Energy-Efficient Incentives Resource Allocation Scheme in

Research ArticleEnergy-Efficient Incentives Resource Allocation Scheme inCooperative Communication System

Zi Yan Liu 1 Pan Mao12 Li Feng3 and Shi Mei Liu1

1College of Big Data and Information Engineering Guizhou University Guiyang 550025 China2Huaxin Consulting Co Ltd Hangzhou 310014 China3State Grid Chongqing Electric Power Company Chongqing 400014 China

Correspondence should be addressed to Zi Yan Liu gzucommgmailcom

Received 21 January 2018 Revised 25 March 2018 Accepted 24 April 2018 Published 5 June 2018

Academic Editor Zheng Chu

Copyright copy 2018 Zi Yan Liu 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

Appropriate resource allocation has great significance to enhance the energy efficiency (EE) for cooperative communication systemThe objective is to allocate the resource to maximize the energy efficiency in single-cell multiuser cooperative communicationsystem We formulate this problem as subcarrier-based resource allocation and solve it with path planning in graph theory A two-level neural network model is designed in which the users and subcarrier are defined as network nodes And then we proposean improved intelligent water drops algorithm combined with Genetic Algorithm boundary condition and initialization rules ofpath soil quantity are put forwardThe simulation results demonstrate that the proposed resource allocation scheme can effectivelyimprove the energy efficiency and enhance QoS performance

1 Introduction

The rapid energy consumption due to the demands of mobilecommunication services has become a subject of global inter-ests from environment perspective On one hand because ofits slow development and limited capacity battery technologybecomes the bottleneck of limiting the development of theportable terminals [1] on the other hand enormous energyconsumption of the communication industry indirectly leadsto the greenhouse gas emission and increases the operatorsrsquooperating costs Statistics show that in 2009 the power con-sumption of three service providers in China was 289 billiondegrees which equals 441 million tons of coal burning By2014 the energy consumption had been up to 671 milliontons with a 52 increase in 5 years [2 3] Compared to otherindustries it is essential for the communication industryto reduce energy consumption Meanwhile as importantsupport for social informatization there is a very broadprospect for the communication industry to promote energyconservation in the society information industry Thereforedesigning a high energy-efficient communication system hasbecome a consensus of the communication industry

Cooperative communication [4 5] is defined as followsin a cell the adjacent devices with single antenna create a vir-tual MIMO system by sharing their antennas with each other[6 7] to achieve the goal of overcoming the multipath fadingand gaining the benefit of multiantenna space diversity Incooperative communication system a reasonable resourceallocation scheme has great significance in improving thespectral efficiency and reducing the energy consumptionThe traditional design of cooperative communication systemmainly focuses on the improvement of system capacity out-age probability spectral efficiency and other performancesWith the scholarsrsquo attention to the energy consumption ofcommunication industry and the concept of green commu-nication the energy efficiency of cooperative communicationsystem is gaining widespread concernWong et al [8] studiedthe network-level resource scheduling scheme in cooperativecommunication system and proposed a cooperative conceptto obtain the higher energy efficiency at the same time theyalso designed high energy-efficient network architecture Aresource allocation scheme on game theory is established itis proven that the cooperation between users can effectivelyimprove the energy efficiency of the system [9] In [10]

HindawiWireless Communications and Mobile ComputingVolume 2018 Article ID 5452120 8 pageshttpsdoiorg10115520185452120

2 Wireless Communications and Mobile Computing

D

12345

N

2

58

SK

S2

S1

R1

RL

R2

Figure 1 Cooperative communication model

an algorithm is proposed to select the optimal relay in asingle-relay cooperation system Compared with the fixedcooperationmethod the energy consumption is significantlyreduced The wireless sensor networks (WSNs) with energyharvesting and cooperative communication were studied in[11] and an energy-efficient scheduling strategy is proposedthe optimal scheduling problem is solved by using a MarkovDecision Process (MDP) In [12] the process of collaborationwas divided into two time slots An algorithm of relayselection and power allocation is proposed for the minimumBER and maximum system capacity respectively Howeverthese studies have achieved the purpose of improving theenergy efficiency from the perspective of energy consump-tion and have not analyzed the energy efficiency in the formof quantitative indicators QoS in the actual scenario is notconsidered especially the demands of high transmission rate

In this paper we consider a single-cell multiuser cooper-ative communication system To meet the demands of QoSperformance a two-level neural network model based onintelligent water drops (IWDs) algorithm is proposed tosolve the problem of resource allocation with optimal energyefficiency And then the IWDs algorithm combined withGenetic Algorithm (GA) is improved to allocate resourcesflexibly and enhance performance

This paper is organized as follows Section 2 introducesthe system model as well as the function of the proposedoptimal problem In Section 3 a novel algorithm is proposedAfter that the optimal energy-efficient resource allocationscheme is presented in Section 4 In Section 5 numericalresults are depicted Finally Section 6 concludes the paper

2 System Model and Problem Formulation

In this section the proposed system model is presentedfollowed by the optimization problem

21 System Model We consider an uplink single-cell mul-tiuser cooperative communication scenario as shown inFigure 1 and the radius is 600 meters It consists of sourcenode 119878 relay node 119877 and destination node 119863 In particularthe source node is119870 users with call requests and 119871 is the relaynode that participates in the cooperative communication andthe destination node is the base station (BS) Orthogonal Fre-quency Division Multiple Access (OFDMA) is the multiple

access scheme and the available bandwidth 119861 is divided into119873 in which the subcarriers are orthogonal to each otherIt is assumed that the relay node has perfect channel stateinformation (CSI) Further the channels are considered aslarge-scale and small-scale fading That is large-scale fadingis defined as path loss and small-scale fading is Rayleigh fad-ing respectively Moreover the forwarding mode is Decode-and-Forward (DF) and the relay network is operated in half-duplex mode

22 Problem Formulation Assume that the channel statesbetween any two terminals (119878119896 119877119897 119863) are independent ofeach other ℎ119899119894119895 (119894 119895 isin 119878119896 119877119897 119863) indicates the channel fadingbetween the device 119894 and 119895 The channel fading between thenodes can be given by [13]

119864(10038161003816100381610038161003816ℎ119899119878119896119863100381610038161003816100381610038162) = 119889minus120572119878119896119863119864 (10038161003816100381610038161003816ℎ119899119878119896119877119897 100381610038161003816100381610038162) = 119889minus120572119878119896119877119897 119864 (10038161003816100381610038161003816ℎ119899119877119897119863100381610038161003816100381610038162) = 119889minus120572119877119897119863

(1)

where |ℎ119899119878119896119863|2 |ℎ119899119878119896119877119897 |2 |ℎ119899119877119897119863|2 denote the CSI on the subcar-rier 119899 from the user 119878119896 to the destination node 119863 from theuser 119878119896 to the relay119877119897 and from the relay119877119897 to the destinationnode 119863 respectively 119889119894119895 (119894 119895 isin 119878119896 119877119897 119863) represents thedistance between the device 119894 and 119895 119864(sdot) is the averageoperator and 120572 isin [3 5] is the channel fading factor

The energy efficiency is defined as follows [14]

120578EE = 119877119875tot (2)

where 119877 represents the total transfer rate and 119875tot denotes thetotal energy consumption

The transmission procedures of cooperative communica-tion system are divided into two time slots [15] the first is thebroadcast slots during which the source node 119878 broadcaststhe information to the relay node 119877 and the destinationnode 119863 the second is the forwarding slot during which therelay node 119877 processes the received signal and forwards it tothe destination node 119863 In the case of dynamic allocationof transmission time slot it is normalized Particularly 119905represents the transmission time in the broadcast slot and(1minus119905) denotes the time of the relay forwarding slot Due to thedifferent environments in which three terminals are locatedthe channel fadings are independent of each other in twodifferent time slots Two hops of information (broadcastingand forwarding) occupy different time slots respectivelyTherefore in two hops we can use the same subcarrierswithout considering the interference between them but thesystem performance may be limited So the subcarriersshould be allocated independently in two time slots whichinvolves the subcarrier pair matching and allocation issue

The distribution coefficient of subcarrier pairs 119862119896119898119899 isin0 1 is introduced firstly 119862119896119898119899 = 1 represents the case wheresubcarrier 119898 (1 le 119898 le 119873) is paired with 119899 (1 le 119899 le 119873)which is noted as the subcarrier pair (119898 119899) That is user 119896

Wireless Communications and Mobile Computing 3

transfers the information on the subcarrier 119898 at the first slotand on the subcarrier 119899 at the second slot On the contrary119862119896119898119899 = 0means that the subcarrier119898 is not paired with 119899

According to the Shannonrsquos equation after the user 119878119896cooperates with the relay node 119877119897 the transfer rate on thesubcarrier pair (119898 119899) can be written as [16]

119877119896119898119899= min119905119861 log2(1 + 119862

119896119898119899120573119896119897119875119878119873119877119897

10038161003816100381610038161003816ℎ119898119878119896119877119897 100381610038161003816100381610038162) 119905119861 log2(1 + 119862119896119898119899119875119878119873119863

10038161003816100381610038161003816ℎ119898119878119896119863100381610038161003816100381610038162) + (1 minus 119905) 119861 log2(1 + 119862119896119898119899120573119896119897119875119897119873119877119897

10038161003816100381610038161003816ℎ119899119877119897119863100381610038161003816100381610038162) (3)

where |ℎ119898119878119896119877119897 |2 |ℎ119898119878119896119863|2 and |ℎ119899119877119897119863|2 respectively represent thechannel coefficients on the subcarrier pair (119898 119899) between thedevices 119894 119895 (119894 119895 isin 119878119896 119877119897 119863) 120573119896119897 is the relay selection factorand 120573119896119897 = 1 indicates that the 119897th relay node participatesin the 119896th user cooperative communication process and viceversa with no participationThe solution of120573119896119897 is the problemof relay selection in cooperative communication system weuse the method ldquodichotomous maprdquo proposed in [17] 119875119878is the transmitting power of the user node and 119875119897 is thetransmitting power of the relay node Respectively 119873119877119897 119873119863are the noises at the relay node 119877 and the destination node119863

The transmission consumption of user 119896 can be written as119875119896 = (120577 (119905119875119878 + (1 minus 119905) 120573119896119897119875119897) + 119875CT + (119905 + 1) 119875CR) 119896 isin 1 2 119870 (4)

where 120577 is the reciprocal of the drain efficiency of the poweramplifier at the transmitter 119875CT is the fixed circuit powerat the transmitter and 119875CR is the fixed circuit power at thereceiver

From (2) the energy efficiency of user 119896 can be defined asfollows

120578119896EE = 119877119896119898119899119875119896 (5)

and the optimal energy-efficient resource allocation is givenby

120578EE = max119862119896119898119899

119870sum119896=1

119873sum119898=1

119873sum119899=1

(min 119905119861 log2 (1 + (119862119896119898119899120573119896119897119875119878119873119877119897) 10038161003816100381610038161003816ℎ119898119878119896119877119897 100381610038161003816100381610038162) 119905119861 log2 (1 + (119862119896119898119899119875119878119873119863) 10038161003816100381610038161003816ℎ119898119878119896119863100381610038161003816100381610038162)

+ (1 minus 119905) 119861 log2 (1 + (119862119896119898119899120573119896119897119875119897119873119877119897) 10038161003816100381610038161003816ℎ119899119877119897119863100381610038161003816100381610038162)sdot (120577 (119905119875119878 + (1 minus 119905) 120573119896119897119875119897) + 119875CT + (119905 + 1) 119875CR)minus1)

(6)

st C1119862119896119898119899 isin 0 1 C2 119873sum119898

119873sum119899

119862119896119898119899 = 1 forall119896

C3 119870sum119896

119862119896119898119899 le 1 forall119898 119899C4 0 le 119875119878 119875119897 le 119875maxC5119877119905119896119897 ge 119877119905min

(7)

where C1 is the subcarrier matching and the allocationcoefficient indicating that the subcarriers have two states ofcooperative communication and noncooperative communi-cation C2 means that a subcarrier is assigned to one user

only C3 nidicates that a user selects only one subcarrier tocooperate C4 is defined as the power limitation betweenusers and relay nodes and C5 is the transmission rate and119877119905min is the minimum transmission rate which is QoS


A

B

C D

Figure 2 Intelligent water drops algorithm

3 Proposed Optimal ResourceAllocation Scheme

In this section we proposed an improved algorithm to solvethe optimal resource allocation formulated in (6)

31 IntelligentWaterDrops Algorithm Intelligent water drops(IWDs) algorithm is an intelligent algorithm introduced byShah-Hosseini inspired by the flow of natural water dropswhich construct a solution by cooperation with each other[18] The IWDs are associated with two properties theamount of soil in the path and the velocity of the IWDsWhenthe water drops pass through different paths the change ofthe soil quantity on the different paths is different due to thedifference of the path distance When the subsequent waterdrops face different available paths they are more likely tochoose a path with less soil in which the IWDs move fasterAs shown in Figure 2 the initial amounts of soil at two pathsare the same when 119905 = 0 different intelligent water dropswill select the paths ACB and ADBwith the same probabilityAfter the intelligent water drops pass these two paths due toa shorter path of ACB the water drops selecting this path willrun faster and carry a lot of soil leading the soil on the pathACB to be less than that on ADB after the initial iteration At119905 = 1 the intelligent water drops will select a shorter pathACB with a greater probability After repeated feedback atmany times the intelligent water drops will find the shortestpath between A and B The intelligent water drops algorithmdraws on the feedback mechanism of changing soil quantityon the path and completes iterative search

32 Improved Intelligent Water Drops Algorithm The maindrawback of the IWDs algorithm is the low speed at theearly stage of training Because the total amount of soil is thesame on all paths the intelligent water drops will randomlyselect a path even if that path is not the optimal one thiswill change the amount of soil on that path resulting in thephenomenon of path dependence in the subsequent iterationprocess Therefore other water drops are inclined more toselect that path and many invalid searching paths appear

Genetic Algorithm (GA) [19] is an adaptive heuristicsearch algorithm based on the biological evolution processreservingwell-adapted individuals in the process of crossoverand mutation and after several evolutions the optimal

solution of the objective function is obtained It starts theiterative process in individual population which makes iteasy to achieve expansion and algorithm fusion GA has thecharacteristics of implicit parallelism and strong global searchability [20] which can quickly seek the solution in searchspace without trapping into the local optimal solution Thelocal search occurs in GA when the value of gItermax istoo small and GA will stop iterative search without findingthe optimal solution then IWDs algorithm begins thusaffecting the searching efficiency In practice GA encounterspremature convergence problems

As mentioned previously IWDs algorithm is prone tomany ineffective searches in the early stage and the localsearch ability of the GA is limited in later period We pro-posed a novel algorithm to improve the IWDs algorithmwith GA That is in the early stage of the process the globalsearch ability of theGA is applied to achieve the rapid optimalsolution which is used as the initial solution of the IWDsalgorithm Finally the global optimal solution is obtained bythe characteristics of fast convergence of IWDs algorithm

The flowchart of improved IWDs algorithm is shown inFigure 3

Twomain problems in improved IWDs algorithm shouldbe mentioned as follows(1) Value of Boundary Condition 119892119868119905119890119903119898119886119909 In the calculationprocess the convergence is different due to different scale ofdataThe boundary condition119892Itermax of GA and IWDs algo-rithm should be determined by the population sizeThe localsearch occurs in GA when the value of 119892Itermax is too smalland GA will stop iterative search without finding the optimalsolution then the algorithm transfers to IWDs algorithmthus affecting the searching efficiency On the contrary if thevalue of 119892Itermax is too large it leads to the slow convergencein GA due to redundant computing Furthermore the earlymaturing of GA causes the phenomenon of path dependencein IWDs algorithm

According to the convergence analysis of GAbased on theMarkov chain model mentioned in [21] the value of 119892Itermaxis expressed as

119892Itermax = 119873 (119891lowast minus 119891lowast (1198831))2radic119870119901119888119901119898119901119904min119891lowast (1198831) (8)

where 119873 and 119870 are the numbers of subcarriers and usersrespectively 119901119888 and 119901119898 are the probability of crossover andmutation 119901119904min is the minimum selection probability of thenonoptimal individual 119891lowast(1198831) is the fitness value of the bestindividual in initial population and 119891lowast is the best individualfitness value in current population(2) Intelligent Water Drops Soil Quantity Initialization GAachieves an optimal solution and many relatively goodsolutions which are received with different weights and areused for the soil initialization of intelligent water drops Thatis

soil (119894 119895) = Intsoil lowast (1 minus 119871120572lowast minus3sum120591=1

119871120573120591120591 ) (119894 119895) isin 119871 (9)


Initialize the population

Calculate the fitness

Choose

Mutation

Crossover

Get a relatively optimal solution

Initialize the parameters of water drops and path soil quantity

Update the parameters of path soil quantity

Put m water drops on n points

Select the order that every water drop goes through according to the path soil quantity

Get a relatively optimal solution of m water drops

Output the optimal solution

START

Genetic algorithm pre-trainning

Y

N

Y

N

gIterlegIteration times

END

IteＬＧＲ

iIterleiIteＬＧＲ

Figure 3 Flowchart of improved intelligent water drops algorithm

U1

U2

U3

UK

c1

c2

c3

cN

middot middot middot

Figure 4 Two-level neural network model

In particular 119871lowast is the optimal solution generated by GA119871120591 (120591 = 1 2 3) refers to the three optimal solutions generatedby GA and sorted by the fitness value 120572 and 120573120591 respectivelyrefer to the weight of solution at the initialization

4 Optimal Resource Allocation Scheme

41 Optimal Resource Allocation Model The model in (6) isthe combinatorial optimization problem which should meeta series of continuous or discrete conditions to obtain theoptimal resource allocation So we present two-level neuralnetwork model to solve the problems of resource allocationin cooperative communication system and then seek theoptimal path with improved IWDs algorithm

The two-level neural networkmodel is shown in Figure 4It is composed of two types of network nodes and theconnection edges between adjacent nodes The basic unit is

rows A row of 119870 master nodes represents the 119870 users anda row of 119873 secondary nodes represents the 119873 unallocatedsubcarriers in cooperative communication systemThe userrsquosprimary nodes are a total of 119870 + 1 rows A new row insertedbetween two adjacent rows represents secondary nodes ofsubcarriers A path between two roles of primary nodespresents the two selected subcarriers and then this path isweighted according to (6) From the first row to the 119870 +1 column a path passing 2119870 nodes represents a possibleallocation scheme In this way the optimal energy-efficientresource allocation in cooperative communication systemtransfers into the path planning in this two-level neuralnetwork The red path in this figure represents a possibleallocation scheme

42 Optimal Resource Allocation Scheme The optimal re-source allocation flow of improved IWDs algorithm is de-scribed as follows

Step 1 Initialize global static parameters the amount ofIWDs119873IWD = 119870Step 2 Pretrain Genetic AlgorithmStep 21 Initialize the population and the static parametersStep 22 Calculate the population fitness according to (6)Step 23 Operate the heredity andmutation on the populationaccording to the boundary conditions of (8) operate themutation and iteration until the end of loop


Table 1 Simulation parameters

Parameters ValuesTotal system bandwidth B 15MHzNumber of subcarrier N 64Subcarrier mean signal to noise ratio 38 dBPath-loss factor 35Maximum delay extension 4 120583mMaximum Doppler shift 30HzChannel status information update cycle 05msMaximum transmit power 119875max 30 dBmCircuit power 119875119888 27 dBmTransmitter power amplifier efficiency 1120585 38User minimum transfer rate 119877min 12Mbps

Step 3 Place119873IWD water drops on primary nodes on the leftas shown in Figure 4 and iterate each intelligent water dropaccording to Steps 4 and 5

Step 4 Iterate intelligent water dropsStep 41 Initialize the intelligent water dropsrsquo parameters andthe soil quantity on the path according to (9)Step 42 Empty the Tabu list and list all subcarriers asassignableStep 43 With the Tabu list calculate the probability of allselectable paths (subcarrier pair scheme) and select the mostsuitable subcarrier pair (119898 119899) as the userrsquos allocation schemeStep 44 Put the subcarrier119898 in Tabu list 1 and subcarrier 119899 inTabu list 2 which indicates that this subcarrier pair has beenoccupied by the system in two time slotsStep 45 Update the amount of path soil and soil carried bythe intelligent water dropsStep 46 Intelligent water drops pass through the path tothe next primary node set the primary node as the initialposition of the next path selection repeat the steps from Step43 to Step 45 until the water drops reach the model on theright side as shown in Figure 4 and this loop ends

Step 5 At the end of the current iteration calculate theoptimal solution of all paths of water drops and updatethe total amount of the path soil according to the optimalsolution

Step 6 Determine whether the number of iterations satisfies119894Iter le 119894Itermax if it does then repeat Steps 4 and 5 otherwiseit goes to the end of program and optimal solution is shown

5 Numerical Analysis

In this section we evaluate the performance of the proposedoptimal resource allocation scheme via simulation on MAT-LAB 2014b The parameters are shown in Table 1

51 Performance Analysis of Proposed Optimal Scheme InFigure 5 if the number of users in system is 10 the

GA Pre solution

30

35

40

45

50

55

60

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

20 30 40 50 60 70 8010Iterations

075gItermax

gItermax

125gItermax

Figure 5 Energy efficiency comparison of different 119892Itermax

relationship between energy efficiency and 119892Itermax isdepicted The proposed algorithm is combined with GA andIWDs algorithm That is after pretraining by GA at earlyperiod of iterations the performance is further improvedby using IWDs at later period It is obvious that when thevalue of boundary condition is 075lowast119892Itermax the prematureconvergence occurs In this case it is inefficient that IWDs areoperated due to GA getting struck at local optimal solutionWhen the value of boundary condition is 125 lowast 119892Itermax theperformance declines and the premature convergence of GAmakes the phenomenon of path dependence in IWDs occur

52 Performance Comparison of Variant Algorithms Figure 6illustrates the performance comparison of improved IWDsalgorithm in terms of energy efficiency Obviously the energyefficiency obtained by employing the proposed algorithm ismuch higher than that of GA and Ant Colony Optimization(ACO) algorithm mentioned in [22] That is the betterperformance is achieved by using improved IWDs algorithm


Imp-IWDACO

IWDGA

20

30

40

50

60

70

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

4 6 8 10 12 14 162User Numbers

Figure 6 Energy efficiency comparison of different user numbers

Imp-IWDACO

IWDGA

100 150 200 250 300 350 400 450 500 55050Different Transmission Distance (m)

10

20

30

40

50

60

70

80

90

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

Figure 7 The system energy efficiency comparison of differenttransmission distance

However the algorithmic stability of IWDs is poor for itslocal optimum Meanwhile the performance increases at thebeginning and then decreases while the number of usersincreases Due to limitation of system resources when thenumber of users increases it results in a lack of systemresources and furthermore the performance degradation

If the number of users is 10 the radius is 600 metersFigure 7 shows the relationship between energy efficiencyand the distance between the users and BS comparison ofdifferent resource allocation schemes It is proven that theproposed improved IWDs algorithm obtains better perfor-mance than that of algorithms in [22] At the same time withthe increasing of transmission distance the energy efficiencyis gradually reduced When the transmission distance isfar the channel condition becomes more severe so the

Imp-IWDACO

IWDGA

20 30 40 50 60 70 8010Iterations

30

35

40

45

50

55

60

65

70

75

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

Figure 8 The convergence rate of different algorithms

transmitting power increases to overcome the path lossresulting in reduced system energy efficiency

Figure 8 shows the convergence rate of different algo-rithms if the number of users is 10 The improved IWDsalgorithm has a faster convergence rate and obtains a betterperformance while the original IWDs algorithm encounterspremature convergence

6 Conclusion

We have addressed the optimal resource allocation for uplinksingle-cell multiuser cooperative communication systemWith the goal of optimizing energy efficiency the improvedIWDs algorithm has faster convergence rate which achievesbetter resource allocation The performance evaluationresults demonstrate the effectiveness of the proposed solu-tion

It is important to notice that a subcarrier is assigned toa user only in this paper In the practical communicationscenario usersrsquo data are abrupt and asymmetric In the futurea real-time resource allocation strategy can be established tomeet the demands of rapid development of mobile services

Data Availability

The network parameter data used to support the findings ofthis study are included within the article

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work was supported by the Natural Science Foun-dation of Guizhou Province (Grant no [2016]1054) JointNatural Science Foundation of Guizhou Province (Grant


no LH[2017]7226) Academic Talent Training and Innova-tion Exploration Project of Guizhou University (Grant no[2017]5788) andGraduate Student Innovation Foundation ofGuizhou University (Grant no 2017015)

References

[1] G Auer V Giannini C Desset I Godor P Skillermark andMOlsson ldquoHowmuch energy is needed to run awireless networkrdquoWireless Communications vol 18 no 5 pp 40ndash49 2011

[2] Y Zhao Y Z Wang and H Y Yang ldquoIntroduction to greencommunication technologiesrdquo Information ampCommunicationsvol 5 pp 254-255 2016

[3] J C Xu ldquoCurrent Situation and Countermeasures of EnergyConservation and Emission Reduction in TelecommunicationIndustry in ChinardquoModern Economic Information vol 7 p 3042014

[4] A Sendonaris E Erkip and B Aazhang ldquoUser cooperationdiversity-part I system descriptionrdquo IEEE Transactions onCommunications vol 51 no 11 pp 1927ndash1938 2003

[5] J N LanemanDN Tse andGWornell ldquoCooperative diversityin wireless networks efficient protocols and outage behaviorrdquoIEEE Transactions on Information Theory vol 50 no 12 pp3062ndash3080 2004

[6] J C Xiao Development of Virtual Massive MIMO ChannelMeasurement System and Analysis of Large Scale Fading BeijingJiaotong University 2016

[7] L Li and C Chigan ldquoA Virtual MIMO based anti-jammingstrategy for cognitive radio networksrdquo inProceedings of the IEEEInternational Conference on Communications pp 1ndash6 2016

[8] C Y Wong R S Cheng K B Letaief and R D Murch ldquoMul-tiuser OFDM with adaptive subcarrier bit and power alloca-tionrdquo IEEE Journal on Selected Areas in Communications vol17 no 10 pp 1747ndash1758 1999

[9] K N Pappi P D Diamantoulakis H Otrok and G K Kara-giannidis ldquoCloud compute-and-forward with relay coopera-tionrdquo IEEE Transactions on Wireless Communications vol 14no 6 pp 3415ndash3428 2015

[10] S Yousaf N Javaid U Qasim N Alrajeh Z A Khan and MAhmed ldquoTowards reliable and energy-efficient incrementalcooperative communication for wireless body area networksrdquoSensors vol 16 no 3 2016

[11] H Li N Jaggi and B Sikdar ldquoRelay scheduling for cooperativecommunications in sensor networks with energy harvestingrdquoIEEE Transactions on Wireless Communications vol 10 no 9pp 2918ndash2928 2011

[12] ADoosti-Aref andA Ebrahimzadeh ldquoAdaptive Relay Selectionand Power Allocation for OFDM Cooperative UnderwaterAcoustic Systemsrdquo IEEETransactions onMobile Computing vol17 no 1 pp 1ndash15 2018

[13] X Yin Long Research on Energy Efficiency Based CooperativeCommunication System 2014

[14] V Rodoplu and T H Meng ldquoBits-per-joule capacity of energy-limited wireless networksrdquo IEEE Transactions onWireless Com-munications vol 6 no 3 pp 857ndash864 2007

[15] W-S Lai T-H Chang and T-S Lee ldquoDistributed dynamicresource allocation for ofdma-based cognitive small cell net-works using a regret-matching game approachrdquo Game TheoryFramework Applied to Wireless Communication Networks 2016

[16] Y Xu Z Bai B Wang et al ldquoEnergy-efficient power allo-cation scheme for multi-relay cooperative communicationsrdquo

in Proceedings of the International Conference on AdvancedCommunication Technology pp 260ndash264 IEEE 2014

[17] Z Y Liu H Tang P Mao S M Liu and L Feng ldquoRelay selec-tion in cooperative communicationwith bipartite graphrdquoAppli-cation Research of Computers vol 4 2018

[18] H Shah-Hosseini ldquoProblem solving by intelligent water dropsrdquoin Proceedings of the Evolutionary Computation CEC IEEECongress pp 3226ndash3231 2007

[19] Y JMa andWX Yun ldquoResearch progress of genetic algorithmrdquoApplication Research of Computers vol 4 pp 1201ndash1210 2012

[20] M Pei Ant Colony Optimization Algorithm in the Allocationof Cloud Computing Resources Shandong Normal University2015

[21] S Q Kuang Parameter Adaptive Controlling and ConvergenceTheory for Genetic Algorithms Central South University 2009

[22] A Zainaldin H Halabian and I Lambadaris ldquoJoint ResourceAllocation and Relay Selection in LTE-Advanced NetworkUsing Hybrid Co-Operative Relaying and Network CodingrdquoIEEE Transactions on Wireless Communications vol 15 no 6pp 4348ndash4361 2016

International Journal of

AerospaceEngineeringHindawiwwwhindawicom Volume 2018

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018


Active and Passive Electronic Components

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawiwwwhindawicom

Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Control Scienceand Engineering

Journal of



Journal ofEngineeringVolume 2018

SensorsJournal of



RotatingMachinery


Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation


Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

Submit your manuscripts atwwwhindawicom


D

12345

N

2

58

SK

S2

S1

R1

RL

R2

Figure 1 Cooperative communication model

an algorithm is proposed to select the optimal relay in asingle-relay cooperation system Compared with the fixedcooperationmethod the energy consumption is significantlyreduced The wireless sensor networks (WSNs) with energyharvesting and cooperative communication were studied in[11] and an energy-efficient scheduling strategy is proposedthe optimal scheduling problem is solved by using a MarkovDecision Process (MDP) In [12] the process of collaborationwas divided into two time slots An algorithm of relayselection and power allocation is proposed for the minimumBER and maximum system capacity respectively Howeverthese studies have achieved the purpose of improving theenergy efficiency from the perspective of energy consump-tion and have not analyzed the energy efficiency in the formof quantitative indicators QoS in the actual scenario is notconsidered especially the demands of high transmission rate

In this paper we consider a single-cell multiuser cooper-ative communication system To meet the demands of QoSperformance a two-level neural network model based onintelligent water drops (IWDs) algorithm is proposed tosolve the problem of resource allocation with optimal energyefficiency And then the IWDs algorithm combined withGenetic Algorithm (GA) is improved to allocate resourcesflexibly and enhance performance

This paper is organized as follows Section 2 introducesthe system model as well as the function of the proposedoptimal problem In Section 3 a novel algorithm is proposedAfter that the optimal energy-efficient resource allocationscheme is presented in Section 4 In Section 5 numericalresults are depicted Finally Section 6 concludes the paper

2 System Model and Problem Formulation

In this section the proposed system model is presentedfollowed by the optimization problem

21 System Model We consider an uplink single-cell mul-tiuser cooperative communication scenario as shown inFigure 1 and the radius is 600 meters It consists of sourcenode 119878 relay node 119877 and destination node 119863 In particularthe source node is119870 users with call requests and 119871 is the relaynode that participates in the cooperative communication andthe destination node is the base station (BS) Orthogonal Fre-quency Division Multiple Access (OFDMA) is the multiple

access scheme and the available bandwidth 119861 is divided into119873 in which the subcarriers are orthogonal to each otherIt is assumed that the relay node has perfect channel stateinformation (CSI) Further the channels are considered aslarge-scale and small-scale fading That is large-scale fadingis defined as path loss and small-scale fading is Rayleigh fad-ing respectively Moreover the forwarding mode is Decode-and-Forward (DF) and the relay network is operated in half-duplex mode

22 Problem Formulation Assume that the channel statesbetween any two terminals (119878119896 119877119897 119863) are independent ofeach other ℎ119899119894119895 (119894 119895 isin 119878119896 119877119897 119863) indicates the channel fadingbetween the device 119894 and 119895 The channel fading between thenodes can be given by [13]

119864(10038161003816100381610038161003816ℎ119899119878119896119863100381610038161003816100381610038162) = 119889minus120572119878119896119863119864 (10038161003816100381610038161003816ℎ119899119878119896119877119897 100381610038161003816100381610038162) = 119889minus120572119878119896119877119897 119864 (10038161003816100381610038161003816ℎ119899119877119897119863100381610038161003816100381610038162) = 119889minus120572119877119897119863

(1)

where |ℎ119899119878119896119863|2 |ℎ119899119878119896119877119897 |2 |ℎ119899119877119897119863|2 denote the CSI on the subcar-rier 119899 from the user 119878119896 to the destination node 119863 from theuser 119878119896 to the relay119877119897 and from the relay119877119897 to the destinationnode 119863 respectively 119889119894119895 (119894 119895 isin 119878119896 119877119897 119863) represents thedistance between the device 119894 and 119895 119864(sdot) is the averageoperator and 120572 isin [3 5] is the channel fading factor

The energy efficiency is defined as follows [14]

120578EE = 119877119875tot (2)

where 119877 represents the total transfer rate and 119875tot denotes thetotal energy consumption

The transmission procedures of cooperative communica-tion system are divided into two time slots [15] the first is thebroadcast slots during which the source node 119878 broadcaststhe information to the relay node 119877 and the destinationnode 119863 the second is the forwarding slot during which therelay node 119877 processes the received signal and forwards it tothe destination node 119863 In the case of dynamic allocationof transmission time slot it is normalized Particularly 119905represents the transmission time in the broadcast slot and(1minus119905) denotes the time of the relay forwarding slot Due to thedifferent environments in which three terminals are locatedthe channel fadings are independent of each other in twodifferent time slots Two hops of information (broadcastingand forwarding) occupy different time slots respectivelyTherefore in two hops we can use the same subcarrierswithout considering the interference between them but thesystem performance may be limited So the subcarriersshould be allocated independently in two time slots whichinvolves the subcarrier pair matching and allocation issue

The distribution coefficient of subcarrier pairs 119862119896119898119899 isin0 1 is introduced firstly 119862119896119898119899 = 1 represents the case wheresubcarrier 119898 (1 le 119898 le 119873) is paired with 119899 (1 le 119899 le 119873)which is noted as the subcarrier pair (119898 119899) That is user 119896




119877119896119898119899= min119905119861 log2(1 + 119862

119896119898119899120573119896119897119875119878119873119877119897

10038161003816100381610038161003816ℎ119898119878119896119877119897 100381610038161003816100381610038162) 119905119861 log2(1 + 119862119896119898119899119875119878119873119863

10038161003816100381610038161003816ℎ119898119878119896119863100381610038161003816100381610038162) + (1 minus 119905) 119861 log2(1 + 119862119896119898119899120573119896119897119875119897119873119877119897

10038161003816100381610038161003816ℎ119899119877119897119863100381610038161003816100381610038162) (3)





120578119896EE = 119877119896119898119899119875119896 (5)


120578EE = max119862119896119898119899

119870sum119896=1

119873sum119898=1

119873sum119899=1

(min 119905119861 log2 (1 + (119862119896119898119899120573119896119897119875119878119873119877119897) 10038161003816100381610038161003816ℎ119898119878119896119877119897 100381610038161003816100381610038162) 119905119861 log2 (1 + (119862119896119898119899119875119878119873119863) 10038161003816100381610038161003816ℎ119898119878119896119863100381610038161003816100381610038162)


(6)

st C1119862119896119898119899 isin 0 1 C2 119873sum119898

119873sum119899

119862119896119898119899 = 1 forall119896

C3 119870sum119896


(7)




A

B

C D















119871120573120591120591 ) (119894 119895) isin 119871 (9)




Choose

Mutation

Crossover








START


Y

N

Y

N


END

IteＬＧＲ



U1

U2

U3

UK

c1

c2

c3

cN




















GA Pre solution

30

35

40

45

50

55

60

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

20 30 40 50 60 70 8010Iterations

075gItermax

gItermax

125gItermax





Imp-IWDACO

IWDGA

20

30

40

50

60

70

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

4 6 8 10 12 14 162User Numbers


Imp-IWDACO

IWDGA


10

20

30

40

50

60

70

80

90

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




Imp-IWDACO

IWDGA

20 30 40 50 60 70 8010Iterations

30

35

40

45

50

55

60

65

70

75

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




6 Conclusion



Data Availability




Acknowledgments




References


























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia





119877119896119898119899= min119905119861 log2(1 + 119862

119896119898119899120573119896119897119875119878119873119877119897

10038161003816100381610038161003816ℎ119898119878119896119877119897 100381610038161003816100381610038162) 119905119861 log2(1 + 119862119896119898119899119875119878119873119863

10038161003816100381610038161003816ℎ119898119878119896119863100381610038161003816100381610038162) + (1 minus 119905) 119861 log2(1 + 119862119896119898119899120573119896119897119875119897119873119877119897

10038161003816100381610038161003816ℎ119899119877119897119863100381610038161003816100381610038162) (3)





120578119896EE = 119877119896119898119899119875119896 (5)


120578EE = max119862119896119898119899

119870sum119896=1

119873sum119898=1

119873sum119899=1

(min 119905119861 log2 (1 + (119862119896119898119899120573119896119897119875119878119873119877119897) 10038161003816100381610038161003816ℎ119898119878119896119877119897 100381610038161003816100381610038162) 119905119861 log2 (1 + (119862119896119898119899119875119878119873119863) 10038161003816100381610038161003816ℎ119898119878119896119863100381610038161003816100381610038162)


(6)

st C1119862119896119898119899 isin 0 1 C2 119873sum119898

119873sum119899

119862119896119898119899 = 1 forall119896

C3 119870sum119896


(7)




A

B

C D















119871120573120591120591 ) (119894 119895) isin 119871 (9)




Choose

Mutation

Crossover








START


Y

N

Y

N


END

IteＬＧＲ



U1

U2

U3

UK

c1

c2

c3

cN




















GA Pre solution

30

35

40

45

50

55

60

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

20 30 40 50 60 70 8010Iterations

075gItermax

gItermax

125gItermax





Imp-IWDACO

IWDGA

20

30

40

50

60

70

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

4 6 8 10 12 14 162User Numbers


Imp-IWDACO

IWDGA


10

20

30

40

50

60

70

80

90

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




Imp-IWDACO

IWDGA

20 30 40 50 60 70 8010Iterations

30

35

40

45

50

55

60

65

70

75

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




6 Conclusion



Data Availability




Acknowledgments




References


























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



A

B

C D















119871120573120591120591 ) (119894 119895) isin 119871 (9)




Choose

Mutation

Crossover








START


Y

N

Y

N


END

IteＬＧＲ



U1

U2

U3

UK

c1

c2

c3

cN




















GA Pre solution

30

35

40

45

50

55

60

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

20 30 40 50 60 70 8010Iterations

075gItermax

gItermax

125gItermax





Imp-IWDACO

IWDGA

20

30

40

50

60

70

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

4 6 8 10 12 14 162User Numbers


Imp-IWDACO

IWDGA


10

20

30

40

50

60

70

80

90

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




Imp-IWDACO

IWDGA

20 30 40 50 60 70 8010Iterations

30

35

40

45

50

55

60

65

70

75

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




6 Conclusion



Data Availability




Acknowledgments




References


























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia





Choose

Mutation

Crossover








START


Y

N

Y

N


END

IteＬＧＲ



U1

U2

U3

UK

c1

c2

c3

cN




















GA Pre solution

30

35

40

45

50

55

60

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

20 30 40 50 60 70 8010Iterations

075gItermax

gItermax

125gItermax





Imp-IWDACO

IWDGA

20

30

40

50

60

70

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

4 6 8 10 12 14 162User Numbers


Imp-IWDACO

IWDGA


10

20

30

40

50

60

70

80

90

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




Imp-IWDACO

IWDGA

20 30 40 50 60 70 8010Iterations

30

35

40

45

50

55

60

65

70

75

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




6 Conclusion



Data Availability




Acknowledgments




References


























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia












GA Pre solution

30

35

40

45

50

55

60

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

20 30 40 50 60 70 8010Iterations

075gItermax

gItermax

125gItermax





Imp-IWDACO

IWDGA

20

30

40

50

60

70

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

4 6 8 10 12 14 162User Numbers


Imp-IWDACO

IWDGA


10

20

30

40

50

60

70

80

90

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




Imp-IWDACO

IWDGA

20 30 40 50 60 70 8010Iterations

30

35

40

45

50

55

60

65

70

75

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




6 Conclusion



Data Availability




Acknowledgments




References


























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



Imp-IWDACO

IWDGA

20

30

40

50

60

70

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)

4 6 8 10 12 14 162User Numbers


Imp-IWDACO

IWDGA


10

20

30

40

50

60

70

80

90

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




Imp-IWDACO

IWDGA

20 30 40 50 60 70 8010Iterations

30

35

40

45

50

55

60

65

70

75

Ener

gy E

ffici

ency

(Mbi

tsJo

ule)




6 Conclusion



Data Availability




Acknowledgments




References


























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia




References


























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia




RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia


Documents

Energy-Efficient Incentives Resource Allocation Scheme in