Download - Design aspects of network assisted device-to-device communications

IEEE Communications Magazine • May 20112 0163-6804/11/$25.00 © 2011 IEEE

INTRODUCTION

Device-to-device (D2D) communications in cel-lular spectrum supported by a cellular infra-structure holds the promise of three types ofgains. The proximity of user equipments (UE)may allow for extreme high bit rates, low delaysand low power consumption [1]. The reuse gainimplies that radio resources may be simultane-ously used by cellular as well as D2D links, tight-ening the reuse factor even of a reuse-1 system[2]. Finally, the hop gain refers to using a singlelink in the D2D mode rather than using both anuplink and a downlink resource when communi-cating via the access point in the cellular mode.Additionally, D2D communications may extendthe cellular coverage and facilitate new types ofwireless peer-to-peer services [2, 3].

However, D2D communications utilizing cel-lular spectrum poses new challenges, becauserelative to cellular communication scenarios, thesystem needs to cope with new interference situ-ations. For example, in an orthogonal frequencydivision multiplexing (OFDM) system in whichD2D communication links may reuse some ofthe OFDM time-frequency resources (physicalresource blocks, PRB), intracell interference isno longer negligible [4].

In addition, in multicell systems, new types of

intercell interference situations have to be dealtwith due to the undesired proximity of D2D andcellular transmitters and receivers. Interestingly,these new types of interference situations areintertwined with the duplexing scheme that thecellular network and the D2D link employ, andalso depend on the spectrum bands and PRBsallocated to D2D links. For example, when aD2D link utilizes some of the cellular uplinkPRBs, a transmitting cellular user equipment(UE) may cause much stronger interference to areceiving UE of a D2D pair in a neighbor cellthan the interference caused to a radio base sta-tion in that same neighbor cell.

Solution approaches to deal with this prob-lem include power control [5], various interfer-ence avoiding multi-antenna transmissiontechniques [6] that can be combined with propermode selection — which decides whether a D2Dcandidate pair should be communicating in D2Dor in cellular mode [7] — and advanced (net-work) coding schemes [1].

The purpose of the current article is to pro-vide a brief overview of the main technical chal-lenges that need to be addressed to realize thepotential gains of D2D communications and pro-pose solution approaches to these challenges.Throughout we will assume OFDM as the trans-mission scheme and the 3GPP Long Term Evo-lution (LTE) system as our baseline for D2Ddesign [10]. The key functions of D2D communi-cations include peer discovery, physical layerprocedures, such as synchronization and refer-ence signal design, and various radio resourcemanagement functions including mode selection,scheduling, PRB allocation, power control, andintra- and intercell interference management.

We structure the article as follows. The nextsection presents D2D scenarios and discussesthe role of a cellular infrastructure in peer dis-covery and radio resource management. Webriefly overview the key design challenges forD2D communications. Next, we focus on peerdiscovery and examine the role of the network inthis important procedure. We discuss radioresource management issues, including modeselection, scheduling, channel quality estimation,and power control in the mixed D2D and cellu-lar environment. We discuss simulation resultsthat indicate that D2D communications helps toimprove the energy efficiency and to reduce the

ABSTRACT

Device-to-device (D2D) communicationsunderlaying a cellular infrastructure has beenproposed as a means of taking advantage of thephysical proximity of communicating devices,increasing resource utilization, and improvingcellular coverage. Relative to the traditional cel-lular methods, there is a need to design newpeer discovery methods, physical layer proce-dures, and radio resource management algo-rithms that help realize the potential advantagesof D2D communications. In this article we usethe 3GPP Long Term Evolution system as abaseline for D2D design, review some of the keydesign challenges, and propose solutionapproaches that allow cellular devices and D2Dpairs to share spectrum resources and therebyincrease the spectrum and energy efficiency oftraditional cellular networks. Simulation resultsillustrate the viability of the proposed design.

ACCEPTED FROM OPEN CALL

Gábor Fodor, Erik Dahlman, Gunnar Mildh, Stefan Parkvall, Norbert Reider, György Miklós,

and Zoltán Turányi, Ericsson Research

Design Aspects of Network AssistedDevice-to-Device Communications

IEEE Communications Magazine • May 2011 3

probability of infeasibility of a target spectrumefficiency. Finally, we draw conclusions andpoint at future works.

DEVICE-TO-DEVICECOMMUNICATION SCENARIOS AND

DESIGN QUESTIONS

D2D SCENARIOS

As illustrated in Fig. 1, one of the key aspects ofD2D communications is the set of spectrumbands in which D2D communications takesplace. Various ad hoc and personal area net-working technologies utilizing unlicensed spec-trum bands such as the industrial, scientific, andmedical bands are available for short range com-munications, including Bluetooth and WiFiDirect. Although such technologies can operatewithout any infrastructure assistance (lower leftpart of Fig. 1), future ad hoc local area networksoperating in unlicensed bands could also benefitfrom a cellular infrastructure providing nodesynchronization and assisting security procedures(lower right).

D2D communications utilizing licensed (andin particular cellular) spectrum has only recentlybeen proposed and studied [1, 2]. According tothis concept, cellular UEs in the proximity ofeach other can exchange information over adirect link rather than transmitting and receivingsignals through a cellular base station. In thiscase, in a national security and public safety situ-ation or outside of cellular coverage, cellulardevices could communicate without networksupport (upper left), similarly to the TerrestrialTrunked Radio (TETRA) technology, althoughsuch solutions have not yet been standardized bymajor cellular standards organizations. Also, adhoc networking in licensed spectrum bands rely-ing on cognitive radio techniques and acting as asecondary user when the primary spectrum uti-lization is low has been studied for long andsome parts are being standardized [9].

D2D communications in cellular spectrumassisted by a network infrastructure (upperright) has recently been proposed as a means toimprove the utilization of cellular spectrumresources and to reduce the energy consumptionof UEs [2, 5]. The expected added value of net-work control, in addition to providing node syn-chronization and security, is to allow D2D pairsto communicate using cellular resources even athigh spectrum utilization, when cognitive radiotechniques would not indicate unused spectrumholes.

The potential gains of D2D communicationsare equally attractive in cellular networks oper-ated in paired as well as unpaired frequencybands. Since in the 3GPP LTE system, for exam-ple, the Frequency Domain Duplexing (FDD)and Time Domain Duplexing (TDD) modes arespecified in the same set of specifications forboth the UE and the base station (eNB), it is anatural requirement for LTE based D2D com-munications that D2D mode should be support-ed in cellular networks operated in either of theduplexing modes. Independently of the duplex-ing mode of the cellular network, the duplexing

mode of the D2D link can be based on frequen-cy or time division (i.e. FDD or TDD for theD2D link itself). Thus, in terms of duplexing, inprinciple four scenarios are possible. However, itis clear that for the UE transmitter and receiverdesign for D2D traffic, a half duplex TDD designbase is advantageous, because it allows transmis-sion and reception by a given UE without havingto implement separate Tx/Rx hardware in thedevices. Because of the expected gains of net-work assisted D2D communications in licensedspectrum, in the rest of the article we focus onthis (upper right of Fig. 1) scenario.

DESIGN QUESTIONSFor cellular network assisted D2D communica-tions, the role of the network is a major designquestion. For example, in the cellular underlayconcept [5–7] the network can play an activerole for mode selection, power control, schedul-ing, and selecting transmission format (modula-tion and coding rates, multi-antennatransmission mode, etc.). In contrast, in theAura-net concept, the role of the network iskept at a minimum, mainly to provide synchro-nization signals to devices [3].

Peer and service discovery is a major issue inD2D communications, since before two devicescan directly communicate with one another, theymust first know (discover) that they are neareach other. Peer discovery without network sup-port is typically time and energy consuming,employing beacon signals and sophisticated scan-ning and security procedures often involvinghigher layers and/or interactions with the enduser. In network assisted mode, it is a designgoal to make such peer discovery and pairingprocedures faster, more efficient in terms ofenergy consumption and more user friendly.

Mode selection means that the networkand/or the D2D pair decide whether the D2Dpair should communicate directly or via the net-

Figure 1. D2D communications scenarios can be categorized in terms of theutilized spectrum resources and the involvement of various network entitiessuch as a cellular base station or a core network. In this article we focus oncellular network assisted D2D communications utilizing licensed spectrumresources.

D2D useslicensed

spectrum

D2D usesother

spectrum(unlicensed

andunused)

D2D scenario:network assisted

D2D scenario:no network

InternetInternet

work. Design issues around mode selectioninclude:• At what time scale mode selection and asso-

ciated channel quality estimations andreporting should operate. Since the radioconditions within the cell and between theD2D pair may change rapidly, the timescale for mode selection cannot be toocoarse. On the other hand, the measure-ments and control signaling required formode selection should be kept at a mini-mum to avoid too much overhead.

• What measurements, reporting mechanisms,and (periodic and/or event triggered,hybrid) algorithms should be used by thedevices and the eNB to select between theD2D and the cellular links. D2D bearer establishment involves the allo-

cation of resources to the D2D link and the pro-tocols to establish, maintain, and terminate D2Dbearers and to ensure proper quality of service(QoS) for the D2D bearer. Some control logicneeds to decide whether UL or DL PRBs shouldbe used for the D2D link and which resourcesshould be used in each direction for the commu-nication between the D2D pair.

The design of power control and scheduling(including retransmissions and hybrid automaticrepeat request (HARQ) operation) involvesdeciding on:• The functional distribution between the net-

work and the D2D pair.• The time scale of interaction between the

network and the D2D pair. One extreme approach is to let the network

schedule on a very short time scale, whereas analternative design is to let the network decide onthe long term resource usage and allow the D2Dpair to schedule their transmissions autonomous-ly (as will be discussed later).

Intercell interference coordination becomes a

major issue in cellular networks supporting D2Dcommunications, since interference needs to bemanaged between multiple cells and between thecellular and the D2D layers. The intercell inter-ference situations can be different and moresevere due to the D2D layer, as illustrated inFig. 2. When the D2D link uses DL resources, atransmitting device may cause high interferenceto a cellular UE in the neighbor cell receivingDL traffic on the same resource. Similarly, whena D2D pair uses uplink cellular resources, thereceiving device of the D2D pair may sufferstrong interference from a cellular UE in theneighbor cell transmitting uplink traffic to itsserving eNB. These situations are in fact similarto the ones that can occur in unsynchronizedTDD systems and problematic because they can-not be easily handled with single cell approachesthat focus on managing interference betweencellular UEs and D2D pairs within the same cell.From this perspective, a cellular operator mayprefer to use UL resources for the D2D layer inorder to better protect the cellular layer fromintercell D2D interference (lower part of Fig. 2).

SERVICE AND PEERDISCOVERY TECHNIQUES

The first step in the establishment of a networkassisted D2D link is that the network and/or theUEs discover the presence of their peer and thedevices are identified as D2D candidates. Peerdiscovery and device pairing are well known pro-cedures in, for example, Bluetooth, where the socalled inquiry process allows a potential masternode to identify devices in range that wish toparticipate in a piconet, whereas the page processallows the master node to establish links towarddesired slave nodes. From a cellular (LTE) net-work perspective, peer discovery has a similarfunctionality as cell search in LTE by which theUE determines the time and frequency parame-ters that are necessary to demodulate the down-link and to determine the cell identity [10] (andthereby the UE effectively “discovers,” i.e.detects the cell).

In both the ad hoc and the cellular cases, thediscovery is made possible by one party transmit-ting a known synchronization or reference signalsequence (which we will refer to as the beacon).Such a beacon is exemplified by the primary andsecondary synchronization sequences in LTE orknown frequency hopping sequences (FHS) andspecific FHS packets in Bluetooth. Irrespectiveof the technology details, the fundamental prob-lem of device discovery is that the two peerdevices have to meet in space, time, and fre-quency. Without any coordination, this can bemade possible via some randomized procedureand one of the peers assuming the responsibilityof sending the beacon. This role arbitration andtransmitting/searching for the beacon are typical-ly time and energy consuming. In the case ofnetwork assisted D2D, however, the network canmediate in the discovery process by recognizingD2D candidates, coordinating the time and fre-quency allocations for sending/scanning for bea-cons, and thereby making the pairing processmore energy efficient and less time consuming.

IEEE Communications Magazine • May 20114

Figure 2. D2D communications gives rise to intracell interference between thecellular and D2D layer and also new types of intercell interference situations,because D2D pairs and cellular UEs may use overlapping time and frequencyresources. The interference situations are different when the D2D link uses DLor UL PRBs (upper and lower figure respectively).

D2Dusing

DLresource

D2DusingU L

resource

Strong interference

Strong interference

TX UE

TX UE

RX UE

RX UE

UE1

UE1

UE2

UE2

eNB-1

eNB-1

eNB-2

eNB-2


Building on these design experiences fromexisting technologies, identifying D2D candi-dates in a network assisted scheme can be basedon the alternatives depicted by Fig. 3. In a-priorischemes, the network (and/or the devices them-selves) detect D2D candidates prior to com-mencing a communication session between thedevices (left hand side schemes of Fig. 3). As anextreme approach (upper left), the network doesnot actively participate in the discovery processother than assigning beacon resources to thedevices. Such beacon assignments are broadcastin the coverage area of the cell so that D2Dservers (transmitting a beacon) as well as D2Dclients (detecting beacons) can readily find oneanother. According to an alternative approach(lower left) the server first registers to the net-work, and the client willing to engage in D2Dcommunications sends a request to the network(e.g. the serving eNB or other network entity(NWE)). Such registration and request messagesmay contain other information such as an ownidentity, a buddy list, or offered/required ser-vices. In this case the NWE takes a more activerole in the discovery process mediating betweenthe server and the client and requesting theD2D server to generate the beacon (Step 3 inthe lower left scheme).

In a-posteriori device discovery, the network(e.g. an eNB) realizes that two communicatingdevices are in the proximity of one another andthereby they are D2D candidates when the com-munication session is already ongoing (in cellu-lar mode) between the UEs (Step 1 on the righthand side of Fig. 3). In the UE assisted a-poste-riori device discovery (upper right) the UEs

agree on a token that is unique to the alreadyongoing communication session (Step 2 in upperright). Notice that the communication path typi-cally goes through different serving and/or pack-et data gateways (S/PGW) (illustrated in thefigure) and therefore relying on the same physi-cal GW identifying D2D candidates is not aviable option. Once the token is established, theUEs register the token at the serving eNB thatcan easily recognize the two UEs as D2D candi-dates. Alternatively, in the radio access networkbased a-posteriori device discovery (lower right)the eNB analyzes the Internet protocol (IP)packets and in particular the source and destina-tion IP addresses to detect D2D pairs communi-cating within the same cell/sector (Step 2 inlower right).

A-priori and a-posteriori schemes are usefulto identify D2D candidates — essentially identi-fying that the two UEs are in the proximity ofone another and/or in the same cell — but theydo not by themselves reveal the actual radio con-ditions between the D2D candidate nodes.Therefore, the next step in the D2D link estab-lishment procedure is to trigger a beacon signalbetween the D2D server and client indicated asthe last step in all four schemes of Fig. 3, whichthe D2D client can then use to report on theD2D link quality to the eNB. Once this piece ofinformation is available at the eNB, it serves asthe basic input to the mode selection that is dis-cussed further in the next section.

Figure 3. Peer discovery techniques include a-priori (left) and a-posteriori (right) methods. In both cases, theinvolvement of the communicating user equipments and various network entities (such as a base station ora packet/serving data gateway) can be different.

3. Sendtokento eNB

4. Scanrequest

3. Scanrequest

5. Beacon

4. Beacon5. Beacon

4. Beaconrequest

3. Beaconrequest

2. IP address analysis in RAN

3. Beaconrequest

1. D2Dserverreg.

2. D2Drequest

2. Beacon

4. Ack

2. Client-server agree on token

MME

MME

S/PGW1 S/PGW2

MME S/PGW1 S/PGW2

1. Communicationvia RAN and CN

1. eNB broadcastingbeacon assignments

From a cellular (LTE)

network perspective,

peer discovery has a

similar functionality

as cell search in LTE

by which the UE

determines the time

and frequency

parameters that

are necessary to

demodulate the

downlink and to

determine the

cell identity.


D2D RESOURCE ALLOCATION ANDMODE SELECTION ALTERNATIVES

AND D2D CHANNEL QUALITYESTIMATION

MODE SELECTION ANDD2D RESOURCE ALLOCATION ALTERNATIVES

To realize the proximity, reuse, and hop gains, itis intuitively clear (and will be illustrated later)that the instantaneous network load, channelconditions, and intercell interference situationshould be taken into account when selecting thebest mode (cellular or direct) and allocatingresources to the cellular and D2D links in thenetwork. Thus, at one extreme (ALT1 of Fig. 4),the eNB collects up-to-date information onchannels, buffer status, and traffic load andselects the mode and allocates resources (e.g.schedules and allocates power for UL/DL) on atime scale that is similar to LTE’s schedulingand transmission time interval of 1 ms [10].

However, channel measurements, reportingand handling the resources at this time scale forall D2D pairs and cellular UEs in the cell mayincrease the signaling and processing overheadand may not scale well as the number of D2Dpairs in the cell increases. Therefore, an alterna-tive approach in which the eNB allocates dedi-cated resources to each D2D pair on a timescale of hundreds of milliseconds, is motivated(ALT-2/A and ALT-2/B of Fig. 4). In alternativeALT-2/A a resource pool is allocated to the bidi-rectional communication between the devices,whereas in alternative ALT-2/B distinct

resources are allocated to each direction of theD2D link.

A third approach is to manage resources onan even longer time scale. In this case the eNBreserves a pool of resources for a set of (maybeall) D2D pairs in that cell. In this approach(ALT3 of Fig. 4) there is a need for an accessmechanism to the resources within the resourcepool that can lead to, for example, WLAN orBluetooth-like medium access schemes. Thisalternative is not so attractive, because it doesnot fully exploit the advantages of network con-trol.

Various hybrid solutions of these three alter-natives are also possible. For example, the timescale for the reallocation of resources and modeselection in ALT-2/A and 2/B could come closeto the time scale of ALT-1 (e.g. once in every 10ms) or of ALT-3 (e.g. once in every second).Our proposal is therefore that ALT-2/B of Fig. 4be used as a resource allocation design base forD2D communications. ALT-2/B implies thatdedicated resources to the two directions of aD2D pair are reserved by the eNB as part of themode selection procedure. In this design bothperiodic and event triggered mode selectionshould be allowed. In the periodic case, themode selection period should be configurable inthe range of several hundreds of ms. In theevent triggered mode selection case the eventcan be at the eNB and/or at the UE.

D2D CHANNEL QUALITY ESTIMATION ANDPOWER CONTROL

For the D2D link, channel estimation for coher-ent detection is relevant only for the PRBs thatare reserved for the D2D link at mode selectionin alternative ALT-2 above and at D2D resourceallocation occasions. For D2D communication, itresembles the task of the LTE UL demodulationreference signals (DMRS) that are embedded(at a predefined OFDM symbol position withinthe resource block) in the transmitted signal[10]. Therefore, for the D2D channel estimation,an embedded reference signal similar to DMRScan be used.

Recall from earlier that the D2D bearer isassumed to use either cellular UL or DLresources in both directions between the twoUEs. Therefore, we can assume D2D channelreciprocity both in TDD and FDD based cellularnetworks. For the sake of channel quality indica-tion (CQI) and estimation (similarly to CQIreporting in LTE), the eNB can assign (and re-assign) client-server roles to the UEs, includingthe case when both UEs transmit and receiveknown DMRSs for channel quality estimation.In our design, the D2D pair is assigned a D2Dpair specific physical identity (PHY ID). DMRSsare then generated by the Server UE based onD2D PHY ID and possibly other parameters,e.g. a D2D DMRS transmit power (scaling fac-tor) or a D2D pair frequency shift. The D2DDMRS parameters are communicated to bothUEs of the D2D pair. To avoid frequencydomain collisions between D2D DMRSs and/orcellular DMRSs in the neighbor cells, the D2DDMRSs need to be carefully designed. D2DDMRS measurements can then be reported to

Figure 4. The role of the network (and in particular the radio base station) canbe quite different in resource allocation, scheduling and power control over theD2D link. The scope of network control and the time scale over which the net-work controls D2D communications is an important design aspect.

eNB schedulesD2D communications

eNBallocates D2Dresources:- Resource pool- Dedicated resources for upstream/downstream

eNB allocates D2Dresource pool for multipleD2D pairs

ALT-1

ALT-3

ALT-2/AALT-2/B


the eNB to facilitate mode selection, power con-trol, and other RRM functions controlled by theeNB.

As will be illustrated in the numerical section,power control is a key RRM function when D2Dand cellular links use overlapping PRBs. Thus,due to the strong near-far effect, power controlplays a key role in mitigating intracell interfer-ence. Since intracell channel quality informationcan be made available at the eNB, intracellpower control can advantageously be enforcedby CQI dependent SINR target setting [11], aswill be explained in the next section.

PERFORMANCE ASPECTSIn this section we consider the uplink transmis-sion of a seven-cell system with intersite distanceISD = 500 m, in which D2D (candidate) pairsand cellular UEs (i.e. UEs transmitting to aneNB) are dropped according to a surface uni-form distribution in a series of Monte Carloexperiments. To gain an insight into the reuseand proximity gains, we are interested in theperformance of the system in two extreme cases.When all the D2D candidates operate in cellularmode, they communicate via their respectiveserving eNBs. In contrast, when the D2D candi-dates in each cell operate in D2D mode, theyuse a direct D2D link using cellular uplinkresources. The cellular UEs (i.e. the UEs thatare not candidates for D2D communications)always transmit to their respective serving eNBsusing a cellular uplink. For resource sharing, weassume that in cellular mode the resources allo-cated to the UEs are orthogonal either in timeor in frequency, while in D2D mode PRBs arereused by cellular and D2D links.

For ease of presentation, we are primarilyinterested in the system performance as thefunction of the maximum D2D distance (i.e. thedistance between the devices), but also as thefunction of the distance between the cellular UEand the serving eNB. The main performancemeasure of interest is the (uplink) power effi-ciency, that is the required sum transmit powerin the system to realize a capacity (spectrum effi-ciency) target. Depending on the actual outcomeof the Monte Carlo experiment, and dependingon the mode selected for the D2D candidates(D2D or cellular) this capacity target may not befeasible, because the required transmit powerwould grow infinitely large. Indeed, the infeasi-bility of a capacity target is a key performancemeasure when minimizing the sum power con-sumption [8, 11]. Therefore, the second perfor-mance measure of interest is the probability ofthe infeasibility of the capacity target.

SETTING THE SINR TARGETSIn order to reach a predefined sum capacity(spectral efficiency) target, we study two solutionapproaches. In the first approach, each link isassigned a predefined SINR target that corre-sponds to the capacity target. This fixed SINRtarget approach is relatively simple, but has thedrawback that it does not consider the actualgeometry of the system. Therefore, a fixed SINRtarget can be suboptimal in terms of the requiredpower to realize that SINR target or in terms of

feasibility (that is, the predefined SINR targetmay not be feasible for some links). In the sec-ond approach, the SINR targets are determinedsuch that the capacity target is reached, but theindividual SINR targets are set differently takinginto account the geometry of the system. Inshort, this second approach assigns higher SINRtarget to links in favorable channel conditionsand thereby it stimulates higher rate allocationson links with lower path loss values. The detailsof the SINR target setting algorithms aredescribed in [11].

CELLULAR UE POSITIONSTo understand the impact of the cellular UElocations within the cell on the system perfor-mance, we perform Monte Carlo simulations inthe following manner. In experiment series-i, thecellular UEs are dropped according to a surfaceuniform distribution within a ring such that theirdistance from their respective serving eNB fallsin the interval of [i ⋅ r, (i + 1) ⋅ r], where r is thegranularity with which the cell radius R is divid-ed (e.g. r = R/10). Thus, in experiment series 1(UE Position 1), the cellular UE is droppedclose to the cell center, whereas in series 10 (UEPosition 10), the cellular UE is dropped withinthe outer ring area at the cell edge. The UEs ofthe D2D candidate pair are dropped accordingto a surface uniform distribution within theentire cell area such that the maximum distancebetween a D2D pair is bounded (specifically bythe value shown on the “Max D2D distance”axes in Fig. 5 and Fig. 6).

Figure 5. Energy efficiency and infeasibility probability in a multicell systemusing fixed SINR targets. When D2D candidates use D2D mode, the gain ofD2D communications heavily depends on the maximum D2D distance andalso on the position of the cellular UE with which the D2D link shares the cel-lular resources (uplink PRB).

Cellular UEposition

Cellular UEposition

Energyefficiency[bps/Hz/mW]

2

1

10

0.1

46

810

100200

300400

500

Infeasibilityratio

Max D2Ddistance [m]

Max D2Ddistance [m]

2

0.5

1.0

0.0

46

810

100200

300400

500

D2D - fixedSINR targetCellular - fixedSINR target


NUMERICAL RESULTS WITHFIXED SINR TARGET SETTING

The results for the case with fixed SINR targetsare shown in Fig. 5. In D2D mode, the requiredsum power and thereby system-wise energy effi-ciency are sensitive not only to the D2D dis-tance, but also to the position of the particularcellular UE with which the D2D candidatereuses the PRB. We see that the system perfor-mance for up to 100 m maximum D2D distance,(especially when the cellular UE is close to thecell center) is significantly better in D2D mode,both in terms of energy efficiency and infeasibili-ty probability.

NUMERICAL RESULTS WITHADAPTIVE SINR TARGET SETTING

The results for the case with adaptive SINR tar-gets are shown in Fig. 6. As we see in the figure,adaptive SINR targets lead to a significantimprovement both in cellular and D2D modesboth in terms of spectrum efficiency and infeasi-bility probability. The reason for this improve-ment is that that the adaptive SINR targetsetting algorithm sets a higher SINR target forlinks with a low path loss value and thereby thealgorithm encourages spending energy on linkswith a high rate utility. More interestingly, theD2D mode shows superior performance evenwhen the D2D distance is high and for all cellu-lar UE positions. The reason for this improve-ment is that adaptive SINR targets are the keyto fully exploit the proximity gain and at the

same time to control the interference betweenthe D2D and the cellular layers.

CONCLUSIONSNetwork assisted D2D communications in cellu-lar spectrum can take advantage of the proximityof communicating devices, allow for reusingresources between D2D pairs and cellular users,and take advantage of the hop gain. These threefactors can lead to power savings, increasedthroughput, and higher spectrum efficiency. Toharvest these potential gains, there is a need tocarefully design peer discovery mechanisms,physical layer procedures, and radio resourcemanagement algorithms that manage the inter-ference between D2D pairs as well as betweenthe D2D and the cellular layers. Our system sim-ulations indicate that proper mode selection andpower control algorithms can play a key role inrealizing the proximity and reuse gains in themixed D2D and cellular environment. To inte-grate D2D piconets and multihop D2D links incellular networks, to explore the possibilities ofdynamic spectrum allocations for D2D commu-nications, and to devise solutions for interopera-tor D2D schemes are interesting topics for futureworks.

ACKNOWLEDGMENTSThe authors would like to thank Dr. ClaesTidestav and Dr. Mikael Prytz, both at EricssonResearch, for the many discussions and valuablecomments on this work. We are also grateful forthe comments by the anonymous reviewers andthe editor that greatly improved the presentationand the contents of the article.

REFERENCES[1] K. Doppler and M. Xiao, Eds., “Innovative Concepts in

Peer-to-Peer and Network Coding,” WINNER+/CELTICDeliverable CELTIC/CP5-026 D1.3, 2008.

[2] K. Doppler et al., “Device-to-Device Communication asan Underlay to LTE-Advanced Networks,” IEEE Com-mun. Mag., vol. 7, no. 12, 2009, pp. 42–49.

[3] M. S. Corson et al., “Towards Proximity-Aware Internet-working,” IEEE Wireless Commun., Dec. 2010, pp.26–33.

[4] T. Peng et al., “Interference Avoidance Mechanisms inthe Hybrid Cellular and Device-to-Device Systems,” IEEEPers. Indoor and Mobile Radio Commun. Symp., Tokyo,Japan, Sept. 2009.

[5] C.-H. Yu et al., “Power Optimization of Device-to-DeviceCommunication Underlaying Cellular Communication,”IEEE Int’l. Conf. Commun., ICC, Dresden, Germany, June2009.

[6] J. P. Jänis et al., “Interference-Avoiding MIMO Schemesfor Device-to-Device Radio Underlaying Cellular Net-works,” IEEE Pers. Indoor and Mobile Radio Commun.Symp., Tokyo, Japan, Sept. 2009.

[7] K. Doppler et al., “Mode Selection for Device-to-DeviceCommunication Underlaying an LTE-Advanced Net-work,” IEEE Wireless Commun. and Networking Conf.,Sydney, Australia, Apr. 2010.

[8] R. Chen et al., “Uplink Power Control in Multi-Cell Spa-tial Multiplexing Wireless Systems,” IEEE Trans. WirelessCommun., vol. 6, no. 7, July 2007, pp. 2700–11.

[9] J. Perez-Romero et al., “A Novel On-Demand CognitivePilot Channel Enabling Dynamic Spectrum Allocation,”Proc. 2nd Int’l. Symp. New Frontiers in Dynamic Spec-trum Access Networks, Apr. 2007.

[10] E. Dahlman, S. Parkvall, and J. Sköld, 4G: LTE/LTE-Advanced for Mobile Broadband, Academic Press, ISBN:012385489X, 2011.

[11] G. Fodor and N. Reider, “A Distributed Power ControlScheme for Network Assisted D2D Communications,”submitted for publication.

Figure 6. Energy efficiency and infeasibility probability in a multicell systemusing adaptive SINR targets. When the SINR targets are properly set, D2Dcommunications has the potential to drastically improve the energy efficiencyand reduce the probability of infeasibility over a wide range of D2D distancesand cellular UE positions.

Cellular UEposition

Energyefficiency[bps/Hz/mW]

Max D2Ddistance [m]

2

1

10

0.1

46

810

100200

300400

500

Cellular UEposition

Infeasibilityratio

Max D2Ddistance [m]

2

0.5

1.0

0.0

46

810

100200

300400

500

D2D - adaptiveSINR targetCellular - adaptiveSINR target


http://gfodor.iblogger.org/upload/conference/D2D-Dis-tributed-Power-Control.pdf

BIOGRAPHIESGABOR FODOR [SM] ([email protected]) received aPh.D. degree in teletraffic theory from the Budapest Uni-versity of Technology and Economics in 1998. Since thenhe has been with Ericsson Research, Kista, Sweden. He iscurrently a master researcher specializing in modeling, per-formance analysis of and protocol development for wire-less access networks. He has published around 50 papersin reviewed conference proceedings and journals and holdsabout 20 patents (granted or pending). He has been oneof the chairs and organizers of the IEEE Broadband Wire-less Access Workshop series since 2007.

ERIK DAHLMAN ([email protected]) is currently thesenior expert in radio access technologies within EricssonResearch. Most recently, he has been involved in the stan-dardization and development of the 3GPP long term evolu-tion (LTE) and its evolution towards LTE-Advanced. He wasalso deeply involved in the development and standardiza-tion of 3G radio access technologies, first in Japan, andlater within the global 3GPP standardization body. He isthe coauthor of the book “3G Evolution — HSPA and LTEfor Mobile Broadband” and its recent follow-up “4G —LTE/LTE-Advanced for Mobile Broadband.”

STEFAN PARKVALL [SM] ([email protected]) (seniormember, IEEE) joined Ericsson Research in 1999 and is cur-rently a principal researcher in the area of radio access,working with research and standardization of cellular tech-nologies. He has been heavily involved in the developmentof HSPA and LTE and is also co-author of “3G Evolution —HSPA and LTE for Mobile Broadband” and “4G — LTE/LTE-Advanced for Mobile Broadband.” In 2009, he received“Stora Teknikpriset” (one of Sweden’s major technicalawards) for his work on HSPA. He holds a Ph.D. from theRoyal Institute of Technology (KTH), Stockholm Sweden in1996. His previous positions include assistant professor incommunication theory at the Royal Institute of Technology,Stockholm, Sweden, and a visiting researcher at Universityof California, San Diego, USA.

GUNNAR MILDH ([email protected]) received hisM.Sc. in electrical engineering from the Royal Institute ofTechnology (KTH), Sweden in 2000. In 2000 he joined Eric-sson Research and has since then been working with stan-dardization and concept development for GSM/EDGE, HSPAand LTE. His focus areas are radio network architecture andprotocols. He is currently an expert in radio network archi-tecture at the Wireless Access Network department, Erics-son Research.

NORBERT REIDER ([email protected]) received theM.Sc. degree in faculty of computer science from theBudapest University of Technology and Economics,Budapest, Hungary, in 2007. He is currently workingtowards the Ph.D. degree at Budapest University of Tech-nology and Economics. In 2010, he joined EricssonResearch, where he is currently a research fellow. He is astudent member of the IEEE. His research interests includeoverall topics related to radio resource management inwireless networks.

GYÖRGY MIKLOS ([email protected]) received hisM.S. and Ph.D. degrees from the Budapest University ofTechnology and Economics in 1997 and 2004, respectively.He currently works as a research fellow at Ericsson Researchin Budapest, Hungary. His research focus includes mobilepacket core architecture and standardization, machine-to-machine and device-to-device communications.

ZOLTAN RICHARD TURANYI ([email protected])received his M.Sc. degree in computer science in 1996from the Budapest University of Technology and Eco-nomics. During the same year, he started his Ph.D. studiesfocusing on packet network performance, QoS, networksimulation and routing.