Keywords: Load Balancing. QoS, Routing Algorithm, Ant Colony Optimization (A CO),Optimized Route, throughput
NomenclatureS Source node in case studyD Destination node in case studyL Number of neighbor's node that located for each
nodePp Choosing probability of current route by ant that
move from j node to i destinationN The number of nodesn Indicate the number of current nodeAP Efficiency of Ants movement on probability
tablePQ Optimize routePf^Q Non-optimize route
h Introduction
With the rapid growth and evolution of wirelesscomputer networks, not only their topologies ischanging and progressing in continuous form, but alsomanner and type of utilization is permanently changingfor various goals. Present routing algorithms, used inwireless computer networks is not suitable to coordinatewith the instant growth. Hence, new algorithms forrouting and load balance are needed in the network [1]-[3]. Wireless ad hoc networks (MANETs) are networksin which all nodes are mobile and communicate witheach other via wireless connections [I].
Nodes can join or leave a specific network at anytime. There is no fixed infrastructure, all nodes areequal and there is no centralized control. There are nodesignated routers; any node can serve as router forothers, and data packets are forwarded from node tonode in a multi-hop fashion [I]. In other hand Wirelessmesh networks (WMNs) consist of static wirelessrouters and gateways, which are directly connected tothe wired infrastriictLire. User stations are connected tothe wired infrastructure via wireless link.
The main goal of routing algorithm is to maintain theQuality of Service (QoS) and to prevent the trafficoverload in some parts ofthe networks [2]-[4]. Dorigoand Di-Caro proposed Ant-Net algorithm first byconsidering only delay in their routing tables. Theirmethod is not suitable for the present requirement ofamount heterogeneous networks consisting of variousnetworks with different goals [4]-[6].
The Asymmetrical Route Synchronization (ARS)algorithm is based on agent, and is proposed fordatagram networks [2], [6]. The ARS supports QoSrouting, resource reservation, and admission controlfunctions using ant-like agents. It works under twoconditions: (/)every node in a network supports theweighted fair queuing algorithm and (/'/) network usersrequire two service class deliveries: datagram and real-time flow. The ARS efficiently allocates networkresources according to user requests. In ASR. for every"£)" and "H", the last two estimation of the trip time
A. A. Moghanjoughi. S. Khatun, M. A. Borhanuddin, R. S. A. Raja Abdullah
from any node k via n to d are stored. Details of thealgorithms and their performance comparison can befound in [7].
The only point that must be considered in thisalgorithm is the agent of controlling the acceptance ofagents similar to ants in addition to control resources.The genetic algorithm defines the main and importantrole to implement this method based on optimumrouting.
In this paper, a new algorithm is proposed byconsidering the QoS, the issue of load balancing androuting for Wireless Mesh Networks. The algorithm isbased on a specific self-organizing behavior of antcolonies, the shortest path discovery, and on the relatedframework of ant colony optimization (ACO) [8]. It hasbeen observed that, ants in a colony can focus onmoving over the shortest among different pathsconnecting their nest to soiu"ce of food [8], [9]. Themain catalyst of this colony-level shortest path behavioris the use of a volatile chemical substance calledpheromone. Ants moving between the nest and a foodsource deposit pheromone, and preferentially movetowards areas of higher pheromone intensity. Shorterpaths can be completed quicker and more frequently bythe ants, and wili therefore be marked with higherpheromone intensity [10].
This paper is organised as follows. Section II,describes related work and routing based on antscolonies behavior. Section 111, describes the proposedalgorithm. Section IV, presents simulation results,followed by the conclusion in section V.
II. Routing Based on Ant's ColoniesBehavior
In this section the detail related literature on ant'scolonies behaviour in terms of ACO routing algorithms,and ill QoS based routing are presented.
//. I. ACO Routing Algorithms
The basic ACO algorithms based idea for routing [8],[IIJ is the possession of routing information throughthe sampling of paths using small control packets,which are called ants. The ants are generatedconcurrently and independently at the nodes, with thetask, to test a path, from "5"' to "D". The ant collectsinformation about the quality of its path (e.g. number ofhops, etc.), and uses this on its way back from "D" to"S" to update the routing information at theintermediate nodes and at "S". Ants always samplecomplete paths, so that routing information can beupdated in a pure Monte Carlo way, without relying onbootstrapping information from one node to the next[12]. The routing tables for each destination contain avector of real-valued entries, one entry for each knownneighbor node from source to destination route. Theseentries are a measure of goodness of going over that
neighbor on the way to a certain destination. They aretermed pheromone variables, and are continuallyupdated according to path quality values calculated bythe ants. The repeated and concurrent generation ofpath-sampling ants results, the availability of a bundleof paths, each with an estimated measure of quality, ateach node. In tum, the ants use the routing tables todefine the path they may sample. At each node theystochastically choose a next hop, giving higherprobability to links with higher pheromone values. Inthe rest of the paper the routing tables are denoted aspheromone tables and vice-versa.
This process is quite similar to the pheromone layingand following behavior of real ant colonies. Like theirnatural counterparts, the artificial ants are in practiceautonomous agents, and through the updating andstochastic following of pheromone tables theyparticipate in a stigmergic conununication process. Theresult is a collective learning behavior, in whichindividual ants have low complexity and littleimportance, while the whole swarm together can collectand maintain up-to date routing information.
The pheromone information is used for routing datapackets, more or less in the same way as for routingants: packets are routed stochastically, giving higherprobability to links with higher pheromone values.Usually, data for the same destination are spread overmultiple paths (but with more packets travel over thebest paths), resulting balanced load in the network. Thismechanism is usually adopted to avoid low qualitypaths, while to make the ants more explorative, so thatthe good paths can be maintained and less good pathscan be avoided. The path exploration is kept separatefrom the ongoing paths. Enough ants are sent to thedifferent destinations to make sure nodes have up-to-date infonnation about the best paths and canautomatically adapt their data load spreading.
II.2. QoS-Based Routing with Ant Colonies Behavior
Many optimizing algorithms for solving the problemsin distributed systems come from Ant's coloniesbehavior. Ants exude a special characteristic"pheromone" when they leave their homes towards foodsource. They can find the shortest route in this way. Incomputer networks, probability in routing tables, whichexisted in all nodes in networks for routing, can act aspheromone. This probability is calculated with ant'straffic or packages [4], [6].
Dorigo and Di Caro show a method, which is calledAnt-Net, based on ant's colonies. In this method theinformation that comes from Ant's appear in each nodelike a routing table and a data structure (LIS') thatinvolve information about local traffic and delayquantities. Fig. I shows data structures of an N-node
A. A. Moghanjoughi, S. Khatun, M. A. Borhanuddin, R. S. A. Raja Abdullah
network used in Ant-Net method where each node has"L" neighbor [5], [6].
The nodes in the network are functioning as routersalso. To accomplish this task, they are using ants togather information about the traffic load in the network.The traffic load depends on the amount of packets in thenetwork (congestion level of the network). Thecongestion level is proportional to the delay that a datapacket experiences during the trip from its source to itsdestination. If the congestion level is high the datapackets will need more time to get to their destination.To collect the infomiation about the traffic ioad on theirpath towards destination, the ants calculate the time thata data packet would experience using the same path.This information is used to update the two datastnictures in a node. These two data structures are: therouting table and the local traffic statistics. The ants areable to read and write in these two structures, while thedata packets are only reading information from therouting table to get to their destination.
Routing table is a local data-base that heips router todecide where to forward data packets. It contains theinformation which specifies the next (neighbour) nodethat should be taken by a data packet to get to anypossible destination in the network. Each routing tableis organized as a set of all the possible destinations (allthe nodes in the network) and the probabilities to reachthese destinations through each of the neighbours of thenode (next hops).
Netwcut Nodes
Pn
Pz,
PLI
P..P22
PL:
P I N
PaN
Pu-iy
DrUy
Fig. I. Data structure used in AnI Net Method [6]
The routing tables are organized like vector distancealgorithms. LTS data structure show the trafficinformation and delay of every destination. In Ant-Net,two different kinds of ants were used: Fnfvi'ard Ant(FA.) and Backward Ant (BA.). These two Ants aredepicted in Fig. 2 [4]-[6].
FAs are the researcher Ants that discover the newroutes, evaluate the route and analyze delay. FA movefrom source to destination during the route and store theinfomiation of route passing time in one stack. Whenthey arrive to their destination, produces BA andtransfers to it all the information that stored during thisroute. Backward ants use the stacks, produced by FA,
and come back to the source in reverse route thenupgrade the LTS structure and routing table for allnodes.
Forward Ant (FA)
Source > » Destination
oBacitward Ant (BA)
Fig. 2. Two types of Ants used in Ant Net
Quantity of P,.; in routing table (see Fig. 1) calculatedas follows: if only one ant arrive (from "/" destination)to node ' / ' , the probability of choosing that route willbe increased base on Eq. (1). In addition, probability ofconnective route to other nodes will be decreased baseon Eq. (2), because the probability table must benormal. Finally, Eq. (3) is correct for all probabilitytable columns [4], [7]:
, Pj.mLD+^
•" 1 + AP
f € Neighbor [n)
j " *• J
„ ={ Neighbor (n)]
0)
(2)
(2.1)
(2.2)
(3)
(3.1)
(3.2)
Age(4)
AP shows the effect of ant movement on probabilitytable entrance and it is created from delay reverse or antage. plus one fix number according to Eq. (4). If ant ageand delay were low, the probability of that route is highand then inclination of other Ant's for moving that routewill be increased.
A. A. Moghanjoughi, S. Khatun, M. A. Borhanuddin. R. S. A. Raja Abdullah
One of the most important problems that exist inrouting algorithms with ant's colonies is statistic routeand not adaptive as per need in the network [15].Statistic means that after Ants choose the /*„ route asoptimum route, the inclination of other ant's tochoosing this route is usually increased. This causestraffic congestion due to increased traffic and at thesame time the probability of use of other routes will bedecreased. Choosing P,, route as the only optimizedroute in network caused undesired result and problemslike no optimization ofP,,.
To solve this statistic nature route problem. MultipleAnt Colony Optimization (MACO) [13] use differentants with their special "pheromone" instead of usingone kind of ant. In MACO, each kind of ant can find theseparate optimum route and this increase routeadaptation and solve the statistic route problem [13].
In addition, Bonabeaul has showed that success ofants when they act gregarious to fmd the shortest routeis proportional. If a group of ants choose non-optimumroute P,,.. by chance, the other ants will choose andfollow the same P,,,, at the end. The concentration of'pheromone' will be increased and the not optimizedroute P,,,, will be used as optimize route. Therefore, non-efficient network can be created [I3]-[15].
In this paper the proposed method attempts decreasethe probability of choosing not optimized route insteadof optimized one by more searches among routes. Also,instead of using one /*„„ route as the optimizing route,use of several optimized routes can solve the congestionproblems caused by a single optimizing route.
III. The Proposed Algorithm
In this section, a simple and effective new algorithmis presented tor load balancing in WMNs. In thisproposed method, the used ants are able to regenerate.In the routing process when an ant from a particularfamily arrives to a junction node with 'm' differentconnective routs, the ant regenerates 'm' successive antsand then send them, one per each connective routeaccordingly based on same considering criterion. Thesenew generations of ants inherit identifier of their family,the number of their generation and the information thattheir parents get during the routing procedure. Thisinformation involves the stack that includes route delay,bandwidth and identification of visited nodes. Thismethod uses different ant's for routing instead of onekind of ant like MACO. In addition, each ant is used forone kind of services for different QoS needs. Toachieve a complete routing table, several kinds of antswith different quality of services is used for each node.
With simple changes, the Ant-Net method can beconverted to methods that prepare QoS, requiredbandwidth and Hop Count (HC) [14] by forcing the FAto use connective routes. Whenever FA passes thestages more than a certain quantity Hn,a», it die and the
routing algorithm of Ant-Net is updated based onnumber of stage [ 15].
In the proposed method, minimum bandwidth (BW)and maximum step (H„ax) ''̂ e considered as QoSparameters. As usual, here also, if FA passes the stagesmore than //^a^ it die and an ant produces 'm' new antswhen it arrives to a junction with 'm' connective routesin it to cover all possible routes. While ants movetowards destination, like Ant-Net method, store theidentification of visited nodes and quantity of QoSroutes in their stack. When an ant arrived to a newnode, it first check whether any co-family ant pass thisnode or not. Any co-family ant already passed the nodemeans that an ant passes this node in shorter route thenthe present ant and the present route is not optimum anddie there. Otherwise the ant like its parent, regeneratesand send new Ant's towards the route that can prepareQoS conditions. This process continues until ants of thatgeneration arrive to destination. When the primary antarrives to destination it finds out, tlrst: whether theroute that it has passed becomes the shortest route ornot, second: the QoS condition. Therefore, the primaryant produced BA ant and transfer the infonnation thathas been gathered during route optimization period andthen die there. Mere BA has duty to use existinginformation stack transferred by FA, follow backwardthe same route created by FA and update the existingrouting tables.
For example, in Fig. 3, an ant in the "a" family at thetime of "t" is going to search optimum route from "S"node to "/)" node. First, this ant generates 3 more antsas there are total 4 routs from "S" to next neighboringnode. These three new ants are considered as 1stgeneration (al) and they will follow the routs SB, SCand SF. But no ant will follow SA route due to longestdistance {i.e.. Lack of QoS). In each stage, when an antmoves forward from one node to next, the generationnumber is increased accordingly, for example a2, a3etc. They (al, a2, a3 etc.) represents ant generation andat the same time number of passing routs from a sourcenodes (shown in Figs, 4-5-6)
A. A. Moghanjoughi. S. Khatun, M. A. Borhanuddin, R. S. A. Raja Abdullah
Fig. 4. Second generation of Ant
A
33
Fig. 5. Third generation of Ant
Fig. 6. Sample of suggestion method
As shown in Fig. 6, when ant a3 arrives from node Fto G and see that a co-family ant (a2) with smallergeneration already passed it, then the route, SCFG willnot be optimized and ant a3 dies there. Same thinghappens for route SEGF.
At destination D three co-family ant's a2, ai havereached there from three different routes BD, FD andGD respectively. It is consider that a2 ant passes theshortest route compared the other Ants and finally therouting table will be updated with SBD as optimumroute for SD connection.
In this method, optimized routes can be found veryquickly and the system will be balanced accordingly.Because of fast searching, the probability of choosingnon-optimizing route instead of optimizing routes isdecreased as expected.
///. I. Simulation Model
The network topology used for simulation isNSFNET model similar to Ant-Net method [4]. Thismodel has 14 nodes and 21 cotinective routes (see Fig.7).
Fig. 7. NSF Network Model
The visual C# language is used to develop and runthe simulation model. The bandwidth of each of allconnective routes is 1.5 Mbits/s.
III.2. Buffer at a Node Model
Every node in the network, functions as a host and arouter simultaneously and it is able also to move fromone place to other. In this section the considered buffersat the nodes model in Mobile AntNet are discussed.
Each connected node has three kinds of buffers:1. Input buffer;2. Output buffer with high- and low-priority queues
for every neighbour;3. WMNs buffer with one queue for each destination
in the network,
Input Suffor t
Incoming
.,Output Sutttr, two queueten each naighbo) nods
Outgoing
WMNs Burief. one queuelor each desllndOon node
Network Node
Fig. 8. A representation ofthe node model when the node hasneighbours
A. A. Moghanjoughi, S. Khatun, M. A. Borhanuddin, R. S. A. Raja Abdullah
The original AntNet uses nodes with two buffers: theinput and the output buffer. In the Mobile AntNet weimplemented an extra buffer - WMNs buffer (see Fig.8).
III. 3. Assumption of Simulation
• The arrival rate and the distribution of packets ineach session follow Poisson distribution.
• Size of BA = NI! + 24 (NH: the number of visitednodes by FA's Ant) bytes.
• The data packet size is exponentially distributed.• The ants inter arrival time is assumed as I second.• Size ofFA = 24 byte.• Maximum numbers of steps or Hma.x = 6. To avoid
increasing load in network, the number of generatedants and the process of generation have consideredshort. The number of created ant's in each node istwo (m=2) and the process generating continuesuntil the final generation.
IV. Simulation Results
The performance of the proposed method isevaluated via simulation. To estimate simulation results,two parameters are studied: packets delay and thenumber of arriving packets to destination (throughput),Figs. 9-10, shows that the delay is decreased and thethroughput is increased in this method compared toAnt-Net method.
Fig. 9 shows the delay compression of proposedmethod, Ant-Net algorithm and conventional Method.The delay mainly depends on the network topology. Itis apparent from the Figure that the delay in theproposed method is quite low cotiipared to Ant-Net. Inaverage, the proposed method is able to reduce 45.47 %and 65.34% delay compared to Ant-Net andconventional method respectively, showing itsefficiency. On the other hand in a similar fashion. Fig.10 shows the throughput efficiency of the proposedmethod. Here throughput is increased 38.58 %compared to Ant-Net.
Fig. 10. Throughput comparison: Ant Net vs. Proposed method
V. Conclusion
In this paper, focus was given on determining thebest routes and bandwidth allocation for various trafflcflows in WMNs to maximize the fair share allocated tothe users. A new algorithm for load balancing withQOS has been introduced based on ant coloniesoptimization algorithms. Two cun-ent problems in Antcolony Algorithms; (/) Statistic network and (ii)choosing a non-optimized route have been eliminatedby considering anl"s and their successive generations,which shows the better load distribution among networknodes. The results show that tlie proposed algorithm,itideed, find near optimal solutions.
References[1]
[2]
Kig. 9. Average packet delay compression between three algorithms;Proposed algorithm. Ant net and conventional
Dorigo M. Stutzle T. Anl Colony Optimization. MIT Press:Cambridge, MA,. (Year of pubtication: 2004 ISBN-I3:978-O-262-04219-2).L. Varzandeh et al.. Load Balancing in QoS Based ComputerNetworking with Ant Net Algorithm, huanattomil Confenrnceon Etecirital Engineering tCEE20il6 Inlcrnatioiial Review onComputer.^ and Software, pp. 94-98, 2006.L. Yong. Z. Guangzhou. S. Fanjun, Adaptive ant based dynamicrouting algorithm, proceeding of the 5lh IEEE Conference onIntelligent Control and Automation, pp. 2694-2697. 20()4.Z. Subing, L. Zemin. A QOS routeing algorithm based on Anialgorithm, IEEE Inlenialioniit Conference tin Cimimiiniciilions,Vol. 5, No. I,p.p.l587-1591. 2001.'
Dorigo M., Di Caro G., Ant Net: A Mobile Agents Approach toAdaptive Routing Technical. Report. IR/DIA- Free BrusselsUniversity, Belgium, 1997.K. M. Sim. W. H. Sun, Ant Colony Optimization for Routingand Load-Balancing: Survey and New Directions, IEEETransiictiou.s on Systems, Man. ond Cybernetics -Part A:Systems ami Humans. Vol. 33. No. 5, pp. 560-572, 2003.A. Pacnt. M. Gadomska. A. Igiclski, Ant-Routing vs Q-Routingin Telecommunication Networks, Proceedings of the 20-thECMS Conference. 2006Dorigo M, Di Caro G, Gambardella LM. Ant algorithms fordiscrete optimizalion. Artificial Life 1999; Vol. 5. No.2, pp. 137-172. 1999.Goss S, Aron S, Deneubourg JL. Pasteeis JM. Self-organizedshortcuts in the Argentine ant. Naturwissenschat^en, Vol. 76,pp. 579-581,1989.
[10] G. Di Caro, F. Ducatelle. L. M. Gambardella, AntHocNct: anadaptive n am re-in spired algorithm for routing in mobile ad hocnetworks, EUROPEAN TRANSACTIONS ON
A. A. Moghanjoughi, S. Khatun, M. A. Borhanuddin, R. S. A. Raja Abdullah
TELECOMhfUNJCATIONS, Euro. Trails. Telecomms.AErr2005, Vol. 16. pp. 443^55, 2005
[11] K. M., W. H. Sun, Multiple ant-colony optimizations fornetwork routing, proceeding ofthe 1st Int. Symp. Cyber world,Tokyo. Japan, pp. 277-281, 2002.
[12] R. Schoonderwoerd, O. Holland, J. Bruten, L. Rothkrantz, Antsfor Load Balancing in Telecommunication Networks, HewlettPackard Lab., Bristol, U.K., Technical Reppon HPL-96-35,1996.
[13] B. Baran, R. Sosa, A new approach for Ant Net routing, 9thInternational Conferences on Computer CommunicationsNetworks, Las Vegas, NV, 2000.
[14] D. Caro, M. Dorigo, Ant Net: Distributed stigmergetJc controlfor communications networks. Journal Arlijicial InielligenceResource, vol. 9, pp. 317-365, 1998.
[15] K. Oida, M. Sekido. An agent-based routing system for QOSquarantines. Proceedings of Ihe IEEE International Conferenceon Systems, Vol. 3, 12-15, pp. 833-838, 1999.
Authors' information'Department of Computer and Communications Systems Engineering,Faculty of Engineering,Universiti Putra Malaysia,43400, Serdang. Selangor,Malaysia.
-MIMOS Berhad.Technology Park Malaysia,57000 Kuala Lumpur,Malaysia.
Ayyoub Akburi Mwghanjoughi received hisB. Eng from IAU University (2006) in Iran andcurrently he doing his M.Sc of IT andcommunication engineering in Faculty ofEngineering at University Putra Malaysia. Alsohe is postgraduate research assistant there. Hewas professional inspector of ICT trainingCenters for state of Yazd in Iran (2004-2006).
He is a member of International Association of Engineers (lAEng) andInternalional Kngineering Consortium (IEC). His research interests aremainly nianagenienl aspects of high-speed and wireless Networks,including ihe area of topology discovering, quality of service (QoS)routing, Wireless Ad-Hoc and Wireltss Mesh Networks. Also AntColony's Optimization (ACO) is his specially interest research area.E-mail: ayyoub.akbariitijamail.com
Associate Professor Dr. Sabira Khatun ispresently working at the Departmetit ofComputer & Communication SystemsEngineering, Faculty of Engineering, andUniversity Putra Malaysia. She is head ofElectronics and Cotnmunications Laboratoryand Research Advisor of Networks ResearchGroup in the departnient. She received ber
PhD in communications and Networking from University PutraMalaysia (UPM) in 2003.She directs research activities within the Wireless Communicationsand Networks (WCN) group and her work in Mobile IPv6 (MIPv6) isgaining importance in Internet-based applications so as to offer secureuninterrupted on-line experience in interactive network applicationssuch as in entciiainmcul. games, video conferencing or videostreaming while on the move and has had worldwide publicity. Shecurrently leads a few research projects on WCN for mobile intemet,medical and 3G/4G applications including UWB, Combined C-SDMAand SDR. She was the proud recipient of Best Inventor of the AsiaPacific Rim 2006, Korea Invention Promotion Association (KIPA)Special Award 2006 for commending excellent crtbrts to create micro-mobility related MIPv6 invention and Excellent Research Award 2006frotn Ministry of Higher Research, Malaysia.
She is an active researcher of Teman project and MyREN ResearchCommunity. She is a member of IEEE and her research interest spansBroadband and Wireless Communications including Cognitive Radio,Software Defined Radio (SDR), MIMO, UWB and IPv6.
Prof. Borhanuddin Mohd. Ali received hisB.Sc (Hons.) Electrical and ElectronicsEngineering from Loughborough University ofTechnology in 1979, his M.Sc and PhD inElectromagnetics Engineering, from theUniversity of Wales (Cardiff), in 1981 and1985, respectively. He became a lecturer at theDepartment of Electronics and Computer
Engineering, University Putra Malaysia in 1985, Associate professorin 1993 and Professor in 2002.He is now on secondment to MIMOS Bhd. a government research labon ICT, heading tlie centre of Wireless Conuiiiniications. Previously,he served as the director of the Institute of Multimedia and Sofhvarc, aCentre of Excelence within tbe same university, 2002-2006, spent 1year at Celcom R&D in 1995as visiting Scientist, and 2 years at Mightin 1996-97 as a Senior Manager, charged witb coming up withresearch and policy direction for Malaysian Telecommunicationindustry. In 1996 he helped to realize tbe fomiation of Teman project,and later was made the Chairman of the MyREN ResearchCommunity, a national research test-bed. He is a Chartered Engineerand a member ofthe lET (previously IEEE), and Senior Member ofIEEE. He was the Chair of IEEE Malaysia Section 2002-2004, andpreviously the Chair of ComSoc Chapter, 1999-2002. His researchinterest spans Wireless and Broadband Communications, and NetworkEngineering.E-mail; borhan(«jmimos.mv
Dr Raja Syamsul A/mir Kaja AbdullahiVLoived the B.Eng (2000) in Electronie andl.lcctrical Engineering and M.Sc (2001) inCommunication System Engineering from theUniversity of Bimiinghatii. U. Kingdom. He(lion received bis PhD (2005) also from theUniversity of Birmingham majoring in RadarSystem.
His research is dedicated lo the Microwave and Radar Systems, IPv6,WiMax, Wireless Sensor Network and has been involved in thedevelopment of hardware and solhvaie for Radar sensors and WirelessSensor Network. He his among the pioneer on developing a practicalForward Scattering Radar System and tbe system has been adopted forvarious applications including civil, military and mcdieal. In 2006, bebecame the founding Head of the Microwave, Millinieter wave andRadar Systems Laboratory (M"aRS) at UPM. whicb has 2 full timeresearcb assistant. 2 PbD and 5 M.Sc students. The laboratoryspecialized in the radar systems, signal processing and microwave.The laboratory also has a strong collaboration witb MicrowaveIntegrated System Laboratory al the University of Birmingham. UK.F.-mail; rsa(dJ.cne.unm.edu.my
![Page 1: Performance Analysis of Ant Colony's Algorithm: Load ...media.web.britannica.com/ebsco/pdf/352/32441352.pdf · node in a multi-hop fashion [I]. In other hand Wireless mesh networks](https://reader033.documents.pub/reader033/viewer/2022052007/601aec8aa4cb96429f679d25/html5/thumbnails/1.jpg)
![Page 2: Performance Analysis of Ant Colony's Algorithm: Load ...media.web.britannica.com/ebsco/pdf/352/32441352.pdf · node in a multi-hop fashion [I]. In other hand Wireless mesh networks](https://reader033.documents.pub/reader033/viewer/2022052007/601aec8aa4cb96429f679d25/html5/thumbnails/2.jpg)
![Page 3: Performance Analysis of Ant Colony's Algorithm: Load ...media.web.britannica.com/ebsco/pdf/352/32441352.pdf · node in a multi-hop fashion [I]. In other hand Wireless mesh networks](https://reader033.documents.pub/reader033/viewer/2022052007/601aec8aa4cb96429f679d25/html5/thumbnails/3.jpg)
![Page 4: Performance Analysis of Ant Colony's Algorithm: Load ...media.web.britannica.com/ebsco/pdf/352/32441352.pdf · node in a multi-hop fashion [I]. In other hand Wireless mesh networks](https://reader033.documents.pub/reader033/viewer/2022052007/601aec8aa4cb96429f679d25/html5/thumbnails/4.jpg)
![Page 5: Performance Analysis of Ant Colony's Algorithm: Load ...media.web.britannica.com/ebsco/pdf/352/32441352.pdf · node in a multi-hop fashion [I]. In other hand Wireless mesh networks](https://reader033.documents.pub/reader033/viewer/2022052007/601aec8aa4cb96429f679d25/html5/thumbnails/5.jpg)
![Page 6: Performance Analysis of Ant Colony's Algorithm: Load ...media.web.britannica.com/ebsco/pdf/352/32441352.pdf · node in a multi-hop fashion [I]. In other hand Wireless mesh networks](https://reader033.documents.pub/reader033/viewer/2022052007/601aec8aa4cb96429f679d25/html5/thumbnails/6.jpg)
![Page 7: Performance Analysis of Ant Colony's Algorithm: Load ...media.web.britannica.com/ebsco/pdf/352/32441352.pdf · node in a multi-hop fashion [I]. In other hand Wireless mesh networks](https://reader033.documents.pub/reader033/viewer/2022052007/601aec8aa4cb96429f679d25/html5/thumbnails/7.jpg)
![Page 8: Performance Analysis of Ant Colony's Algorithm: Load ...media.web.britannica.com/ebsco/pdf/352/32441352.pdf · node in a multi-hop fashion [I]. In other hand Wireless mesh networks](https://reader033.documents.pub/reader033/viewer/2022052007/601aec8aa4cb96429f679d25/html5/thumbnails/8.jpg)
