119894 it can take the equal signTherefore each bidder mustbe truthful to obtain the maximum system energy efficiencyThe proof is finished
4 Performance Evaluation
In this section we built the system-level simulation plat-form according to the 3GPP LTE-Advanced simulationmethodology [23] Based on this platform we validate theperformances of the proposed reverse auction based GOscheme with comparison algorithms in the small cell HetNetdownlink scenario
41 Simulation Setting Performance Metrics and ComparisonAlgorithms Theconsidered simulation scenario in this papercomprises 19-hexagonal macrocells with 3 sectors per macro-cell In each sector there is one small cell cluster deployedwith shared spectrum manner The small cell cluster is agroup of densely deployed small cells We deploy 23 usersin the coverage of small cell clusters while the remainingusers are distributed in the coverage area of macro cellsThe users are uniformly distributed Moreover as mentionedabove the bandwidth resource granularity in the simulationis one RB In the initial state each user is served by the BSwhich can provide the highest downlink RSRP Once a newtraffic offloading requirement is requested the reverse auc-tion based GO scheme is triggered The detailed simulationparameters are according to 3GPP LTE-Advanced small cellHetNet evaluation methodology [24] These parameters arelisted in Table 1
The performance metrics include the system energyefficiency offloading gain and throughput In this paper themetric of offloading gain (120574gain) is defined as
120574gain =120591offloading
120591total (26)
where 120591offloading denotes the offloaded throughput and 120591totaldenotes the total system throughput The offloading gain is
Table 1 Simulation parameters
Simulation parameter ValueCarrier frequency 2GHzSystem bandwidth 10MHzTotal transmission power ofmacrocell 46 dBm
Total transmission power of small cell 30 dBmPath-loss of macrocell Pl = 283 + 220log
10(119889)
Path-loss of small cell Pl = 305 + 367log10(119889)
Small cell number per cluster 4sim10Small cell cluster number permacrocell 3
User number per macrocell 60
Antenna gain of macrocells 17 dBiAntenna gain of small cells 5 dBiTraffic model FTP Model 1Power spectrum density of thermalnoise minus174 dBmHz
User throughput threshold 5Mbps
a more straightforward notation about howmuch traffic loadcould be offloaded to improve the energy efficiency
In order to evaluate the performances of the proposedGOscheme we compare it with the TOFFR algorithm proposedin [13] and the incentivized scheme proposed in [14] whichhave been introduced in the related works The simulationresults are given as below
42 Impacts of Small Cell Numbers on Energy EfficiencyAccording to 3GPP simulation assumptions there are 23UEs distributed in the coverage of small cell cluster whilethe remainingUEs are uniformly distributed in the remainingarea of macrocells In this section the impacts of deploymentdensity of small cells in a cluster on the system energyefficiency are investigated According to 3GPP simulationmethodology the number of small cells in a cluster variesfrom 4 to 10
As shown in Figure 2 the system energy efficiency ofdifferent algorithms versus small cell numbers per cluster isdemonstrated We can observe that the system energy effi-ciency is improved with the increasing of small cell numbersThe reason lies that small cells usually can provide higherenergy efficiency than macrocells due to lower transmissionpower attenuations in hot spot deployment scenarios Whenthere exist more small cells inside one macrocell more usertraffic could be offloaded potentially to small cellsThereforehigher system energy efficiency could be further achieved
But the system energy efficiency increases slightly whenthe number of small cells is large This is because when thesmall cells are deployed more densely the intercell interfer-enceswill bemore severe indicating the requirement of largertransmission power to ensure the same user throughputMoreover from results in Figure 2 we have proved that theproposedGO scheme outperforms the TOFFR algorithm andincentivized scheme regardless of small cell density deployed
8 Mobile Information Systems
5 6 7 8 9 104Number of small cells
TOFFRwo offloading
GO schemeIncentivized scheme
5
6
7
8
9
10
11
12
13Sy
stem
ener
gy effi
cien
cy (k
bps
W)
Figure 2 System energy efficiency versus small cell numbers percluster
per cluster In the TOFFR algorithm the fractional frequencyreuse scheme is adopted to improve UE performances betterin the cell edge The UE located in the central area cannotcontribute to the energy efficiency improvement The incen-tivized scheme focuses on maximizing the offloading utilitythat purchases the available unused bandwidth in femtocellsSo the proposed GO scheme has better energy efficiencyperformance than both of them
We also compare these three offloading algorithms withthe situation of no offloading (denoted as wo offloading insimulations) as the baseline It is obvious thatwhen offloadingschemes are adopted more user traffic originally served bymacrocell will be actively offloaded to small cells There-fore all of the offloading schemes including the proposedGO scheme TOFFR algorithm and incentivized algorithmachieve higher system energy efficiency
43 Impacts of Small Cell Numbers on Offloading GainAs shown in Figure 3 the offloading gain versus differentsmall cell numbers per cluster is dipicted We can observethat the offloading gain increases with small cell numbersdue to the capacity growth with the increase of small cellnumbers per cluster Besides in order to maximize thesystem energy efficiency there is limitation on the amountof offloaded throughputs as demonstrated in Figure 3Whenthe offloading gain reaches 62 the rise against the small cellnumbers becomes rather slowThe simulation results furthersuggest that the proposed GO scheme outperforms theTOFFR algorithm and incentivized scheme in terms of notonly energy efficiency but also the offloading gain becausein TOFFR algorithm and incentivized scheme the offloadingis mainly focused on the cell edge users which limits theperformance improvements Moreover Figure 3 shows thatthe increasing of the small cell numbers of all these threeschemes will reach a plateau The reason lies that there arealways several specific users out of the coverage of the small
TOFFRGO schemeIncentivized scheme
4 5 6 7 8 9 10 113Number of small cells per cluster
0
10
20
30
40
50
60
70
Offl
oadi
ng g
ain
()
Figure 3 Offloading gain versus small cell numbers per cluster
91011121314151617181920
Smal
l cel
l clu
ster t
hrou
ghpu
t (M
bps)
5 6 7 8 9 104Number of small cells per cluster
TOFFRwo offloading
GO schemeIncentivized scheme
Figure 4 Small cell throughput versus small cell numbers percluster
cell clusters in the simulations Just as mentioned before23 UEs are deployed in the coverage of small cell clusterand the simulation results also show that the offloading gainlimitation of all three algorithms can only reach to near23 which in turn indicates the offloading limitation will bedecided by the distribution and location of users
44 Impact of Small Cell Numbers on Small Cell ThroughputIn this section the impacts on the throughput of small cellcluster versus different small cell numbers are investigatedFrom Figure 4 we can see the throughput increases with theincrease of small cell numbers per cluster The reason is thatmore users will be served by the small cells after offloadingprocesses Moreover the throughput increments grow slowly
Mobile Information Systems 9
when small cell numbers per cluster are relatively large Thisis because of the increase of intercell interferences caused bydenser small cell deployments Finally the results in Figure 4prove that the proposed GO scheme outperforms TOFFRalgorithm and incentivized algorithm also in terms of smallcell cluster throughput
5 Conclusion
This paper aims to solve the problem regarding how toperformoffloading in the small cell HetNet deployments withoptimization on maximizing the system energy efficiencyThe reverse auction theory has been implemented with theproposed GO scheme design to decide the offloading targetBS or BSs with coordination transmission enabled technol-ogy The reverse auction model is formulated by multiplesellers (BSs) and a single buyer (offloading user) with thefirst price sealed bid mechanism The BS coordination trans-missions are also supported for multiple winning biddersscenarios According to the proposed reverse auction basedGO scheme the energy efficiency optimization problemwith constraints of user guaranteed throughput thresholdbandwidth occupation and transmission power limitationis solved by Dynamic Programming method with KKTconditions The Individual Rationality and Truthfulness ofthe VCG based reverse auction model are also proved in thepaper System-level simulations have been conducted to ver-ify the effectiveness of the proposed GO scheme according to3GPP LTE-Advanced evaluation methodologies The perfor-mances when applying theGO scheme comparison schemesand the baseline without offloading situation are evaluatedwith performance metrics of energy efficiency offloadinggain and throughput The simulation results prove that theproposed GO scheme can achieve supreme performances
Notation Definition
b = b1 b2 b119894 b119897 BidsB = B
1B2 B
119895 B
119899 Bandwidth allocation results
P = P1P2 P
119895 P
119899 Power allocation results
B119895= 1198611
119895 1198612
119895 119861
119897119895
119895 119895th BS bandwidth allocation
resultsP119895= 1198751
119895 1198752
119895 119875
119897119895
119895 119895th BS power allocation
results119861119894
119895 119894th subcarrier in the 119895th BS
119875119894
119895 Transmission power on 119861119894
119895
b119899= b1 b2 b119899 Bids sent by first 119899 BSs
120578 Energy efficiency
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper is supported by the National High Technol-ogy Research and Development Program of China no2014AA01A701 Nature and Science Foundation of China
under Grants nos 61471068 and 61421061 InternationalCollaboration Project no 2015DFT10160 andNationalMajorProject no 2016ZX03001009-003
References
[1] T Nakamura S Nagata A Benjebbour et al ldquoTrends in smallcell enhancements in LTE advancedrdquo IEEE CommunicationsMagazine vol 51 no 2 pp 98ndash105 2013
[2] W Ni and I B Collings ldquoA new adaptive small-cell architec-turerdquo IEEE Journal on Selected Areas in Communications vol31 no 5 pp 829ndash839 2013
[3] L Hanzo H Haas S Imre D OrsquoBrien M Rupp and LGyongyosi ldquoWireless myths realities and futures from 3G4Gto optical and quantum wirelessrdquo Proceedings of the IEEE vol100 pp 1853ndash1888 2012
[4] TheMETIS 2020 ProjectmdashLaying the Foundation of 5G httpswwwmetis2020com
[5] ldquoSmall cell market statusrdquo White Paper Informa and Small CellForum 1 2013
[6] D Calin H Claussen and H Uzunalioglu ldquoOn femto deploy-ment architectures and macrocell offloading benefits in jointmacro-femto deploymentsrdquo IEEE Communications Magazinevol 48 no 1 pp 26ndash32 2010
[7] S-I Sou ldquoMobile data offloading with policy and chargingcontrol in 3GPP core networkrdquo IEEE Transactions on VehicularTechnology vol 62 no 7 pp 3481ndash3486 2013
[8] J Korhonen T Savolainen A Y Ding and M Kojo ldquoTowardnetwork controlled IP traffic offloadingrdquo IEEE CommunicationsMagazine vol 51 no 3 pp 96ndash102 2013
[9] H Elsawy E Hossain and S Camorlinga ldquoTraffic offloadingtechniques in two-tier femtocell networksrdquo in Proceedings of theIEEE International Conference onCommunications (ICC rsquo13) pp6086ndash6090 IEEE Budapest Hungary June 2013
[10] Z Lu P Sinha and R Srikant ldquoEasyBid enabling cellularoffloading via small playersrdquo in Proceedings of the 33rd IEEEConference on Computer Communications (IEEE INFOCOMrsquo14) pp 691ndash699 IEEE Toronto Canada May 2014
[11] M Usman A Vastberg and T Edler ldquoEnergy efficient highcapacityHETNETby offloading highQoSusers through femtordquoin Proceedings of the 17th IEEE International Conference onNetworks (ICON rsquo11) pp 19ndash24 Singapore December 2011
[12] P Chandhar and S S Das ldquoAnalytical evaluation of offloadinggain in macrocell-femtocell OFDMA networksrdquo in Proceedingsof the IEEE 77th Vehicular Technology Conference (VTC Springrsquo13) pp 1ndash6 June 2013
[13] Q Liu G Feng and S Qin ldquoEnergy-efficient traffic offloadingin Macro-Pico networksrdquo in Proceedings of the 22nd Wirelessand Optical Communications Conference (WOCC rsquo13) pp 236ndash241 IEEE Chongqing China May 2013
[14] Y Jia M Zhao K Wang and W Zhou ldquoAn incentivizedoffloading mechanism via truthful auction in heterogeneousnetworksrdquo in Proceedings of the 6th International Conference onWireless Communications and Signal Processing (WCSP rsquo14) pp1ndash6 Hefei China October 2014
[15] L Gao G Iosifidis J Huang L Tassiulas and D Li ldquoBargain-ing-based mobile data offloadingrdquo IEEE Journal on SelectedAreas in Communications vol 32 no 6 pp 1114ndash1125 2014
[16] F Zhang W Zhang and Q Ling ldquoNon-cooperative game forcapacity offloadrdquo IEEE Transactions on Wireless Communica-tions vol 11 no 4 pp 1565ndash1575 2012
10 Mobile Information Systems
[17] X Xu H Zhang X Dai and X Tao ldquoOptimal Energy EfficientOffloading in small cell HetNet with auctionrdquo in Proceedingsof the 9th International Conference on Communications andNetworking in China (CHINACOM rsquo14) pp 335ndash340MaomingChina August 2014
[18] D P Bertsekas D A Castanon and H Tsaknakis ldquoReverseauction and the solution of inequality constrained assignmentproblemsrdquo SIAM Journal on Optimization vol 3 no 2 pp 268ndash297 1993
[19] M Khaledi and A A Abouzeid ldquoDynamic spectrum sharingauction with time-evolving channel qualitiesrdquo IEEE Transac-tions onWireless Communications vol 14 no 11 pp 5900ndash59122015
[20] F Shen D Li P-H Lin and E Jorswieck ldquoAuction basedspectrum sharing for hybrid access in macro-femtocell net-works under QoS requirementsrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC rsquo15) pp3335ndash3340 IEEE London UK June 2015
[21] X Zhuo W Gao G Cao and S Hua ldquoAn incentive frameworkfor cellular traffic offloadingrdquo IEEE Transactions on MobileComputing vol 13 no 3 pp 541ndash555 2014
[22] W Vickrey ldquoCounterspeculation auctions and competitivesealed tendersrdquo The Journal of Finance vol 16 no 1 pp 8ndash371961
[23] 3GPP-TR36814 (v1110) ldquoEvolved Universal Terrestrial RadioAccess (E-UTRA) Further advancements for E-UTRA physicallayer aspectsrdquo 2013
[24] 3GPP ldquoSmall cell enhancements for E-UTRA and EUTRANphysical layer aspects (release 12)rdquo 3GPP TR 36872 2013
[25] 3GPP R1-130744 ldquoWF on evaluation assumptions for SCEphysical layerrdquo Huawei HiSilicon CATR CMCC 2013
[26] 3GPP ldquoStudy on small cell enhancements for EUTRA and E-UTRAN higher layer aspectsrdquo 3GPP TR 36842 (v1200) 2013
[27] 3GPP TS 36300 ldquoTechnical Specification Group Radio AccessNetwork Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Net-work (EUTRAN) Overall description Stage 2 (Release 12)rdquo2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
6 Mobile Information Systems
(1) Let 1205780= 0
(2) for 119895 = 1 to 119899 do(3) if Bidder b119895 is a macrocell BS then(4) for 119875 = 1 to 10119875
119898do
(5) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(6) end for(7) for 119894 = 1 to 119897
119895do
(8) if 119875119894119895gt 0 then
(9) 119861119894
119895= 119861
(10) else(11) 119861
119894
119895= 0
(12) end if(13) end for(14) else if Bidder b119895 is a small cell BS then(15) for 119875 = 1 to 10119875
119904do
(16) P119895= argmaxb119895 120578b119895 = 120578b119895 + 120578b119895minus1
(17) end for(18) for 119894 = 1 to 119897
119895do
(19) if 119875119894119895gt 0 then
(20) 119861119894
119895= 119861
(21) else(22) 119861
119894
119895= 0
(23) end if(24) end for(25) end if(26) end for
Algorithm 1 Reverse auction 119860119897119897119900119888119886119905119894119900119899 (119899 b119899)
Renumber all RB groups 119888119909is the capacity of the 119909
stage which denotes the 119894th RB group of the 119895th BS and thecorresponding value is V
119909
The capacity of the 119909 stage is
119888119909= 119908 [119894 119895] = 119861
119894
119895log(1 +
119875119894
119895119867119895
119868119894
119895+ 1198730119861119894
119895
) (18)
The value of the corresponding V119909is
V119909= V [119894 119895] = 119875119894
119895 (19)
The iterative equation will be
119865 [119909 119888] = min 119865 [119909 minus 1 119888] 119865 [119909 minus 1 119888 minus 119888119909] + V119909 (20)
In (20) 119865[119909 119888] is the minimum value of the transmissionpower in the stage 119909 119888 denotes the remaining space ofthe pack according to the current stage 119888
119909denotes the
provided capacity when choosing the 119909 stage V119909denotes the
provided transmission power when choosing the 119909 stageThecomputational complexity is reduced to 119874(119862sum log119872
119894)
The proposed 119860119897119897119900119888119886119905119894119900119899 algorithm is illustrated inAlgorithm 1 with B = B
1B2 B
119899 and P = P
1P2
P119899In Algorithm 1 the transmission power of each bidder is
chosen firstly and the corresponding bandwidth is decidedbased on the transmission power allocation results As
for 119895 = 1 to 119899 do(2) if 119895th BS is a winning bidder then
Reverse Auction - 119860119897119897119900119888119886119905119894119900119899 (119899 119895B b119894)(4) 119901
119895= 120578B119899b119894 minus (120578B119899 minus 120578b119894 )
else(6) 119901
119895= 0
end if(8) end for
Algorithm 2 Reverse auction 119875119903119894119888119894119899119892 (119899 b119899BP 120578)
mentioned before 01W is chosen as the basic transmissionpower unit and the integer data is needed in the DynamicProgramming Therefore the range of 119875 is defined from 1
to 10119875119898or 10119875
119904 The equation in Line (5) and Line (16) of
Algorithm 1 mean that the 119895th bidderrsquos transmission powershould be chosen to achieve the largest 120578b119895 and guarantee theoffloaded userrsquos throughput threshold 120591
119903at the same time
where 120578b119895 denotes the energy efficiency of the 119895th bidderAfter implementing the Dynamic Programming the resultsof Algorithm 1 P10 and B will be the optimal allocationsolution for the proposed reverse auction process
322 Pricing In traditional 119875119903119894119888119894119899119892 algorithms the biddersare encouraged to set their own bids truthfully as illustratedbefore So in this paper the same energy efficiency that thecorresponding bidder achieves is paid back With regard tothe offloading user throughput threshold 120591
119903 we define 120578
1and
1205782as (21) and (22) as follows
1205781= 120578B119899b119894 = max
B119895B119894P119895P119894
119862system
119864 (119875119905)
(21)
1205782= 120578B119899 minus 120578b119894 = (max
B119895P119895
119862system
119864 (119875119905)) minus
119862b119894
119864 (P119894) (22)
where 1205781denotes the system energy efficiency under the
optimal 119860119897119897119900119888119886119905119894119900119899 solution without the presence of the 119894thBS The 120578
2denotes the system energy efficiency except for
the 119894th BS under current optimal119860119897119897119900119888119886119905119894119900119899 resultsThen theopportunity cost of the 119894th BS is defined as the differencebetween 120578
1and 1205782 just as illustrated in (23) [19] as follows
119901119894= 1205781minus 1205782= 120578B119899b119894 minus (120578B119899 minus 120578b119894) (23)
The 119875119903119894119888119894119899119892 algorithm is given as Algorithm 2
323 Properties In this section the properties of theproposed reverse auction model are analyzed Accordingto the VCG based reverse auction model the IndividualRa-tionality and the 119879119903119906119905ℎ119891119906119897119899119890119904119904 properties need to be proved
IndividualRationality When the utility of each participatingbidder in the119875119903119894119888119894119899119892 stage is greater than zero this algorithmis individual rational for each winning bidder Namely
119901119894= 120578B119899b119894 minus (120578B119899 minus 120578b119894) ge 0 (24)
Mobile Information Systems 7
119879119903119906119905ℎ119891119906119897119899119890119904119904 For each bidder the Truthfulness means thateach bidderrsquos bid price is equal to its private value This isa weakly dominant strategy If BSrsquos bidding is untrue theenergy efficiency will be unlikely the biggest In order toget the maximum energy efficiency the allocation should beformulated as follows
119901119895= 120578B119899b119895 minus (120578B119899 minus 120578b119895)
120575 = 119901119895minus 119901119894
= 120578B119899b119895 minus (120578B119899 minus 120578b119895) minus [120578B119899b119894 minus (120578B119899 minus 120578b119894)]
= 120578B119899b119895 minus 120578B119899 + 120578b119895 minus 120578B119899b119894 + 120578B119899 minus 120578b119894
= 120578B119899b119895 + 120578b119895 minus 120578B119899b119894 minus 120578b119894
= (120578B119899b119895 + 120578b119895) minus (120578B119899b119894 + 120578b119894)
(25)
Based on the proposed model in this paper because 119901119895le
119901119894and120575 le 0 thismeans 120578B119899b119895+120578b119895 le 120578B119899b119894+120578b119894 If and only
if 119895 = 119894 it can take the equal signTherefore each bidder mustbe truthful to obtain the maximum system energy efficiencyThe proof is finished
4 Performance Evaluation
In this section we built the system-level simulation plat-form according to the 3GPP LTE-Advanced simulationmethodology [23] Based on this platform we validate theperformances of the proposed reverse auction based GOscheme with comparison algorithms in the small cell HetNetdownlink scenario
41 Simulation Setting Performance Metrics and ComparisonAlgorithms Theconsidered simulation scenario in this papercomprises 19-hexagonal macrocells with 3 sectors per macro-cell In each sector there is one small cell cluster deployedwith shared spectrum manner The small cell cluster is agroup of densely deployed small cells We deploy 23 usersin the coverage of small cell clusters while the remainingusers are distributed in the coverage area of macro cellsThe users are uniformly distributed Moreover as mentionedabove the bandwidth resource granularity in the simulationis one RB In the initial state each user is served by the BSwhich can provide the highest downlink RSRP Once a newtraffic offloading requirement is requested the reverse auc-tion based GO scheme is triggered The detailed simulationparameters are according to 3GPP LTE-Advanced small cellHetNet evaluation methodology [24] These parameters arelisted in Table 1
The performance metrics include the system energyefficiency offloading gain and throughput In this paper themetric of offloading gain (120574gain) is defined as
120574gain =120591offloading
120591total (26)
where 120591offloading denotes the offloaded throughput and 120591totaldenotes the total system throughput The offloading gain is
Table 1 Simulation parameters
Simulation parameter ValueCarrier frequency 2GHzSystem bandwidth 10MHzTotal transmission power ofmacrocell 46 dBm
Total transmission power of small cell 30 dBmPath-loss of macrocell Pl = 283 + 220log
10(119889)
Path-loss of small cell Pl = 305 + 367log10(119889)
Small cell number per cluster 4sim10Small cell cluster number permacrocell 3
User number per macrocell 60
Antenna gain of macrocells 17 dBiAntenna gain of small cells 5 dBiTraffic model FTP Model 1Power spectrum density of thermalnoise minus174 dBmHz
User throughput threshold 5Mbps
a more straightforward notation about howmuch traffic loadcould be offloaded to improve the energy efficiency
In order to evaluate the performances of the proposedGOscheme we compare it with the TOFFR algorithm proposedin [13] and the incentivized scheme proposed in [14] whichhave been introduced in the related works The simulationresults are given as below
42 Impacts of Small Cell Numbers on Energy EfficiencyAccording to 3GPP simulation assumptions there are 23UEs distributed in the coverage of small cell cluster whilethe remainingUEs are uniformly distributed in the remainingarea of macrocells In this section the impacts of deploymentdensity of small cells in a cluster on the system energyefficiency are investigated According to 3GPP simulationmethodology the number of small cells in a cluster variesfrom 4 to 10
As shown in Figure 2 the system energy efficiency ofdifferent algorithms versus small cell numbers per cluster isdemonstrated We can observe that the system energy effi-ciency is improved with the increasing of small cell numbersThe reason lies that small cells usually can provide higherenergy efficiency than macrocells due to lower transmissionpower attenuations in hot spot deployment scenarios Whenthere exist more small cells inside one macrocell more usertraffic could be offloaded potentially to small cellsThereforehigher system energy efficiency could be further achieved
But the system energy efficiency increases slightly whenthe number of small cells is large This is because when thesmall cells are deployed more densely the intercell interfer-enceswill bemore severe indicating the requirement of largertransmission power to ensure the same user throughputMoreover from results in Figure 2 we have proved that theproposedGO scheme outperforms the TOFFR algorithm andincentivized scheme regardless of small cell density deployed
8 Mobile Information Systems
5 6 7 8 9 104Number of small cells
TOFFRwo offloading
GO schemeIncentivized scheme
5
6
7
8
9
10
11
12
13Sy
stem
ener
gy effi
cien
cy (k
bps
W)
Figure 2 System energy efficiency versus small cell numbers percluster
per cluster In the TOFFR algorithm the fractional frequencyreuse scheme is adopted to improve UE performances betterin the cell edge The UE located in the central area cannotcontribute to the energy efficiency improvement The incen-tivized scheme focuses on maximizing the offloading utilitythat purchases the available unused bandwidth in femtocellsSo the proposed GO scheme has better energy efficiencyperformance than both of them
We also compare these three offloading algorithms withthe situation of no offloading (denoted as wo offloading insimulations) as the baseline It is obvious thatwhen offloadingschemes are adopted more user traffic originally served bymacrocell will be actively offloaded to small cells There-fore all of the offloading schemes including the proposedGO scheme TOFFR algorithm and incentivized algorithmachieve higher system energy efficiency
43 Impacts of Small Cell Numbers on Offloading GainAs shown in Figure 3 the offloading gain versus differentsmall cell numbers per cluster is dipicted We can observethat the offloading gain increases with small cell numbersdue to the capacity growth with the increase of small cellnumbers per cluster Besides in order to maximize thesystem energy efficiency there is limitation on the amountof offloaded throughputs as demonstrated in Figure 3Whenthe offloading gain reaches 62 the rise against the small cellnumbers becomes rather slowThe simulation results furthersuggest that the proposed GO scheme outperforms theTOFFR algorithm and incentivized scheme in terms of notonly energy efficiency but also the offloading gain becausein TOFFR algorithm and incentivized scheme the offloadingis mainly focused on the cell edge users which limits theperformance improvements Moreover Figure 3 shows thatthe increasing of the small cell numbers of all these threeschemes will reach a plateau The reason lies that there arealways several specific users out of the coverage of the small
TOFFRGO schemeIncentivized scheme
4 5 6 7 8 9 10 113Number of small cells per cluster
0
10
20
30
40
50
60
70
Offl
oadi
ng g
ain
()
Figure 3 Offloading gain versus small cell numbers per cluster
91011121314151617181920
Smal
l cel
l clu
ster t
hrou
ghpu
t (M
bps)
5 6 7 8 9 104Number of small cells per cluster
TOFFRwo offloading
GO schemeIncentivized scheme
Figure 4 Small cell throughput versus small cell numbers percluster
cell clusters in the simulations Just as mentioned before23 UEs are deployed in the coverage of small cell clusterand the simulation results also show that the offloading gainlimitation of all three algorithms can only reach to near23 which in turn indicates the offloading limitation will bedecided by the distribution and location of users
44 Impact of Small Cell Numbers on Small Cell ThroughputIn this section the impacts on the throughput of small cellcluster versus different small cell numbers are investigatedFrom Figure 4 we can see the throughput increases with theincrease of small cell numbers per cluster The reason is thatmore users will be served by the small cells after offloadingprocesses Moreover the throughput increments grow slowly
Mobile Information Systems 9
when small cell numbers per cluster are relatively large Thisis because of the increase of intercell interferences caused bydenser small cell deployments Finally the results in Figure 4prove that the proposed GO scheme outperforms TOFFRalgorithm and incentivized algorithm also in terms of smallcell cluster throughput
5 Conclusion
This paper aims to solve the problem regarding how toperformoffloading in the small cell HetNet deployments withoptimization on maximizing the system energy efficiencyThe reverse auction theory has been implemented with theproposed GO scheme design to decide the offloading targetBS or BSs with coordination transmission enabled technol-ogy The reverse auction model is formulated by multiplesellers (BSs) and a single buyer (offloading user) with thefirst price sealed bid mechanism The BS coordination trans-missions are also supported for multiple winning biddersscenarios According to the proposed reverse auction basedGO scheme the energy efficiency optimization problemwith constraints of user guaranteed throughput thresholdbandwidth occupation and transmission power limitationis solved by Dynamic Programming method with KKTconditions The Individual Rationality and Truthfulness ofthe VCG based reverse auction model are also proved in thepaper System-level simulations have been conducted to ver-ify the effectiveness of the proposed GO scheme according to3GPP LTE-Advanced evaluation methodologies The perfor-mances when applying theGO scheme comparison schemesand the baseline without offloading situation are evaluatedwith performance metrics of energy efficiency offloadinggain and throughput The simulation results prove that theproposed GO scheme can achieve supreme performances
Notation Definition
b = b1 b2 b119894 b119897 BidsB = B
1B2 B
119895 B
119899 Bandwidth allocation results
P = P1P2 P
119895 P
119899 Power allocation results
B119895= 1198611
119895 1198612
119895 119861
119897119895
119895 119895th BS bandwidth allocation
resultsP119895= 1198751
119895 1198752
119895 119875
119897119895
119895 119895th BS power allocation
results119861119894
119895 119894th subcarrier in the 119895th BS
119875119894
119895 Transmission power on 119861119894
119895
b119899= b1 b2 b119899 Bids sent by first 119899 BSs
120578 Energy efficiency
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper is supported by the National High Technol-ogy Research and Development Program of China no2014AA01A701 Nature and Science Foundation of China
under Grants nos 61471068 and 61421061 InternationalCollaboration Project no 2015DFT10160 andNationalMajorProject no 2016ZX03001009-003
References
[1] T Nakamura S Nagata A Benjebbour et al ldquoTrends in smallcell enhancements in LTE advancedrdquo IEEE CommunicationsMagazine vol 51 no 2 pp 98ndash105 2013
[2] W Ni and I B Collings ldquoA new adaptive small-cell architec-turerdquo IEEE Journal on Selected Areas in Communications vol31 no 5 pp 829ndash839 2013
[3] L Hanzo H Haas S Imre D OrsquoBrien M Rupp and LGyongyosi ldquoWireless myths realities and futures from 3G4Gto optical and quantum wirelessrdquo Proceedings of the IEEE vol100 pp 1853ndash1888 2012
[4] TheMETIS 2020 ProjectmdashLaying the Foundation of 5G httpswwwmetis2020com
[5] ldquoSmall cell market statusrdquo White Paper Informa and Small CellForum 1 2013
[6] D Calin H Claussen and H Uzunalioglu ldquoOn femto deploy-ment architectures and macrocell offloading benefits in jointmacro-femto deploymentsrdquo IEEE Communications Magazinevol 48 no 1 pp 26ndash32 2010
[7] S-I Sou ldquoMobile data offloading with policy and chargingcontrol in 3GPP core networkrdquo IEEE Transactions on VehicularTechnology vol 62 no 7 pp 3481ndash3486 2013
[8] J Korhonen T Savolainen A Y Ding and M Kojo ldquoTowardnetwork controlled IP traffic offloadingrdquo IEEE CommunicationsMagazine vol 51 no 3 pp 96ndash102 2013
[9] H Elsawy E Hossain and S Camorlinga ldquoTraffic offloadingtechniques in two-tier femtocell networksrdquo in Proceedings of theIEEE International Conference onCommunications (ICC rsquo13) pp6086ndash6090 IEEE Budapest Hungary June 2013
[10] Z Lu P Sinha and R Srikant ldquoEasyBid enabling cellularoffloading via small playersrdquo in Proceedings of the 33rd IEEEConference on Computer Communications (IEEE INFOCOMrsquo14) pp 691ndash699 IEEE Toronto Canada May 2014
[11] M Usman A Vastberg and T Edler ldquoEnergy efficient highcapacityHETNETby offloading highQoSusers through femtordquoin Proceedings of the 17th IEEE International Conference onNetworks (ICON rsquo11) pp 19ndash24 Singapore December 2011
[12] P Chandhar and S S Das ldquoAnalytical evaluation of offloadinggain in macrocell-femtocell OFDMA networksrdquo in Proceedingsof the IEEE 77th Vehicular Technology Conference (VTC Springrsquo13) pp 1ndash6 June 2013
[13] Q Liu G Feng and S Qin ldquoEnergy-efficient traffic offloadingin Macro-Pico networksrdquo in Proceedings of the 22nd Wirelessand Optical Communications Conference (WOCC rsquo13) pp 236ndash241 IEEE Chongqing China May 2013
[14] Y Jia M Zhao K Wang and W Zhou ldquoAn incentivizedoffloading mechanism via truthful auction in heterogeneousnetworksrdquo in Proceedings of the 6th International Conference onWireless Communications and Signal Processing (WCSP rsquo14) pp1ndash6 Hefei China October 2014
[15] L Gao G Iosifidis J Huang L Tassiulas and D Li ldquoBargain-ing-based mobile data offloadingrdquo IEEE Journal on SelectedAreas in Communications vol 32 no 6 pp 1114ndash1125 2014
[16] F Zhang W Zhang and Q Ling ldquoNon-cooperative game forcapacity offloadrdquo IEEE Transactions on Wireless Communica-tions vol 11 no 4 pp 1565ndash1575 2012
10 Mobile Information Systems
[17] X Xu H Zhang X Dai and X Tao ldquoOptimal Energy EfficientOffloading in small cell HetNet with auctionrdquo in Proceedingsof the 9th International Conference on Communications andNetworking in China (CHINACOM rsquo14) pp 335ndash340MaomingChina August 2014
[18] D P Bertsekas D A Castanon and H Tsaknakis ldquoReverseauction and the solution of inequality constrained assignmentproblemsrdquo SIAM Journal on Optimization vol 3 no 2 pp 268ndash297 1993
[19] M Khaledi and A A Abouzeid ldquoDynamic spectrum sharingauction with time-evolving channel qualitiesrdquo IEEE Transac-tions onWireless Communications vol 14 no 11 pp 5900ndash59122015
[20] F Shen D Li P-H Lin and E Jorswieck ldquoAuction basedspectrum sharing for hybrid access in macro-femtocell net-works under QoS requirementsrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC rsquo15) pp3335ndash3340 IEEE London UK June 2015
[21] X Zhuo W Gao G Cao and S Hua ldquoAn incentive frameworkfor cellular traffic offloadingrdquo IEEE Transactions on MobileComputing vol 13 no 3 pp 541ndash555 2014
[22] W Vickrey ldquoCounterspeculation auctions and competitivesealed tendersrdquo The Journal of Finance vol 16 no 1 pp 8ndash371961
[23] 3GPP-TR36814 (v1110) ldquoEvolved Universal Terrestrial RadioAccess (E-UTRA) Further advancements for E-UTRA physicallayer aspectsrdquo 2013
[24] 3GPP ldquoSmall cell enhancements for E-UTRA and EUTRANphysical layer aspects (release 12)rdquo 3GPP TR 36872 2013
[25] 3GPP R1-130744 ldquoWF on evaluation assumptions for SCEphysical layerrdquo Huawei HiSilicon CATR CMCC 2013
[26] 3GPP ldquoStudy on small cell enhancements for EUTRA and E-UTRAN higher layer aspectsrdquo 3GPP TR 36842 (v1200) 2013
[27] 3GPP TS 36300 ldquoTechnical Specification Group Radio AccessNetwork Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Net-work (EUTRAN) Overall description Stage 2 (Release 12)rdquo2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mobile Information Systems 7
119879119903119906119905ℎ119891119906119897119899119890119904119904 For each bidder the Truthfulness means thateach bidderrsquos bid price is equal to its private value This isa weakly dominant strategy If BSrsquos bidding is untrue theenergy efficiency will be unlikely the biggest In order toget the maximum energy efficiency the allocation should beformulated as follows
119901119895= 120578B119899b119895 minus (120578B119899 minus 120578b119895)
120575 = 119901119895minus 119901119894
= 120578B119899b119895 minus (120578B119899 minus 120578b119895) minus [120578B119899b119894 minus (120578B119899 minus 120578b119894)]
= 120578B119899b119895 minus 120578B119899 + 120578b119895 minus 120578B119899b119894 + 120578B119899 minus 120578b119894
= 120578B119899b119895 + 120578b119895 minus 120578B119899b119894 minus 120578b119894
= (120578B119899b119895 + 120578b119895) minus (120578B119899b119894 + 120578b119894)
(25)
Based on the proposed model in this paper because 119901119895le
119901119894and120575 le 0 thismeans 120578B119899b119895+120578b119895 le 120578B119899b119894+120578b119894 If and only
if 119895 = 119894 it can take the equal signTherefore each bidder mustbe truthful to obtain the maximum system energy efficiencyThe proof is finished
4 Performance Evaluation
In this section we built the system-level simulation plat-form according to the 3GPP LTE-Advanced simulationmethodology [23] Based on this platform we validate theperformances of the proposed reverse auction based GOscheme with comparison algorithms in the small cell HetNetdownlink scenario
41 Simulation Setting Performance Metrics and ComparisonAlgorithms Theconsidered simulation scenario in this papercomprises 19-hexagonal macrocells with 3 sectors per macro-cell In each sector there is one small cell cluster deployedwith shared spectrum manner The small cell cluster is agroup of densely deployed small cells We deploy 23 usersin the coverage of small cell clusters while the remainingusers are distributed in the coverage area of macro cellsThe users are uniformly distributed Moreover as mentionedabove the bandwidth resource granularity in the simulationis one RB In the initial state each user is served by the BSwhich can provide the highest downlink RSRP Once a newtraffic offloading requirement is requested the reverse auc-tion based GO scheme is triggered The detailed simulationparameters are according to 3GPP LTE-Advanced small cellHetNet evaluation methodology [24] These parameters arelisted in Table 1
The performance metrics include the system energyefficiency offloading gain and throughput In this paper themetric of offloading gain (120574gain) is defined as
120574gain =120591offloading
120591total (26)
where 120591offloading denotes the offloaded throughput and 120591totaldenotes the total system throughput The offloading gain is
Table 1 Simulation parameters
Simulation parameter ValueCarrier frequency 2GHzSystem bandwidth 10MHzTotal transmission power ofmacrocell 46 dBm
Total transmission power of small cell 30 dBmPath-loss of macrocell Pl = 283 + 220log
10(119889)
Path-loss of small cell Pl = 305 + 367log10(119889)
Small cell number per cluster 4sim10Small cell cluster number permacrocell 3
User number per macrocell 60
Antenna gain of macrocells 17 dBiAntenna gain of small cells 5 dBiTraffic model FTP Model 1Power spectrum density of thermalnoise minus174 dBmHz
User throughput threshold 5Mbps
a more straightforward notation about howmuch traffic loadcould be offloaded to improve the energy efficiency
In order to evaluate the performances of the proposedGOscheme we compare it with the TOFFR algorithm proposedin [13] and the incentivized scheme proposed in [14] whichhave been introduced in the related works The simulationresults are given as below
42 Impacts of Small Cell Numbers on Energy EfficiencyAccording to 3GPP simulation assumptions there are 23UEs distributed in the coverage of small cell cluster whilethe remainingUEs are uniformly distributed in the remainingarea of macrocells In this section the impacts of deploymentdensity of small cells in a cluster on the system energyefficiency are investigated According to 3GPP simulationmethodology the number of small cells in a cluster variesfrom 4 to 10
As shown in Figure 2 the system energy efficiency ofdifferent algorithms versus small cell numbers per cluster isdemonstrated We can observe that the system energy effi-ciency is improved with the increasing of small cell numbersThe reason lies that small cells usually can provide higherenergy efficiency than macrocells due to lower transmissionpower attenuations in hot spot deployment scenarios Whenthere exist more small cells inside one macrocell more usertraffic could be offloaded potentially to small cellsThereforehigher system energy efficiency could be further achieved
But the system energy efficiency increases slightly whenthe number of small cells is large This is because when thesmall cells are deployed more densely the intercell interfer-enceswill bemore severe indicating the requirement of largertransmission power to ensure the same user throughputMoreover from results in Figure 2 we have proved that theproposedGO scheme outperforms the TOFFR algorithm andincentivized scheme regardless of small cell density deployed
8 Mobile Information Systems
5 6 7 8 9 104Number of small cells
TOFFRwo offloading
GO schemeIncentivized scheme
5
6
7
8
9
10
11
12
13Sy
stem
ener
gy effi
cien
cy (k
bps
W)
Figure 2 System energy efficiency versus small cell numbers percluster
per cluster In the TOFFR algorithm the fractional frequencyreuse scheme is adopted to improve UE performances betterin the cell edge The UE located in the central area cannotcontribute to the energy efficiency improvement The incen-tivized scheme focuses on maximizing the offloading utilitythat purchases the available unused bandwidth in femtocellsSo the proposed GO scheme has better energy efficiencyperformance than both of them
We also compare these three offloading algorithms withthe situation of no offloading (denoted as wo offloading insimulations) as the baseline It is obvious thatwhen offloadingschemes are adopted more user traffic originally served bymacrocell will be actively offloaded to small cells There-fore all of the offloading schemes including the proposedGO scheme TOFFR algorithm and incentivized algorithmachieve higher system energy efficiency
43 Impacts of Small Cell Numbers on Offloading GainAs shown in Figure 3 the offloading gain versus differentsmall cell numbers per cluster is dipicted We can observethat the offloading gain increases with small cell numbersdue to the capacity growth with the increase of small cellnumbers per cluster Besides in order to maximize thesystem energy efficiency there is limitation on the amountof offloaded throughputs as demonstrated in Figure 3Whenthe offloading gain reaches 62 the rise against the small cellnumbers becomes rather slowThe simulation results furthersuggest that the proposed GO scheme outperforms theTOFFR algorithm and incentivized scheme in terms of notonly energy efficiency but also the offloading gain becausein TOFFR algorithm and incentivized scheme the offloadingis mainly focused on the cell edge users which limits theperformance improvements Moreover Figure 3 shows thatthe increasing of the small cell numbers of all these threeschemes will reach a plateau The reason lies that there arealways several specific users out of the coverage of the small
TOFFRGO schemeIncentivized scheme
4 5 6 7 8 9 10 113Number of small cells per cluster
0
10
20
30
40
50
60
70
Offl
oadi
ng g
ain
()
Figure 3 Offloading gain versus small cell numbers per cluster
91011121314151617181920
Smal
l cel
l clu
ster t
hrou
ghpu
t (M
bps)
5 6 7 8 9 104Number of small cells per cluster
TOFFRwo offloading
GO schemeIncentivized scheme
Figure 4 Small cell throughput versus small cell numbers percluster
cell clusters in the simulations Just as mentioned before23 UEs are deployed in the coverage of small cell clusterand the simulation results also show that the offloading gainlimitation of all three algorithms can only reach to near23 which in turn indicates the offloading limitation will bedecided by the distribution and location of users
44 Impact of Small Cell Numbers on Small Cell ThroughputIn this section the impacts on the throughput of small cellcluster versus different small cell numbers are investigatedFrom Figure 4 we can see the throughput increases with theincrease of small cell numbers per cluster The reason is thatmore users will be served by the small cells after offloadingprocesses Moreover the throughput increments grow slowly
Mobile Information Systems 9
when small cell numbers per cluster are relatively large Thisis because of the increase of intercell interferences caused bydenser small cell deployments Finally the results in Figure 4prove that the proposed GO scheme outperforms TOFFRalgorithm and incentivized algorithm also in terms of smallcell cluster throughput
5 Conclusion
This paper aims to solve the problem regarding how toperformoffloading in the small cell HetNet deployments withoptimization on maximizing the system energy efficiencyThe reverse auction theory has been implemented with theproposed GO scheme design to decide the offloading targetBS or BSs with coordination transmission enabled technol-ogy The reverse auction model is formulated by multiplesellers (BSs) and a single buyer (offloading user) with thefirst price sealed bid mechanism The BS coordination trans-missions are also supported for multiple winning biddersscenarios According to the proposed reverse auction basedGO scheme the energy efficiency optimization problemwith constraints of user guaranteed throughput thresholdbandwidth occupation and transmission power limitationis solved by Dynamic Programming method with KKTconditions The Individual Rationality and Truthfulness ofthe VCG based reverse auction model are also proved in thepaper System-level simulations have been conducted to ver-ify the effectiveness of the proposed GO scheme according to3GPP LTE-Advanced evaluation methodologies The perfor-mances when applying theGO scheme comparison schemesand the baseline without offloading situation are evaluatedwith performance metrics of energy efficiency offloadinggain and throughput The simulation results prove that theproposed GO scheme can achieve supreme performances
Notation Definition
b = b1 b2 b119894 b119897 BidsB = B
1B2 B
119895 B
119899 Bandwidth allocation results
P = P1P2 P
119895 P
119899 Power allocation results
B119895= 1198611
119895 1198612
119895 119861
119897119895
119895 119895th BS bandwidth allocation
resultsP119895= 1198751
119895 1198752
119895 119875
119897119895
119895 119895th BS power allocation
results119861119894
119895 119894th subcarrier in the 119895th BS
119875119894
119895 Transmission power on 119861119894
119895
b119899= b1 b2 b119899 Bids sent by first 119899 BSs
120578 Energy efficiency
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper is supported by the National High Technol-ogy Research and Development Program of China no2014AA01A701 Nature and Science Foundation of China
under Grants nos 61471068 and 61421061 InternationalCollaboration Project no 2015DFT10160 andNationalMajorProject no 2016ZX03001009-003
References
[1] T Nakamura S Nagata A Benjebbour et al ldquoTrends in smallcell enhancements in LTE advancedrdquo IEEE CommunicationsMagazine vol 51 no 2 pp 98ndash105 2013
[2] W Ni and I B Collings ldquoA new adaptive small-cell architec-turerdquo IEEE Journal on Selected Areas in Communications vol31 no 5 pp 829ndash839 2013
[3] L Hanzo H Haas S Imre D OrsquoBrien M Rupp and LGyongyosi ldquoWireless myths realities and futures from 3G4Gto optical and quantum wirelessrdquo Proceedings of the IEEE vol100 pp 1853ndash1888 2012
[4] TheMETIS 2020 ProjectmdashLaying the Foundation of 5G httpswwwmetis2020com
[5] ldquoSmall cell market statusrdquo White Paper Informa and Small CellForum 1 2013
[6] D Calin H Claussen and H Uzunalioglu ldquoOn femto deploy-ment architectures and macrocell offloading benefits in jointmacro-femto deploymentsrdquo IEEE Communications Magazinevol 48 no 1 pp 26ndash32 2010
[7] S-I Sou ldquoMobile data offloading with policy and chargingcontrol in 3GPP core networkrdquo IEEE Transactions on VehicularTechnology vol 62 no 7 pp 3481ndash3486 2013
[8] J Korhonen T Savolainen A Y Ding and M Kojo ldquoTowardnetwork controlled IP traffic offloadingrdquo IEEE CommunicationsMagazine vol 51 no 3 pp 96ndash102 2013
[9] H Elsawy E Hossain and S Camorlinga ldquoTraffic offloadingtechniques in two-tier femtocell networksrdquo in Proceedings of theIEEE International Conference onCommunications (ICC rsquo13) pp6086ndash6090 IEEE Budapest Hungary June 2013
[10] Z Lu P Sinha and R Srikant ldquoEasyBid enabling cellularoffloading via small playersrdquo in Proceedings of the 33rd IEEEConference on Computer Communications (IEEE INFOCOMrsquo14) pp 691ndash699 IEEE Toronto Canada May 2014
[11] M Usman A Vastberg and T Edler ldquoEnergy efficient highcapacityHETNETby offloading highQoSusers through femtordquoin Proceedings of the 17th IEEE International Conference onNetworks (ICON rsquo11) pp 19ndash24 Singapore December 2011
[12] P Chandhar and S S Das ldquoAnalytical evaluation of offloadinggain in macrocell-femtocell OFDMA networksrdquo in Proceedingsof the IEEE 77th Vehicular Technology Conference (VTC Springrsquo13) pp 1ndash6 June 2013
[13] Q Liu G Feng and S Qin ldquoEnergy-efficient traffic offloadingin Macro-Pico networksrdquo in Proceedings of the 22nd Wirelessand Optical Communications Conference (WOCC rsquo13) pp 236ndash241 IEEE Chongqing China May 2013
[14] Y Jia M Zhao K Wang and W Zhou ldquoAn incentivizedoffloading mechanism via truthful auction in heterogeneousnetworksrdquo in Proceedings of the 6th International Conference onWireless Communications and Signal Processing (WCSP rsquo14) pp1ndash6 Hefei China October 2014
[15] L Gao G Iosifidis J Huang L Tassiulas and D Li ldquoBargain-ing-based mobile data offloadingrdquo IEEE Journal on SelectedAreas in Communications vol 32 no 6 pp 1114ndash1125 2014
[16] F Zhang W Zhang and Q Ling ldquoNon-cooperative game forcapacity offloadrdquo IEEE Transactions on Wireless Communica-tions vol 11 no 4 pp 1565ndash1575 2012
10 Mobile Information Systems
[17] X Xu H Zhang X Dai and X Tao ldquoOptimal Energy EfficientOffloading in small cell HetNet with auctionrdquo in Proceedingsof the 9th International Conference on Communications andNetworking in China (CHINACOM rsquo14) pp 335ndash340MaomingChina August 2014
[18] D P Bertsekas D A Castanon and H Tsaknakis ldquoReverseauction and the solution of inequality constrained assignmentproblemsrdquo SIAM Journal on Optimization vol 3 no 2 pp 268ndash297 1993
[19] M Khaledi and A A Abouzeid ldquoDynamic spectrum sharingauction with time-evolving channel qualitiesrdquo IEEE Transac-tions onWireless Communications vol 14 no 11 pp 5900ndash59122015
[20] F Shen D Li P-H Lin and E Jorswieck ldquoAuction basedspectrum sharing for hybrid access in macro-femtocell net-works under QoS requirementsrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC rsquo15) pp3335ndash3340 IEEE London UK June 2015
[21] X Zhuo W Gao G Cao and S Hua ldquoAn incentive frameworkfor cellular traffic offloadingrdquo IEEE Transactions on MobileComputing vol 13 no 3 pp 541ndash555 2014
[22] W Vickrey ldquoCounterspeculation auctions and competitivesealed tendersrdquo The Journal of Finance vol 16 no 1 pp 8ndash371961
[23] 3GPP-TR36814 (v1110) ldquoEvolved Universal Terrestrial RadioAccess (E-UTRA) Further advancements for E-UTRA physicallayer aspectsrdquo 2013
[24] 3GPP ldquoSmall cell enhancements for E-UTRA and EUTRANphysical layer aspects (release 12)rdquo 3GPP TR 36872 2013
[25] 3GPP R1-130744 ldquoWF on evaluation assumptions for SCEphysical layerrdquo Huawei HiSilicon CATR CMCC 2013
[26] 3GPP ldquoStudy on small cell enhancements for EUTRA and E-UTRAN higher layer aspectsrdquo 3GPP TR 36842 (v1200) 2013
[27] 3GPP TS 36300 ldquoTechnical Specification Group Radio AccessNetwork Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Net-work (EUTRAN) Overall description Stage 2 (Release 12)rdquo2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
8 Mobile Information Systems
5 6 7 8 9 104Number of small cells
TOFFRwo offloading
GO schemeIncentivized scheme
5
6
7
8
9
10
11
12
13Sy
stem
ener
gy effi
cien
cy (k
bps
W)
Figure 2 System energy efficiency versus small cell numbers percluster
per cluster In the TOFFR algorithm the fractional frequencyreuse scheme is adopted to improve UE performances betterin the cell edge The UE located in the central area cannotcontribute to the energy efficiency improvement The incen-tivized scheme focuses on maximizing the offloading utilitythat purchases the available unused bandwidth in femtocellsSo the proposed GO scheme has better energy efficiencyperformance than both of them
We also compare these three offloading algorithms withthe situation of no offloading (denoted as wo offloading insimulations) as the baseline It is obvious thatwhen offloadingschemes are adopted more user traffic originally served bymacrocell will be actively offloaded to small cells There-fore all of the offloading schemes including the proposedGO scheme TOFFR algorithm and incentivized algorithmachieve higher system energy efficiency
43 Impacts of Small Cell Numbers on Offloading GainAs shown in Figure 3 the offloading gain versus differentsmall cell numbers per cluster is dipicted We can observethat the offloading gain increases with small cell numbersdue to the capacity growth with the increase of small cellnumbers per cluster Besides in order to maximize thesystem energy efficiency there is limitation on the amountof offloaded throughputs as demonstrated in Figure 3Whenthe offloading gain reaches 62 the rise against the small cellnumbers becomes rather slowThe simulation results furthersuggest that the proposed GO scheme outperforms theTOFFR algorithm and incentivized scheme in terms of notonly energy efficiency but also the offloading gain becausein TOFFR algorithm and incentivized scheme the offloadingis mainly focused on the cell edge users which limits theperformance improvements Moreover Figure 3 shows thatthe increasing of the small cell numbers of all these threeschemes will reach a plateau The reason lies that there arealways several specific users out of the coverage of the small
TOFFRGO schemeIncentivized scheme
4 5 6 7 8 9 10 113Number of small cells per cluster
0
10
20
30
40
50
60
70
Offl
oadi
ng g
ain
()
Figure 3 Offloading gain versus small cell numbers per cluster
91011121314151617181920
Smal
l cel
l clu
ster t
hrou
ghpu
t (M
bps)
5 6 7 8 9 104Number of small cells per cluster
TOFFRwo offloading
GO schemeIncentivized scheme
Figure 4 Small cell throughput versus small cell numbers percluster
cell clusters in the simulations Just as mentioned before23 UEs are deployed in the coverage of small cell clusterand the simulation results also show that the offloading gainlimitation of all three algorithms can only reach to near23 which in turn indicates the offloading limitation will bedecided by the distribution and location of users
44 Impact of Small Cell Numbers on Small Cell ThroughputIn this section the impacts on the throughput of small cellcluster versus different small cell numbers are investigatedFrom Figure 4 we can see the throughput increases with theincrease of small cell numbers per cluster The reason is thatmore users will be served by the small cells after offloadingprocesses Moreover the throughput increments grow slowly
Mobile Information Systems 9
when small cell numbers per cluster are relatively large Thisis because of the increase of intercell interferences caused bydenser small cell deployments Finally the results in Figure 4prove that the proposed GO scheme outperforms TOFFRalgorithm and incentivized algorithm also in terms of smallcell cluster throughput
5 Conclusion
This paper aims to solve the problem regarding how toperformoffloading in the small cell HetNet deployments withoptimization on maximizing the system energy efficiencyThe reverse auction theory has been implemented with theproposed GO scheme design to decide the offloading targetBS or BSs with coordination transmission enabled technol-ogy The reverse auction model is formulated by multiplesellers (BSs) and a single buyer (offloading user) with thefirst price sealed bid mechanism The BS coordination trans-missions are also supported for multiple winning biddersscenarios According to the proposed reverse auction basedGO scheme the energy efficiency optimization problemwith constraints of user guaranteed throughput thresholdbandwidth occupation and transmission power limitationis solved by Dynamic Programming method with KKTconditions The Individual Rationality and Truthfulness ofthe VCG based reverse auction model are also proved in thepaper System-level simulations have been conducted to ver-ify the effectiveness of the proposed GO scheme according to3GPP LTE-Advanced evaluation methodologies The perfor-mances when applying theGO scheme comparison schemesand the baseline without offloading situation are evaluatedwith performance metrics of energy efficiency offloadinggain and throughput The simulation results prove that theproposed GO scheme can achieve supreme performances
Notation Definition
b = b1 b2 b119894 b119897 BidsB = B
1B2 B
119895 B
119899 Bandwidth allocation results
P = P1P2 P
119895 P
119899 Power allocation results
B119895= 1198611
119895 1198612
119895 119861
119897119895
119895 119895th BS bandwidth allocation
resultsP119895= 1198751
119895 1198752
119895 119875
119897119895
119895 119895th BS power allocation
results119861119894
119895 119894th subcarrier in the 119895th BS
119875119894
119895 Transmission power on 119861119894
119895
b119899= b1 b2 b119899 Bids sent by first 119899 BSs
120578 Energy efficiency
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper is supported by the National High Technol-ogy Research and Development Program of China no2014AA01A701 Nature and Science Foundation of China
under Grants nos 61471068 and 61421061 InternationalCollaboration Project no 2015DFT10160 andNationalMajorProject no 2016ZX03001009-003
References
[1] T Nakamura S Nagata A Benjebbour et al ldquoTrends in smallcell enhancements in LTE advancedrdquo IEEE CommunicationsMagazine vol 51 no 2 pp 98ndash105 2013
[2] W Ni and I B Collings ldquoA new adaptive small-cell architec-turerdquo IEEE Journal on Selected Areas in Communications vol31 no 5 pp 829ndash839 2013
[3] L Hanzo H Haas S Imre D OrsquoBrien M Rupp and LGyongyosi ldquoWireless myths realities and futures from 3G4Gto optical and quantum wirelessrdquo Proceedings of the IEEE vol100 pp 1853ndash1888 2012
[4] TheMETIS 2020 ProjectmdashLaying the Foundation of 5G httpswwwmetis2020com
[5] ldquoSmall cell market statusrdquo White Paper Informa and Small CellForum 1 2013
[6] D Calin H Claussen and H Uzunalioglu ldquoOn femto deploy-ment architectures and macrocell offloading benefits in jointmacro-femto deploymentsrdquo IEEE Communications Magazinevol 48 no 1 pp 26ndash32 2010
[7] S-I Sou ldquoMobile data offloading with policy and chargingcontrol in 3GPP core networkrdquo IEEE Transactions on VehicularTechnology vol 62 no 7 pp 3481ndash3486 2013
[8] J Korhonen T Savolainen A Y Ding and M Kojo ldquoTowardnetwork controlled IP traffic offloadingrdquo IEEE CommunicationsMagazine vol 51 no 3 pp 96ndash102 2013
[9] H Elsawy E Hossain and S Camorlinga ldquoTraffic offloadingtechniques in two-tier femtocell networksrdquo in Proceedings of theIEEE International Conference onCommunications (ICC rsquo13) pp6086ndash6090 IEEE Budapest Hungary June 2013
[10] Z Lu P Sinha and R Srikant ldquoEasyBid enabling cellularoffloading via small playersrdquo in Proceedings of the 33rd IEEEConference on Computer Communications (IEEE INFOCOMrsquo14) pp 691ndash699 IEEE Toronto Canada May 2014
[11] M Usman A Vastberg and T Edler ldquoEnergy efficient highcapacityHETNETby offloading highQoSusers through femtordquoin Proceedings of the 17th IEEE International Conference onNetworks (ICON rsquo11) pp 19ndash24 Singapore December 2011
[12] P Chandhar and S S Das ldquoAnalytical evaluation of offloadinggain in macrocell-femtocell OFDMA networksrdquo in Proceedingsof the IEEE 77th Vehicular Technology Conference (VTC Springrsquo13) pp 1ndash6 June 2013
[13] Q Liu G Feng and S Qin ldquoEnergy-efficient traffic offloadingin Macro-Pico networksrdquo in Proceedings of the 22nd Wirelessand Optical Communications Conference (WOCC rsquo13) pp 236ndash241 IEEE Chongqing China May 2013
[14] Y Jia M Zhao K Wang and W Zhou ldquoAn incentivizedoffloading mechanism via truthful auction in heterogeneousnetworksrdquo in Proceedings of the 6th International Conference onWireless Communications and Signal Processing (WCSP rsquo14) pp1ndash6 Hefei China October 2014
[15] L Gao G Iosifidis J Huang L Tassiulas and D Li ldquoBargain-ing-based mobile data offloadingrdquo IEEE Journal on SelectedAreas in Communications vol 32 no 6 pp 1114ndash1125 2014
[16] F Zhang W Zhang and Q Ling ldquoNon-cooperative game forcapacity offloadrdquo IEEE Transactions on Wireless Communica-tions vol 11 no 4 pp 1565ndash1575 2012
10 Mobile Information Systems
[17] X Xu H Zhang X Dai and X Tao ldquoOptimal Energy EfficientOffloading in small cell HetNet with auctionrdquo in Proceedingsof the 9th International Conference on Communications andNetworking in China (CHINACOM rsquo14) pp 335ndash340MaomingChina August 2014
[18] D P Bertsekas D A Castanon and H Tsaknakis ldquoReverseauction and the solution of inequality constrained assignmentproblemsrdquo SIAM Journal on Optimization vol 3 no 2 pp 268ndash297 1993
[19] M Khaledi and A A Abouzeid ldquoDynamic spectrum sharingauction with time-evolving channel qualitiesrdquo IEEE Transac-tions onWireless Communications vol 14 no 11 pp 5900ndash59122015
[20] F Shen D Li P-H Lin and E Jorswieck ldquoAuction basedspectrum sharing for hybrid access in macro-femtocell net-works under QoS requirementsrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC rsquo15) pp3335ndash3340 IEEE London UK June 2015
[21] X Zhuo W Gao G Cao and S Hua ldquoAn incentive frameworkfor cellular traffic offloadingrdquo IEEE Transactions on MobileComputing vol 13 no 3 pp 541ndash555 2014
[22] W Vickrey ldquoCounterspeculation auctions and competitivesealed tendersrdquo The Journal of Finance vol 16 no 1 pp 8ndash371961
[23] 3GPP-TR36814 (v1110) ldquoEvolved Universal Terrestrial RadioAccess (E-UTRA) Further advancements for E-UTRA physicallayer aspectsrdquo 2013
[24] 3GPP ldquoSmall cell enhancements for E-UTRA and EUTRANphysical layer aspects (release 12)rdquo 3GPP TR 36872 2013
[25] 3GPP R1-130744 ldquoWF on evaluation assumptions for SCEphysical layerrdquo Huawei HiSilicon CATR CMCC 2013
[26] 3GPP ldquoStudy on small cell enhancements for EUTRA and E-UTRAN higher layer aspectsrdquo 3GPP TR 36842 (v1200) 2013
[27] 3GPP TS 36300 ldquoTechnical Specification Group Radio AccessNetwork Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Net-work (EUTRAN) Overall description Stage 2 (Release 12)rdquo2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mobile Information Systems 9
when small cell numbers per cluster are relatively large Thisis because of the increase of intercell interferences caused bydenser small cell deployments Finally the results in Figure 4prove that the proposed GO scheme outperforms TOFFRalgorithm and incentivized algorithm also in terms of smallcell cluster throughput
5 Conclusion
This paper aims to solve the problem regarding how toperformoffloading in the small cell HetNet deployments withoptimization on maximizing the system energy efficiencyThe reverse auction theory has been implemented with theproposed GO scheme design to decide the offloading targetBS or BSs with coordination transmission enabled technol-ogy The reverse auction model is formulated by multiplesellers (BSs) and a single buyer (offloading user) with thefirst price sealed bid mechanism The BS coordination trans-missions are also supported for multiple winning biddersscenarios According to the proposed reverse auction basedGO scheme the energy efficiency optimization problemwith constraints of user guaranteed throughput thresholdbandwidth occupation and transmission power limitationis solved by Dynamic Programming method with KKTconditions The Individual Rationality and Truthfulness ofthe VCG based reverse auction model are also proved in thepaper System-level simulations have been conducted to ver-ify the effectiveness of the proposed GO scheme according to3GPP LTE-Advanced evaluation methodologies The perfor-mances when applying theGO scheme comparison schemesand the baseline without offloading situation are evaluatedwith performance metrics of energy efficiency offloadinggain and throughput The simulation results prove that theproposed GO scheme can achieve supreme performances
Notation Definition
b = b1 b2 b119894 b119897 BidsB = B
1B2 B
119895 B
119899 Bandwidth allocation results
P = P1P2 P
119895 P
119899 Power allocation results
B119895= 1198611
119895 1198612
119895 119861
119897119895
119895 119895th BS bandwidth allocation
resultsP119895= 1198751
119895 1198752
119895 119875
119897119895
119895 119895th BS power allocation
results119861119894
119895 119894th subcarrier in the 119895th BS
119875119894
119895 Transmission power on 119861119894
119895
b119899= b1 b2 b119899 Bids sent by first 119899 BSs
120578 Energy efficiency
Competing Interests
The authors declare that they have no competing interests
Acknowledgments
This paper is supported by the National High Technol-ogy Research and Development Program of China no2014AA01A701 Nature and Science Foundation of China
under Grants nos 61471068 and 61421061 InternationalCollaboration Project no 2015DFT10160 andNationalMajorProject no 2016ZX03001009-003
References
[1] T Nakamura S Nagata A Benjebbour et al ldquoTrends in smallcell enhancements in LTE advancedrdquo IEEE CommunicationsMagazine vol 51 no 2 pp 98ndash105 2013
[2] W Ni and I B Collings ldquoA new adaptive small-cell architec-turerdquo IEEE Journal on Selected Areas in Communications vol31 no 5 pp 829ndash839 2013
[3] L Hanzo H Haas S Imre D OrsquoBrien M Rupp and LGyongyosi ldquoWireless myths realities and futures from 3G4Gto optical and quantum wirelessrdquo Proceedings of the IEEE vol100 pp 1853ndash1888 2012
[4] TheMETIS 2020 ProjectmdashLaying the Foundation of 5G httpswwwmetis2020com
[5] ldquoSmall cell market statusrdquo White Paper Informa and Small CellForum 1 2013
[6] D Calin H Claussen and H Uzunalioglu ldquoOn femto deploy-ment architectures and macrocell offloading benefits in jointmacro-femto deploymentsrdquo IEEE Communications Magazinevol 48 no 1 pp 26ndash32 2010
[7] S-I Sou ldquoMobile data offloading with policy and chargingcontrol in 3GPP core networkrdquo IEEE Transactions on VehicularTechnology vol 62 no 7 pp 3481ndash3486 2013
[8] J Korhonen T Savolainen A Y Ding and M Kojo ldquoTowardnetwork controlled IP traffic offloadingrdquo IEEE CommunicationsMagazine vol 51 no 3 pp 96ndash102 2013
[9] H Elsawy E Hossain and S Camorlinga ldquoTraffic offloadingtechniques in two-tier femtocell networksrdquo in Proceedings of theIEEE International Conference onCommunications (ICC rsquo13) pp6086ndash6090 IEEE Budapest Hungary June 2013
[10] Z Lu P Sinha and R Srikant ldquoEasyBid enabling cellularoffloading via small playersrdquo in Proceedings of the 33rd IEEEConference on Computer Communications (IEEE INFOCOMrsquo14) pp 691ndash699 IEEE Toronto Canada May 2014
[11] M Usman A Vastberg and T Edler ldquoEnergy efficient highcapacityHETNETby offloading highQoSusers through femtordquoin Proceedings of the 17th IEEE International Conference onNetworks (ICON rsquo11) pp 19ndash24 Singapore December 2011
[12] P Chandhar and S S Das ldquoAnalytical evaluation of offloadinggain in macrocell-femtocell OFDMA networksrdquo in Proceedingsof the IEEE 77th Vehicular Technology Conference (VTC Springrsquo13) pp 1ndash6 June 2013
[13] Q Liu G Feng and S Qin ldquoEnergy-efficient traffic offloadingin Macro-Pico networksrdquo in Proceedings of the 22nd Wirelessand Optical Communications Conference (WOCC rsquo13) pp 236ndash241 IEEE Chongqing China May 2013
[14] Y Jia M Zhao K Wang and W Zhou ldquoAn incentivizedoffloading mechanism via truthful auction in heterogeneousnetworksrdquo in Proceedings of the 6th International Conference onWireless Communications and Signal Processing (WCSP rsquo14) pp1ndash6 Hefei China October 2014
[15] L Gao G Iosifidis J Huang L Tassiulas and D Li ldquoBargain-ing-based mobile data offloadingrdquo IEEE Journal on SelectedAreas in Communications vol 32 no 6 pp 1114ndash1125 2014
[16] F Zhang W Zhang and Q Ling ldquoNon-cooperative game forcapacity offloadrdquo IEEE Transactions on Wireless Communica-tions vol 11 no 4 pp 1565ndash1575 2012
10 Mobile Information Systems
[17] X Xu H Zhang X Dai and X Tao ldquoOptimal Energy EfficientOffloading in small cell HetNet with auctionrdquo in Proceedingsof the 9th International Conference on Communications andNetworking in China (CHINACOM rsquo14) pp 335ndash340MaomingChina August 2014
[18] D P Bertsekas D A Castanon and H Tsaknakis ldquoReverseauction and the solution of inequality constrained assignmentproblemsrdquo SIAM Journal on Optimization vol 3 no 2 pp 268ndash297 1993
[19] M Khaledi and A A Abouzeid ldquoDynamic spectrum sharingauction with time-evolving channel qualitiesrdquo IEEE Transac-tions onWireless Communications vol 14 no 11 pp 5900ndash59122015
[20] F Shen D Li P-H Lin and E Jorswieck ldquoAuction basedspectrum sharing for hybrid access in macro-femtocell net-works under QoS requirementsrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC rsquo15) pp3335ndash3340 IEEE London UK June 2015
[21] X Zhuo W Gao G Cao and S Hua ldquoAn incentive frameworkfor cellular traffic offloadingrdquo IEEE Transactions on MobileComputing vol 13 no 3 pp 541ndash555 2014
[22] W Vickrey ldquoCounterspeculation auctions and competitivesealed tendersrdquo The Journal of Finance vol 16 no 1 pp 8ndash371961
[23] 3GPP-TR36814 (v1110) ldquoEvolved Universal Terrestrial RadioAccess (E-UTRA) Further advancements for E-UTRA physicallayer aspectsrdquo 2013
[24] 3GPP ldquoSmall cell enhancements for E-UTRA and EUTRANphysical layer aspects (release 12)rdquo 3GPP TR 36872 2013
[25] 3GPP R1-130744 ldquoWF on evaluation assumptions for SCEphysical layerrdquo Huawei HiSilicon CATR CMCC 2013
[26] 3GPP ldquoStudy on small cell enhancements for EUTRA and E-UTRAN higher layer aspectsrdquo 3GPP TR 36842 (v1200) 2013
[27] 3GPP TS 36300 ldquoTechnical Specification Group Radio AccessNetwork Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Net-work (EUTRAN) Overall description Stage 2 (Release 12)rdquo2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
10 Mobile Information Systems
[17] X Xu H Zhang X Dai and X Tao ldquoOptimal Energy EfficientOffloading in small cell HetNet with auctionrdquo in Proceedingsof the 9th International Conference on Communications andNetworking in China (CHINACOM rsquo14) pp 335ndash340MaomingChina August 2014
[18] D P Bertsekas D A Castanon and H Tsaknakis ldquoReverseauction and the solution of inequality constrained assignmentproblemsrdquo SIAM Journal on Optimization vol 3 no 2 pp 268ndash297 1993
[19] M Khaledi and A A Abouzeid ldquoDynamic spectrum sharingauction with time-evolving channel qualitiesrdquo IEEE Transac-tions onWireless Communications vol 14 no 11 pp 5900ndash59122015
[20] F Shen D Li P-H Lin and E Jorswieck ldquoAuction basedspectrum sharing for hybrid access in macro-femtocell net-works under QoS requirementsrdquo in Proceedings of the IEEEInternational Conference on Communications (ICC rsquo15) pp3335ndash3340 IEEE London UK June 2015
[21] X Zhuo W Gao G Cao and S Hua ldquoAn incentive frameworkfor cellular traffic offloadingrdquo IEEE Transactions on MobileComputing vol 13 no 3 pp 541ndash555 2014
[22] W Vickrey ldquoCounterspeculation auctions and competitivesealed tendersrdquo The Journal of Finance vol 16 no 1 pp 8ndash371961
[23] 3GPP-TR36814 (v1110) ldquoEvolved Universal Terrestrial RadioAccess (E-UTRA) Further advancements for E-UTRA physicallayer aspectsrdquo 2013
[24] 3GPP ldquoSmall cell enhancements for E-UTRA and EUTRANphysical layer aspects (release 12)rdquo 3GPP TR 36872 2013
[25] 3GPP R1-130744 ldquoWF on evaluation assumptions for SCEphysical layerrdquo Huawei HiSilicon CATR CMCC 2013
[26] 3GPP ldquoStudy on small cell enhancements for EUTRA and E-UTRAN higher layer aspectsrdquo 3GPP TR 36842 (v1200) 2013
[27] 3GPP TS 36300 ldquoTechnical Specification Group Radio AccessNetwork Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Net-work (EUTRAN) Overall description Stage 2 (Release 12)rdquo2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Submit your manuscripts athttpwwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Journal of
Computer Networks and Communications
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ArtificialNeural Systems
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014