ue-initiated cell reselection game for cell load...

Research ArticleUE-Initiated Cell Reselection Game for Cell Load Balancing ina Wireless Network

Jaesung Park

Department of Information Security University of Suwon San 2-2 Wau-ri Bongdam-eup Hwaseong-siGyeonggi-do 445-743 Republic of Korea

Correspondence should be addressed to Jaesung Park jaesungparksuwonackr

Received 2 November 2017 Accepted 24 December 2017 Published 24 January 2018

Academic Editor Hideyuki Takahashi

Copyright copy 2018 Jaesung ParkThis is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

A user changes its serving cell if the quality of experience (QoE) provided by the current serving cell is not satisfactory Since usersreselect cells to increase their QoEs selfishly the system resource efficiency can be deteriorated and a system can be unstable if usersare not driven to cooperate appropriately In this paper inspired by the minority game (MG) model we design a UE-initiated cellreselection policy The MG has a salient characteristic that the number of players who win the game converges to a prespecifiedvalue even though players act selfishly without knowing the actions taken by the other players Using the MG model we devise arule by which each UE plays a cell reselection game We also design a criterion that a system controller uses to determine the resultof a game and public information sent by a system controller to induce implicit cooperation among UEs The simulation resultsshow that compared with noncooperative method the proposed method increases not only the system performance such as cellload balance index and system utility but also the performance of UEs in terms of a downlink data rate and an outage probabilityreceived from a system

1 Introduction

Over a few decades wireless data traffic has been increasedexponentially To cope with the traffic demands variousefforts have been made in many directions to increasenetwork capacity [1] For example MIMO is adopted [2] ina based station and user equipment (UE) small cells aredeployed [3] heterogeneous wireless networks are integrated[4] and new spectrums are secured [5]

Among those increasing cell density is an intuitiveway to increase spectral efficiency Operators deploy basestations (ie cells) and configure their capacities based on theexpected traffic demands in an area However since the trafficdemands vary widely in time and space a mismatch betweenthe deployed resources and the actual traffic demands hap-pens frequently over time Therefore balancing loads amongcells become one of the important issues in awireless networkto increase resource efficiency

In a wireless network cell loads can be controlled at twodifferent phases The first phase is when a UE accesses awireless network A UE needs to be associated with a cell to

receive a service from a network Unlike conventional SINR-based association cell loads can be balanced if a UE knowsthe loads of neighboring cells and selects the least loaded cellThe load-aware cell association is a mapping problem thatfinds pairs of cells and UEs optimizing system performanceindex [6 7] However the channel condition between a celland a UE is dynamic and activities of UEs are also dynamicTherefore load unbalance problem occurs even though UEsselect serving cells based on the cell loads when they firstaccess a network Therefore at the second phase UEs servedby an overloaded cell needs to be transferred to underloadedcells to balance loads among cells In this paper we focus onthe second phase of cell load balancing

In [8 9] an overloaded cell borrows resources fromlightly loaded neighboring cells to resolve cell load unbalanceproblem However channel borrowing method limits theamount of radio resource that can be reused because of thecochannel interference problem

Basically to balance loads among cells some UEs in acongested cell must change their serving cells while the otherUEs in the same cell remain in the cell To achieve this goal

HindawiWireless Communications and Mobile ComputingVolume 2018 Article ID 2712357 8 pageshttpsdoiorg10115520182712357

2 Wireless Communications and Mobile Computing

transmission power control methods are proposed whichadjust transmit power of a set of cells according to the loadsof the cells [10 11] An overloaded cell decreases its coverageby reducing transmit power while underloaded neighboringcells increase their transmit power to expand cell coverageHowever since transmit power of a cell controls the physicalcell boundary coverage holes may occur if the transmitpowers of relevant cells are not adjusted precisely On theother hand mobility load balancing (MLB) is proposed tocontrol the logical cell boundary bymanipulating a parametercalled cell individual offset (CIO) [12] The CIO determinesthe coverage areas between a pair of cells If we denote CIObetween a cell 119894 and a cell 119895 by CIO119894119895 by reducing CIO119894119895 thecoverage of a cell 119894 is reduced while that of cell 119895 increases cor-respondingly Various methods are proposed to adjust CIOsbetween a congested cell and its neighboring cells [13 14]

MLB methods proposed in the literature are drivenby a system In other words cells or a system controllerdetermines the time to redistribute loads among cells andUEs that need to change serving cells Generally cells anda controller exchange necessary information periodically toachieve load balancing UEs passively follow the decisionsmade by cells or a controller in a system However as theservice quality that users expect from a wireless networkincreases users play more active role in a cell reselectionprocess If a cell is congested the service level provided toUEs associated with the cell decreases Since there are a fewavailable cells around a user a user changes its serving cellif the quality of experience (QoE) perceived in the currentserving cell is unsatisfactory

UEs in a congested cell reselect cells to increase theirown QoEs without any coordination among the UEs in thecell Therefore if too many UEs change serving cells thecongested cell can be underloaded rapidly However sincethe locations of UEs in the same cell are similar it is verylikely that they select the same cell as their new serving cellThen the newly selected cell becomes congested right after itaccommodates theUEs in congested neighboring cells whichdrive UEs in the cell to change serving cells accordingly Ifthe process repeats itself a system becomes unstable and thesystem resource efficiency decreasesTherefore the decisionsmade by UEs need to be coordinated by a system to achievesystem stability Since a cell loadmanagementmethod shouldnot restrict the scalability of a system the coordination mustbe arranged in a distributed manner without any messageexchanges among UEs

In this paper we design a UE-initiated cell reselectionpolicy for cell load balancing by adopting the minority game(MG) model The MG model guarantees that the number ofplayers whowin the game converges to a prespecified value ina macroscopic level even though players act selfishly withoutknowing the actions taken by the other players [15 16] Byadopting the MG model we devise a cell reselection ruleEach UE selects one of two actionsmdashchanging its serving cellor staying at the current serving cellmdashselfishly to maximizeits QoE according to the previous game results A systemspecifies a performance goal before operating In this paperwe use the degree of cell load balance as the system goalThe degree of cell load balance changes by the selections

made by UEs a system monitors the degree of cell loadbalance periodically and informs UEs whether the degreeof cell load balance is acceptable by a system or not Sincethe feedback information affects the decisions made by UEsthis manifests an implicit cooperation among UEs even ifthey decide their actions only with their local informationThrough the implicit cooperation among UEs the numberof UEs choosing to change serving cells is determined sothat the degree of cell load balance converges to the targetvalue which avoids wide fluctuation of cell loads Since it isnot necessary for each UE to know the selections made byother UEs our cell reselection policy does not incur signalingoverhead among UEs

The rest of the paper is organized as follows In Section 2we review related works on cell load balancing methods InSection 3 we detail a UE-initiated cell reselection policy Weevaluate the performance of the proposed method throughsimulations in Section 4 Section 5 concludes the paper

2 Related Works

Various technical approaches haven been taken for cell loadbalancing by reselecting pairs of UEs and their serving cellsHeuristic methods are proposed in [17ndash19] In [17] a ripplealgorithm is adopted to resolve load unbalance problemThe authors in [18] propose a heuristic algorithm to solve aload balancing problem in HetNets which are known to beNP-hard In [19] dynamic hysteresis adjustment method isproposed to harmonize the handover parameter optimizationand load balancing for an LTE self-organizing network

In [20ndash22] learning methods are used to bias cellsadaptively in a distributed manner In [20] Q-learning isused to determine the best step size used for CIO adjustmentbetween a congested cell and its neighboring cells In [21]a reinforcement Q-learning algorithm is proposed to resolvecell congestion problem by predicting the load status of eachcell By taking a statistical learning approach the authorsin [22] propose a cell range expansion method for LTEheterogeneous networks Stochastic geometry is also used tomodel and analyze a load balancing algorithm [23 24]Unlikethe optimization methods that maximize the utility functionfor the current network configurations stochastic geometryperforms optimization over the average utility

Game theory is applied to cell load balancing in [25ndash27]Using game models interactive decision-making processesare devised However in these works actions to be taken byother players are predicted by the perfect deductive rational-ity of each player In other words it is assumed that eachplayer has enough information andmakes decisions based onprecise logical reasoningTherefore large signaling overheadis incurred for each player to acquire enough information

We adoptMG to devise a cell reselection policy Howeverunlike the noncooperative and cooperative game theoriesMG does not assume that players acts fully rationally Insteadof a deductive reasoning players in a MG uses inductivereasoning to anticipate the actions taken by the other playersIn addition each player in MG need not know the actionstaken by other players The only information shared by allplayers is a result of game which is binary information In

Wireless Communications and Mobile Computing 3

addition our method differs from above cell load balancemethods in that the cell reselection is made not by a systembut by UEs

3 Cell Reselection Game

31 Minority Game The minority game model explains thephenomenon that even if each player behaves selfishly tomaximize its profit the number of winners converges to thehalf of the total number of players with small fluctuationsaround the equilibriumAMG is played repeatedly at discreteiteration steps Let 119873 be the number of players participatingin a MG Generally 119873 is supposed to be an odd numberWhen a game begins each player selects one group betweentwo groups (say group 0 and group 1) based only on itsstrategy and a past game history Each player maintains themost recent119867 game results

Before joining a game each player has a set of strategiesEach strategy is associated with a score The score representsthe appropriateness of the strategy and is updated accordingto a game resultWhenever each player plays a game it selectsa strategy that has the highest score and decides a group basedon the past history and the selected strategy

Then the players in a group whose number of players isminor win the game A result of a game is encoded by one bitwhich represents theminor group and is shared among all theplayers so that they can update the game history and scoresof strategies A player updates each score for each strategyregardless of whether a strategy is used to choose a group ornot

Since each player does not know the groups that otherplayers select there is no coordination among the playerswhen they decide on a group However as the game is playedrepeatedly implicit cooperation among players emerges bythe game result shared among all the players which convergesthe number of players in a minority group to 1198732 In[28] the properties of the minority game are investigatedanalyticallyThe authors derive an exact solution of themodeland show that not only the individual utility but also theglobal efficiency improves It was shown in [29] that MGgives a near-to-optimal solution when multiple selfish agentscontend for common resources

32 System Model We consider an LTE network integratedwith SDN [30 31] In Notations section we summarize thenotations used in the paper Figure 1 shows a system modelWe assume that cell reselection game is played at discreteiteration steps

We assume that each UE 119909 selects a cell 119894 to be associatedwith based on the downlink (DL) data rate (119903119909119894) that 119894 canprovide 119909 If 119903119909119894 is larger than QoE requirement 119903119905 119909 remainsassociated with 119894 If we denote the transmission power of acell 119894 and the channel gain between 119909 and 119894 by 119875119894 and 119866119909119894respectively SINR at 119909 from 119894 becomes

120593119909119894 = 1198751198941198661199091198941198730 + sum119895isin119862119894 119875119895119866119909119895 (1)

where 1198730 is the power of the white Gaussian noise Theamount of resources 119894 allocated to 119909 depends on the type of

SDN controller

eNB

Backhaul network

UE

eNBeNB

eNB

Cell

Figure 1 System model

scheduler that 119894 uses In this paper we assume that all cellsemploy the proportional fair scheduler Then from [32] thenumber of resources called RB (Resource Block) in the LTEspecifications allocated to 119909 by 119894 is given as

120601119909119894 = 11987711989411987210038161003816100381610038161198801198941003816100381610038161003816 sum119910isin1198801198941119910 (2)

where119877119894119872 is themaximumnumber of RBs that 119894 has and119880119894 isthe set of UEs associated with 119894 |119880119894| represents the cardinalityof the set119880119894Then if we denote the bandwidth of a RB as119882RB119903119909119894 is given as

119903119909119894 = 120601119909119894119882RBlog2 (1 + 120593119909119894) (3)

Thus 119903119909119894 can be increased if other UEs in a congestedcell 119894 change their serving cells or 119909 can find a cell that canprovide higher DL data rate by compensating for the decreasein SINR with the increase in 120593119909119894 UEs whose 119903119909119894 is smallerthan 119903119905 take part in the cell reselection game Since UEscannot exchange information for cooperation each UE mustdetermines autonomously whether to change serving cell ornot We assume that each UE makes the decision once pereach game step

An SDN controller manages a set of cells denoted by 119862At the end of each game step an SDN controller collects cellload information from each cell 119894 in119862 and calculates cell loadbalance index If we denote the load of a cell 119894 at the end ofthe 119896th step by 120588119894(119896) the cell load balance index at the end ofthe 119896th step is defined as

119868lb (119896) =radicsum119894isin119862sum119895isin119862 (120588119894 (119896) minus 120588119895 (119896))2

|119862| (4)


As 119868lb(119896) becomes smaller it means that the loads amongcells are more balanced

The load of a cell can be defined in a few ways [33 34]However to focus on the cell reselection process we use theintuitive definition of a cell load which is the RB utilizationratio From (3) the number of RBs to satisfy 119903119909119894 ge 119903119905 is givenas

120601119909119894 = 119903119905119882RBlog2 (1 + 120593119909119894) (5)

Therefore a cell load is defined as

120588119894 = sum119909isin119880119894 120601119909119894119877119894119872sum119910isin119880119894 1119910

(6)

However we note that any other definition for a cell loadcan easily replace (6) We assume that time is synchronizedamong all UEs and an SDN controller We also assumethat the backhaul networks connecting UEs and eNBs iscongestion free

33 UE-Initiated Cell Reselection Game

331 Initialization Before participating in a cell reselectiongame each UE initializes their strategy tables Each UE 119909has a set of binary actions minus1 1 The action ldquominus1rdquo encodesthe determination that a UE chooses to stay at the currentserving cell while the action ldquo1rdquo means that a UE determinesto change its serving cell A strategy table defines a mappingbetween a game history and a recommended action In otherwords the 119910th strategy table of a UE 119909(119878119909119910) has 2119867 entriesEach entry corresponds to each instance of the past119867 gamehistory and specifies an action to be taken if a past gamehistory is given Namely

119878119909119910 minus1 1119867 997888rarr minus1 1 (7)

Since the size of game history is 119867 there can be 22119867strategy tables Each UE 119909 selects a set of 119878119909 = 1198781199091 119878119909119910 119878119909|119878119909| strategy tables randomly from 22119867 possiblestrategy tables and maintains 119878119909 throughout the game onceit is selected

Each strategy table 119878119909119910 is associated with a score 119904119909119910The initial score for each 119878119909119910 is randomly chosen from [0 1]according to the uniform distribution and 119904119909119910 is updatedevery time a game result is determined

332 Cell Reselection Game At the beginning of each gamestep 119896 UEs whose 119903119909119894 is less than 119903119905 take part in a game AUE playing a cell reselection game chooses the strategy tablethat has the highest score and uses it to determine its actionIn other words the best strategy of a UE 119909 is determined as

119910lowast = argmax119910=1|119878119909|

(119904119909119910 (119896)) (8)

Then 119909 selects an action 119886119909(119896) according to 119878119909119910lowast and ℎ119896where ℎ119896 is the gamehistory at the beginning of the 119896th step If

we denote the action specified in the strategy table 119878119909119910 whenthe game history is ℎ119896 by 119891119909(119878119909119910 ℎ119896) 119909 selects an action as

119886119909 (119896) = 119891119909 (119878119909119910lowast ℎ119896) (9)

At the end of the 119896th step an SDN controller calculates119868lb(119896) and compares it with a threshold value 119868119905Then an SDNcontroller broadcasts a binary information 119892(119896) to UEs whoplayed the game during the 119896th step The binary information119892(119896) means 119868lb(119896) ge 119868119905 if 119892(119896) = 1 Since the cell loadbalance index is higher than 119868119905moreUEsneed to change theirserving cells By disseminating 119892(119896) = 1 an SDN controllerencourages UEs to change their serving cell On the contrary119892(119896) = 0 means 119868lb(119896) lt 119868119905 119868lb(119896) lt 119868119905 implies that cell loadsare balanced Therefore an SDN controller inspires UEs tostay by sending 119892(119896) = 0

Since all systems broadcast system information period-ically the increase in signaling overhead by adding a binaryfeedback119892(119896) to the system information is trivial In additionthe signaling overhead of the proposed method is not relatedto the number of UEs playing a game

After receiving 119892(119896) 119909 updates the history information asfollows

ℎ119896+1 = ℎ119896 mod 2119867 + 119892 (119896) (10)

119909 also updates the scores of strategy tables as follows119904119909119910 (119896 + 1) = 119904119909119910 (119896) minus (2119892 (119896) minus 1) 119886119909 (119896) forall119910 isin 119878119909 (11)

We note that all the scores for all the members in 119878119909are updated regardless of whether a strategy table 119878119909119910 isused to determine 119886119909(119896) or not The same procedure repeatswhenever UEs play a game

4 Performance Evaluation

In this section through simulations we evaluate the per-formance of the proposed method by comparing it withthe method in which UEs select new serving cells selfishly(henceforth we will call this method as NonCo) For sim-ulation studies we construct a 3-tiered cell topology using119862 = 37 cells In Figure 2 we depict the simulation topologyExcept the cells located at the third tier all the cells have6 neighboring cells The cells are homogeneous in that theyhave the same system parameters

We configure the system parameters according to therecommendations by 3GPP [35] The intercell distance is setto 500m We assume the frequency reuse factor is one Thesystem bandwidth is set to 5MHz which is translated to25 RBs According to the LTE specification the bandwidthof a RB is configured as 180KHz The transmit power of acell is set to 46 dBm The antenna height and antenna gainof each cell are set to 32m and 15 dBi respectively Those ofeach UE are set to 15m and 2 dBi respectively To modelthe path loss from a cell to a UE we applied the model 1281+ 376 log(max(119889 119889ref )) where 119889 is the distance between asender and a receiver in km and 119889ref is the path-loss referencedistance which is 0035 km We adopt a lognormal shadowfading model with mean zero and standard deviation of 8 dB


C0

C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

C11

C12

C13

C14

C15

C16

C17

C18

C19

C20

C21

C22

C23

C24

C25

C26

C27

C28

C29

C30

C31

C32

C33

C34

C35

C36

Figure 2 Cell topology

5 10 15 20 25 30 35 400Steps

05

1

15

2

25

3

35

4

45

5

55

Cel

l loa

d

Proposed c7NonCo c7

Proposed c9

NonCo c9

Proposed c17NonCo c17

Figure 3 Cell loads variations of initially congested cells over timewhen 119903119905 = 256Kbps and 119868119905 = 025

1198730 is set to minus11145 dBm We set the minimum coupling lossto 70 dB

Initially we overloaded 7 cells 1198880 1198887 1198889 11988811 11988813 11988815 11988817by deploying 200 UEs in each of these cells while makingthe other 30 cells underloaded by deploying 20 UEs in eachof them The locations of UEs in each cell are randomlyselected according to the uniform distribution We configure119867 = 10 and |119878119909| = 2 for all UEs Since the amount of UEsjoining a game depends on 119903119905 we experimentally configure119868119905 according to 119903119905 For 119903119905 = 256Kbps and 512 Kbps 119868119905 isconfigured as 025 and 13 respectively We investigate howthe proposed method balances cell loads by varying 119903119905

Figure 3 shows the cell load variations of initially con-gested cells over time and Figure 4 shows the variation of cellload balance index when 119903119905 = 256Kbps and 119868119905 = 025 Cell

ProposedNonCo

010203040506070809

11112

Cel

l loa

d-ba

lanc

e ind

ex

5 10 15 20 25 30 35 40 45 500Steps

Figure 4 Variation of cell load balance index over time when 119903119905 =256Kbps and 119868119905 = 025

16000

16500

17000

17500

18000Sy

stem

util

ity

5 10 15 20 25 30 35 40 45 500Steps

Proposed rt = 512 KbpsNonCo rt = 512 Kbps

Proposed rt = 256KbpsNonCo rt = 256Kbps

Figure 5 Variation of system utility over time

loads fluctuate more widely when NonCo is used than whenthe proposed method is used Since UEs selfishly select newserving cells when NonCo is used the load of a congestedcell can be released rapidly However it is more likely that thelightly loaded cell 119894 is overloaded after it accommodates UEsassociated with its neighboring cells Then the DL rates ofUEs in 119894will decrease whichmakesUEs in 119894 attempt to changeserving cells If this process repeats itself the load of a celloscillates over timeWhen the proposedmethod is used eventhough UEs selfishly select whether to change their servingcells or not using its strategy tables the cooperation amongUEs is manifestedTherefore the number of UEs who chooseto change serving cells is controlled tomake 119868lb(119896) converge to119868119905 in a distributedmanner withoutmessage exchanges amongUEs

Figure 5 shows the utility of a system over time Thesystem utility is defined as sum119894isin119862sum119909isin119880119894 log(119903119909119894) We can


02

025

03

035

04

045

05

055

06

065

07

Prop

ortio

n

5 10 15 20 25 30 35 40 45 500Steps



Figure 6 Proportion of UEs whose DL rate is smaller than 119903119905

observe in Figure 5 that the proposedmethod can increase thesystemutility whichmeans that the proposedmethod canusesystem resources more efficiently In addition as 119903119905 becomeslarge the systemutility fluctuates widely whenNonCo is usedwhile the proposed method maintains the system utility at astable level because the proposed method prevents UEs fromchanging serving cells unnecessarily Accordingly as shownin Figure 6 the proportion of UEs whose data rates are lessthan 119903119905 becomes smaller when the proposed method is usedthan when NonCo is used

To analyze the effect of the proposed method on theperformance of UEs we investigate 119903119909119894 and the outageprobability of UE The outage probability is defined as theprobability of 119903119909119894 gt 119903119905 From [36] the outage probability ofa UE 119909 associated with a cell 119894 is given as follows

119874119909119894 = 1minus 119890minus119882RB1198730119860119909119894119866119909119894119875119894 prod

119895isin119862119894

1119866119909119895119875119895119860119909119894119866119909119894119875119894 + 1119866119909119895119875119895

(12)

where 119860119909119894 = 2119903119905119903119909119894 minus 1Figure 7 shows the cumulative proportion of UEs accord-

ing to 119903119909119894 when 119903119905 is 256KbpsThe two methods can improvethe initial distribution They can decrease the proportion ofUEs whose DL data rates are less than 119903119905 while increasingthe proportion of UEs whose DL data rates are larger than710Kbps However the proposed method improves NonCoFor example the proportion of UEs whose DL data ratesare less than 256Kbps is 032 initially NonCo decreases theproportion to 025 and the proposed method decreases theproportion further to 022

Figure 8 shows the cumulative proportion of119874119909119894 when 119903119905= 256Kbps We can observe that the proposed method candecrease the outage probability by the implicit cooperationamong UEs competing system resources

InitialProposedNonCo

500 1000 1500 20000DL rate rxi (Kbps)

0

01

02

03

04

05

06

07

08

09

1

Prob

(X

lex

)

Figure 7 Cumulative proportion of UEs according to 119903119909119894 when 119903119905 =256Kbps


02 04 06 08 10Outage probability

0

01

02

03

04

05

06

07

08

09

1Pr

ob (X

lex

)

Figure 8 Cumulative proportion of UEs according to 119874119909119894 when 119903119905= 256Kbps

5 Conclusions

In this paper we proposed a UE-initiated cell reselectionpolicy for load balancing Inspired by the minority gametheory we design a cell reselection strategy that each UE useswhenever it needs to decide whether to change its servingcell or to stay at the current serving cell Each UE makes adecision selfishly to maximize its profit irrespective of thedecisions made by other UEs However the minority gameplayed among UEs drives the number of UEs who selectto change serving cells to a stable state where the cell loadbalance index of a system converges to a prespecified value Inaddition the proposedmethod does not require any signalingmessages among UEs except the binary feedback from an


SDN controller Therefore the signaling overhead incurredby the proposed method is trivial

Through extensive simulation studies we showed that theproposedmethod increases not only the systemperformancesuch as the load balance level among cells and system utilitybut also the performance of UEs in terms of the DL data rateand the outage probability

Even though the minority game converges the numberof players who win a game to a given value the numberof players winning a game fluctuates around a predefinedtarget value Therefore as a future work we will devisean algorithm that controls the fluctuation to guarantee asystem performance index used to determine a game resultAs another future work we will extend our method tothe environment where there is no entity such as an SDNcontroller that provides a global information so that ourmethod works in a fully distributed manner

Notations

120593119909119894 SINR at a UE 119909 from a cell 119894120601119909119894 The number of RBs allocated to UE 119909 by a

cell 119894119903119909119894 Downlink data rate provided to UE 119909 by a

cell 119894120588119894 Load of a cell 119894119868lb(119896) Cell load balance index at the end of the

119896th step119903119905 QoE threshold of a UE119868119905 Cell load balance threshold119867 The size of game history119878119909 Set of strategy tables of a UE 119909119878119909119910 The 119910th strategy table of a UE 119909119904119909119910 Score value of 119878119909119910119891119909(119878119909119910 ℎ119896) Action specified in 119878119909119910 given game history

of ℎ119896119892(119896) Result of game played at the 119896th step119874119909119894 Outage probability of a UE 119909 in a cell 119894Conflicts of Interest

The author declares that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This research was supported by the Basic Science ResearchProgram through the National Research Foundation ofKorea (NRF) funded by the Ministry of Education (NRF-2015R1D1A1A01060117)

References

[1] A Imran A Zoha and A Abu-Dayya ldquoChallenges in 5G howto empower SONwith big data for enabling 5Grdquo IEEE Networkvol 28 no 6 pp 27ndash33 2014

[2] S Yang and L Hanzo ldquoFifty years of MIMO detection theroad to large-scale MIMOsrdquo IEEE Communications Surveys ampTutorials vol 17 no 4 pp 1941ndash1988 2015

[3] T Nakamura S Nagata A Benjebbour et al ldquoTrends in smallcell enhancements in LTE advancedrdquo IEEE CommunicationsMagazine vol 51 no 2 pp 98ndash105 2013

[4] X Zhang Y Zhang R YuWWang andM Guizani ldquoEnhanc-ing spectral-energy efficiency for LTE-Advanced heterogeneousnetworks a users social pattern perspectiverdquo IEEE WirelessCommunications Magazine vol 21 no 2 pp 10ndash17 2014

[5] T Wang G Li J Ding Q Miao J Li and Y Wang ldquo5GSpectrum is China readyrdquo IEEE Communications Magazinevol 53 no 7 pp 58ndash65 2015

[6] Q Ye B Rong Y Chen M Al-Shalash C Caramanis and J GAndrews ldquoUser association for load balancing in heterogeneouscellular networksrdquo IEEE Transactions on Wireless Communica-tions vol 12 no 6 pp 2706ndash2716 2013

[7] J G Andrews S Singh Q Ye X Lin and H S Dhillon ldquoAnoverview of load balancing in HetNets old myths and openproblemsrdquo IEEEWireless CommunicationsMagazine vol 21 no2 pp 18ndash25 2014

[8] S K Das S K Sen and R Jayaram ldquoA Structured ChannelBorrowing Scheme for Dynamic Load Balancing in CellularNetworksrdquo in Proceedings of the 17th International Conferenceon Distributed Computing Systems pp 116ndash123 1997

[9] S Mitra and S DasBit ldquoLoad balancing strategy using dynamicchannel assignment and channel borrowing in cellular mobileenvironmentrdquo in Proceedings of the 2000 IEEE InternationalConference on Personal Wireless Communications pp 278ndash282December 2000

[10] C Ma R Yin G Yu and J Zhang ldquoReference signal powercontrol for load balancing in downlink LTE-A self-organizingnetworksrdquo in Proceedings of the IEEE 23rd International Sym-posium on Personal Indoor and Mobile Radio CommunicationsPIMRC rsquo12 pp 460ndash464 September 2012

[11] S Yang W Zhan and X Zhao ldquoVirtual cell-breathing basedload balancing in downlink LTE-A self-optimizing networksrdquoinProceedings of the in Proceedings of IEEE International Confer-ence on Wireless Communications Signal Processing WCSP pp1ndash6 2012

[12] H Hu J Zhang X Zheng Y Yang and P Wu ldquoSelf-configuration and self-optimization for LTE networksrdquo IEEECommunications Magazine vol 48 no 2 pp 94ndash100 2010

[13] S S Mwanje L C Schmelz and A Mitschele-Thiel ldquoCognitivecellular networks a q-learning framework for self-organizingnetworksrdquo IEEE Transactions on Network and Service Manage-ment vol 31 no 1 pp 85ndash98 2016

[14] Z Altman S Sallem R Nasri B Sayrac andM Clerc ldquoParticleswarm optimization for mobility load balancing SON in LTEnetworksrdquo in Proceedings of the IEEE Wireless Communicationsand Networking Conference Workshops (WCNCW rsquo14) pp 172ndash177 IEEE Istanbul Turkey April 2014

[15] D Challet and Y-C Zhang ldquoEmergence of cooperation andorganization in an evolutionary gamerdquo Physica A StatisticalMechanics and its Applications vol 246 no 3-4 pp 407ndash4181997

[16] A C C Coolen The Mathematical Theory of Minority GamesStatistical Mechanics of Interacting Agents Oxford UniversityPress Oxford UK 2005

[17] W Li X Duan S Jia L Zhang Y Liu and J Lin ldquoA dynamichysteresis-adjusting algorithm in LTE self-organization net-worksrdquo in Proceedings of the IEEE 75th Vehicular TechnologyConference (VTC Spring rsquo12) pp 1ndash5 Yokohama Japan May2012


[18] W Tangtrongpairoj A Jansang and A Phonphoem ldquoRipplealgorithm An LTE load balance mechanism with distributedand heuristic approachrdquo in Proceedings of the 2013 10th Inter-national Joint Conference on Computer Science and SoftwareEngineering JCSSE rsquo13 pp 111ndash115 May 2013

[19] E Rakotomanana and F Gagnon ldquoFair Load Balancing inHeterogeneous Cellular Networksrdquo in Proceedings of the IEEEInternational Conference on Ubiquitous Wireless BroadbandICUWB rsquo15 pp 1ndash5 October 2015

[20] S S Mwanje and A Mitschele-Thiel ldquoA Q-learning strategyfor LTE mobility Load Balancingrdquo in Proceedings of the 2013IEEE 24th Annual International Symposium on Personal Indoorand Mobile Radio Communications PIMRC rsquo13 pp 2154ndash2158September 2013

[21] J Xu L Tang Q Chen and L Yi ldquoStudy on Based Rein-forcement Q-Learning for Mobile Load Balancing Techniquesin LTE-A HetNetsrdquo in Proceedings of IEEE 17th InternationalConference on Computational Science and Engineering (CSE)pp 1766ndash1771 December 2014

[22] C A S Franco and J R B De Marca ldquoLoad balancing in self-organized heterogeneous LTE networks a statistical learningapproachrdquo in Proceedings of the 7th IEEE Latin-American Con-ference on Communications LATINCOM rsquo15 pp 1ndash5 November2015

[23] S Singh H S Dhillon and J G Andrews ldquoOffloading in het-erogeneous networks modeling analysis and design insightsrdquoIEEE Transactions on Wireless Communications vol 12 no 5pp 2484ndash2497 2013

[24] H Elsawy E Hossain and S Camorlinga ldquoTraffic offloadingtechniques in two-tier femtocell networksrdquo in Proceedings of theIEEE International Conference onCommunications (ICC rsquo13) pp6086ndash6090 IEEE Budapest Hungary June 2013

[25] A Awada B Wegmann I Viering and A Klein ldquoA game-theoretic approach to load balancing in cellular radio networksrdquoin Proceedings of the IEEE 21st International Symposium onPersonal Indoor andMobile Radio Communications PIMRC rsquo10pp 1184ndash1189 September 2010

[26] C Yang ldquoConcurrent mobility load balancing in LTE self-organized networksrdquo in Proceedings of the 21st InternationalConference on Telecommunications ICT rsquo14 pp 288ndash292 May2014

[27] Y Jiang M Yuan Y Bao et al ldquoA game model based on cellload in LTE self-optimizing networkrdquo in Proceedings of the IEEEAdvanced Information Technology Electronic and AutomationControl Conference IAEAC rsquo15 pp 451ndash454 December 2015

[28] M Marsili D Challet and R Zecchina ldquoExact solution ofa modified El Farolrsquos bar problem efficiency and the roleof market impactrdquo Physica A Statistical Mechanics and itsApplications vol 280 no 3 pp 522ndash553 2000

[29] Y She and H-F Leung ldquoAn adaptive strategy for allocation ofresources with gradually or abruptly changing capacitiesrdquo inProceedings of the 20th IEEE International Conference on Toolswith Artificial Intelligence ICTAI rsquo08 pp 415ndash422 November2008

[30] J Costa-Requena ldquoSDN integration in LTE mobile backhaulnetworksrdquo in Proceedings of the 28th International Conferenceon Information Networking (ICOIN rsquo14) pp 264ndash269 IEEEPhuket Thailand February 2014

[31] A Basta W Kellerer M Hoffmann K Hoffmann and E-DSchmidt ldquoA virtual SDN-enabled LTE EPC architecture a casestudy for S-P-gateways functionsrdquo in Proceedings of the IEEE

SDN for Future Networks and Services (SDN4FNS rsquo13) pp 1ndash7November 2013

[32] H J Kushner and P AWhiting ldquoConvergence of proportional-fair sharing algorithms under general conditionsrdquo IEEE Trans-actions onWireless Communications vol 3 no 4 pp 1250ndash12592004

[33] R Kwan R Arnott R Paterson R Trivisonno and M KubotaldquoOn mobility load balancing for LTE systemsrdquo in Proceedingsof the 2010 IEEE 72nd Vehicular Technology Conference FallVTC2010-Fall pp 1ndash5 September 2010

[34] P Szilagyi Z Vincze and C Vulkan ldquoEnhanced mobility loadbalancing optimisation in LTErdquo in Proceedings of the IEEE 23rdInternational Symposium on Personal Indoor and Mobile RadioCommunications PIMRC rsquo12 pp 997ndash1003 September 2012

[35] ldquoLTE Evolved Universal Terrestrial Radio Access (E-UTRA)Radio Frequency (RF) System Scenariosrdquo Tech Rep GPP TR36942 2009 V820

[36] H Boostanimehr and V K Bhargava ldquoUnified and distributedQos-driven cell association algorithms in heterogeneous net-worksrdquo IEEE Transactions on Wireless Communications vol 14no 3 pp 1650ndash1662 2015

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


transmission power control methods are proposed whichadjust transmit power of a set of cells according to the loadsof the cells [10 11] An overloaded cell decreases its coverageby reducing transmit power while underloaded neighboringcells increase their transmit power to expand cell coverageHowever since transmit power of a cell controls the physicalcell boundary coverage holes may occur if the transmitpowers of relevant cells are not adjusted precisely On theother hand mobility load balancing (MLB) is proposed tocontrol the logical cell boundary bymanipulating a parametercalled cell individual offset (CIO) [12] The CIO determinesthe coverage areas between a pair of cells If we denote CIObetween a cell 119894 and a cell 119895 by CIO119894119895 by reducing CIO119894119895 thecoverage of a cell 119894 is reduced while that of cell 119895 increases cor-respondingly Various methods are proposed to adjust CIOsbetween a congested cell and its neighboring cells [13 14]

MLB methods proposed in the literature are drivenby a system In other words cells or a system controllerdetermines the time to redistribute loads among cells andUEs that need to change serving cells Generally cells anda controller exchange necessary information periodically toachieve load balancing UEs passively follow the decisionsmade by cells or a controller in a system However as theservice quality that users expect from a wireless networkincreases users play more active role in a cell reselectionprocess If a cell is congested the service level provided toUEs associated with the cell decreases Since there are a fewavailable cells around a user a user changes its serving cellif the quality of experience (QoE) perceived in the currentserving cell is unsatisfactory

UEs in a congested cell reselect cells to increase theirown QoEs without any coordination among the UEs in thecell Therefore if too many UEs change serving cells thecongested cell can be underloaded rapidly However sincethe locations of UEs in the same cell are similar it is verylikely that they select the same cell as their new serving cellThen the newly selected cell becomes congested right after itaccommodates theUEs in congested neighboring cells whichdrive UEs in the cell to change serving cells accordingly Ifthe process repeats itself a system becomes unstable and thesystem resource efficiency decreasesTherefore the decisionsmade by UEs need to be coordinated by a system to achievesystem stability Since a cell loadmanagementmethod shouldnot restrict the scalability of a system the coordination mustbe arranged in a distributed manner without any messageexchanges among UEs

In this paper we design a UE-initiated cell reselectionpolicy for cell load balancing by adopting the minority game(MG) model The MG model guarantees that the number ofplayers whowin the game converges to a prespecified value ina macroscopic level even though players act selfishly withoutknowing the actions taken by the other players [15 16] Byadopting the MG model we devise a cell reselection ruleEach UE selects one of two actionsmdashchanging its serving cellor staying at the current serving cellmdashselfishly to maximizeits QoE according to the previous game results A systemspecifies a performance goal before operating In this paperwe use the degree of cell load balance as the system goalThe degree of cell load balance changes by the selections

made by UEs a system monitors the degree of cell loadbalance periodically and informs UEs whether the degreeof cell load balance is acceptable by a system or not Sincethe feedback information affects the decisions made by UEsthis manifests an implicit cooperation among UEs even ifthey decide their actions only with their local informationThrough the implicit cooperation among UEs the numberof UEs choosing to change serving cells is determined sothat the degree of cell load balance converges to the targetvalue which avoids wide fluctuation of cell loads Since it isnot necessary for each UE to know the selections made byother UEs our cell reselection policy does not incur signalingoverhead among UEs

The rest of the paper is organized as follows In Section 2we review related works on cell load balancing methods InSection 3 we detail a UE-initiated cell reselection policy Weevaluate the performance of the proposed method throughsimulations in Section 4 Section 5 concludes the paper

2 Related Works

Various technical approaches haven been taken for cell loadbalancing by reselecting pairs of UEs and their serving cellsHeuristic methods are proposed in [17ndash19] In [17] a ripplealgorithm is adopted to resolve load unbalance problemThe authors in [18] propose a heuristic algorithm to solve aload balancing problem in HetNets which are known to beNP-hard In [19] dynamic hysteresis adjustment method isproposed to harmonize the handover parameter optimizationand load balancing for an LTE self-organizing network

In [20ndash22] learning methods are used to bias cellsadaptively in a distributed manner In [20] Q-learning isused to determine the best step size used for CIO adjustmentbetween a congested cell and its neighboring cells In [21]a reinforcement Q-learning algorithm is proposed to resolvecell congestion problem by predicting the load status of eachcell By taking a statistical learning approach the authorsin [22] propose a cell range expansion method for LTEheterogeneous networks Stochastic geometry is also used tomodel and analyze a load balancing algorithm [23 24]Unlikethe optimization methods that maximize the utility functionfor the current network configurations stochastic geometryperforms optimization over the average utility

Game theory is applied to cell load balancing in [25ndash27]Using game models interactive decision-making processesare devised However in these works actions to be taken byother players are predicted by the perfect deductive rational-ity of each player In other words it is assumed that eachplayer has enough information andmakes decisions based onprecise logical reasoningTherefore large signaling overheadis incurred for each player to acquire enough information

We adoptMG to devise a cell reselection policy Howeverunlike the noncooperative and cooperative game theoriesMG does not assume that players acts fully rationally Insteadof a deductive reasoning players in a MG uses inductivereasoning to anticipate the actions taken by the other playersIn addition each player in MG need not know the actionstaken by other players The only information shared by allplayers is a result of game which is binary information In










120593119909119894 = 1198751198941198661199091198941198730 + sum119895isin119862119894 119875119895119866119909119895 (1)


SDN controller

eNB

Backhaul network

UE

eNBeNB

eNB

Cell



120601119909119894 = 11987711989411987210038161003816100381610038161198801198941003816100381610038161003816 sum119910isin1198801198941119910 (2)


119903119909119894 = 120601119909119894119882RBlog2 (1 + 120593119909119894) (3)




|119862| (4)




120601119909119894 = 119903119905119882RBlog2 (1 + 120593119909119894) (5)


120588119894 = sum119909isin119880119894 120601119909119894119877119894119872sum119910isin119880119894 1119910

(6)




119878119909119910 minus1 1119867 997888rarr minus1 1 (7)




119910lowast = argmax119910=1|119878119909|

(119904119909119910 (119896)) (8)



119886119909 (119896) = 119891119909 (119878119909119910lowast ℎ119896) (9)




ℎ119896+1 = ℎ119896 mod 2119867 + 119892 (119896) (10)







C0

C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

C11

C12

C13

C14

C15

C16

C17

C18

C19

C20

C21

C22

C23

C24

C25

C26

C27

C28

C29

C30

C31

C32

C33

C34

C35

C36


5 10 15 20 25 30 35 400Steps

05

1

15

2

25

3

35

4

45

5

55

Cel

l loa

d

Proposed c7NonCo c7

Proposed c9

NonCo c9






ProposedNonCo

010203040506070809

11112

Cel

l loa

d-ba

lanc

e ind

ex

5 10 15 20 25 30 35 40 45 500Steps


16000

16500

17000

17500

18000Sy

stem

util

ity

5 10 15 20 25 30 35 40 45 500Steps







02

025

03

035

04

045

05

055

06

065

07

Prop

ortio

n

5 10 15 20 25 30 35 40 45 500Steps







119895isin119862119894

1119866119909119895119875119895119860119909119894119866119909119894119875119894 + 1119866119909119895119875119895

(12)





500 1000 1500 20000DL rate rxi (Kbps)

0

01

02

03

04

05

06

07

08

09

1

Prob

(X

lex

)




0

01

02

03

04

05

06

07

08

09

1Pr

ob (X

lex

)


5 Conclusions






Notations







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











120593119909119894 = 1198751198941198661199091198941198730 + sum119895isin119862119894 119875119895119866119909119895 (1)


SDN controller

eNB

Backhaul network

UE

eNBeNB

eNB

Cell



120601119909119894 = 11987711989411987210038161003816100381610038161198801198941003816100381610038161003816 sum119910isin1198801198941119910 (2)


119903119909119894 = 120601119909119894119882RBlog2 (1 + 120593119909119894) (3)




|119862| (4)




120601119909119894 = 119903119905119882RBlog2 (1 + 120593119909119894) (5)


120588119894 = sum119909isin119880119894 120601119909119894119877119894119872sum119910isin119880119894 1119910

(6)




119878119909119910 minus1 1119867 997888rarr minus1 1 (7)




119910lowast = argmax119910=1|119878119909|

(119904119909119910 (119896)) (8)



119886119909 (119896) = 119891119909 (119878119909119910lowast ℎ119896) (9)




ℎ119896+1 = ℎ119896 mod 2119867 + 119892 (119896) (10)







C0

C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

C11

C12

C13

C14

C15

C16

C17

C18

C19

C20

C21

C22

C23

C24

C25

C26

C27

C28

C29

C30

C31

C32

C33

C34

C35

C36


5 10 15 20 25 30 35 400Steps

05

1

15

2

25

3

35

4

45

5

55

Cel

l loa

d

Proposed c7NonCo c7

Proposed c9

NonCo c9






ProposedNonCo

010203040506070809

11112

Cel

l loa

d-ba

lanc

e ind

ex

5 10 15 20 25 30 35 40 45 500Steps


16000

16500

17000

17500

18000Sy

stem

util

ity

5 10 15 20 25 30 35 40 45 500Steps







02

025

03

035

04

045

05

055

06

065

07

Prop

ortio

n

5 10 15 20 25 30 35 40 45 500Steps







119895isin119862119894

1119866119909119895119875119895119860119909119894119866119909119894119875119894 + 1119866119909119895119875119895

(12)





500 1000 1500 20000DL rate rxi (Kbps)

0

01

02

03

04

05

06

07

08

09

1

Prob

(X

lex

)




0

01

02

03

04

05

06

07

08

09

1Pr

ob (X

lex

)


5 Conclusions






Notations







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





120601119909119894 = 119903119905119882RBlog2 (1 + 120593119909119894) (5)


120588119894 = sum119909isin119880119894 120601119909119894119877119894119872sum119910isin119880119894 1119910

(6)




119878119909119910 minus1 1119867 997888rarr minus1 1 (7)




119910lowast = argmax119910=1|119878119909|

(119904119909119910 (119896)) (8)



119886119909 (119896) = 119891119909 (119878119909119910lowast ℎ119896) (9)




ℎ119896+1 = ℎ119896 mod 2119867 + 119892 (119896) (10)







C0

C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

C11

C12

C13

C14

C15

C16

C17

C18

C19

C20

C21

C22

C23

C24

C25

C26

C27

C28

C29

C30

C31

C32

C33

C34

C35

C36


5 10 15 20 25 30 35 400Steps

05

1

15

2

25

3

35

4

45

5

55

Cel

l loa

d

Proposed c7NonCo c7

Proposed c9

NonCo c9






ProposedNonCo

010203040506070809

11112

Cel

l loa

d-ba

lanc

e ind

ex

5 10 15 20 25 30 35 40 45 500Steps


16000

16500

17000

17500

18000Sy

stem

util

ity

5 10 15 20 25 30 35 40 45 500Steps







02

025

03

035

04

045

05

055

06

065

07

Prop

ortio

n

5 10 15 20 25 30 35 40 45 500Steps







119895isin119862119894

1119866119909119895119875119895119860119909119894119866119909119894119875119894 + 1119866119909119895119875119895

(12)





500 1000 1500 20000DL rate rxi (Kbps)

0

01

02

03

04

05

06

07

08

09

1

Prob

(X

lex

)




0

01

02

03

04

05

06

07

08

09

1Pr

ob (X

lex

)


5 Conclusions






Notations







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



C0

C1

C2

C3

C4

C5

C6

C7

C8

C9

C10

C11

C12

C13

C14

C15

C16

C17

C18

C19

C20

C21

C22

C23

C24

C25

C26

C27

C28

C29

C30

C31

C32

C33

C34

C35

C36


5 10 15 20 25 30 35 400Steps

05

1

15

2

25

3

35

4

45

5

55

Cel

l loa

d

Proposed c7NonCo c7

Proposed c9

NonCo c9






ProposedNonCo

010203040506070809

11112

Cel

l loa

d-ba

lanc

e ind

ex

5 10 15 20 25 30 35 40 45 500Steps


16000

16500

17000

17500

18000Sy

stem

util

ity

5 10 15 20 25 30 35 40 45 500Steps







02

025

03

035

04

045

05

055

06

065

07

Prop

ortio

n

5 10 15 20 25 30 35 40 45 500Steps







119895isin119862119894

1119866119909119895119875119895119860119909119894119866119909119894119875119894 + 1119866119909119895119875119895

(12)





500 1000 1500 20000DL rate rxi (Kbps)

0

01

02

03

04

05

06

07

08

09

1

Prob

(X

lex

)




0

01

02

03

04

05

06

07

08

09

1Pr

ob (X

lex

)


5 Conclusions






Notations







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



02

025

03

035

04

045

05

055

06

065

07

Prop

ortio

n

5 10 15 20 25 30 35 40 45 500Steps







119895isin119862119894

1119866119909119895119875119895119860119909119894119866119909119894119875119894 + 1119866119909119895119875119895

(12)





500 1000 1500 20000DL rate rxi (Kbps)

0

01

02

03

04

05

06

07

08

09

1

Prob

(X

lex

)




0

01

02

03

04

05

06

07

08

09

1Pr

ob (X

lex

)


5 Conclusions






Notations







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






Notations







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

























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia


ue-initiated cell reselection game for cell load...

Documents