Context-aware Cluster Based Devce-to-Devce Communcaton to Serve Machne Type Communcatons J Langha, Lu Man, Hans D. Schotten Char of Wreless Communcaton, Unversty of Kaserslautern, Germany {j,manlu,schotten}@et.un-kl.de arxv:1708.09661v1 [cs.ni] 31 Aug 2017 Abstract Bllons of Machne Type Communcaton (MTC) devces are foreseen to be deployed n next ten years and therefore potentally open a new market for next generaton wreless network. However, MTC applcatons have dfferent characterstcs and requrements compared wth the servces provded by legacy cellular networks. For nstance, an MTC devce sporadcally requres to transmt a small data packet contanng nformaton generated by sensors. At the same tme, due to the massve deployment of MTC devces, t s neffcent to charge ther batteres manually and thus a long battery lfe s requred for MTC devces. In ths sense, legacy networks desgned to serve human-drven traffcs n real tme can not support MTC effcently. In order to mprove the avalablty and battery lfe of MTC devces, context-aware devce-to-devce (D2D) communcaton s exploted n ths paper. By applyng D2D communcaton, some MTC users can serve as relays for other MTC users who experence bad channel condtons. Moreover, sgnalng schemes are also desgned to enable the collecton of context nformaton and support the proposed D2D communcaton scheme. Last but not least, a system level smulator s mplemented to evaluate the system performance of the proposed technologes and a large performance gan s shown by the numercal results. I. INTRODUCTION As one of the new emergng servces, Machne Type Communcaton (MTC) [1] [2] s consdered by many researchers and experts as an mportant servce n the comng ffth generaton (5G) cellular network. In the 4G network, legacy LTE-A network was desgned to offer hgh data rate, low latency, hgh spectrum effcency and hgh system capacty. Thus, 4G network experences techncal challenges to offer MTC servces snce dfferent consderatons and requrements are posed on MTC servces, e.g., a massve deployment of devces, a low devce complexty and a long devce battery lfe. In the thrd Generaton Partnershp Project (3GPP), related studes are conducted to evolve legacy network to meet the requrements of MTC servces. For nstance, a new user equpment (UE) category (UE category 0 ) s ntroduced to reduce devce complexty and power consumpton n [2]. In addton, an extended dscontnuous recepton (DRX) s consdered n [3] to reduce battery consumpton where longer sleep cycles are exploted and optmzed for delay-tolerant MTC applcatons. Moreover, removal of the power amplfer s proposed n [2] to reduce devce cost. However, uplnk c Copyrght 2017 IEEE coverage s reduced n ths case due to a lower maxmal transmsson power. Another challenge faced by MTC s the extra penetraton loss of 20 db due to the deep-n-door deployment of users (UEs) [4]. These MTC UEs located deep-n-door are referred as remote UEs due to ther bad cellular channel condtons. Compared wth base staton (BS), snce a lower transmsson power s avalable at UE, t s more challengng to mantan the network coverage n uplnk. Exstng soluton to mantan uplnk coverage s to ether use narrow band transmsson or explot massve transmsson tme nterval (TTI) bundlng [4]. Though both schemes help n enhancng the MTC avalablty, a large resource usage at system level and a battery dran at devce level are deduced. In another work [5], t s proposed to explot the 3GPP defned relay nodes (RNs) to mprove coverage for MTC. However, RNs are usually exploted for moble broadband servce and they normally locate n areas wth busy human actvtes. Thus, deployment of RNs n deep-n-door and rural area cannot always be assumed. As one of the crtcal techncal enablers for 5G cellular network, devce-to-devce (D2D) communcaton [6] [7] opens the opportunty to mprove the performance of cellular networks. The motvaton of explotng D2D communcaton was to ether offload cellular traffc to local nformaton exchange procedure or to enable a drect D2D communcaton to acheve low latency [8]. In research work [9] [10], the appled scenaro of D2D communcaton s extended to MTC servces. In these work, a normal cellphone wth D2D dscovery and communcaton capablty acts as a relay for other sensors. Together wth academc communty, 3GPP also consders explotng cellphones as relays for MTC applcatons [11]. In ths approach, several drawbacks exst, as lsted n the followng tems. 1) D2D dscovery procedure s carred out every tme when a remote UE s paged or has uplnk date n ts buffer to transmt. In ths way, extra power consumpton s deduced. 2) Snce the D2D parng s performed n a dstrbuted manner wthout any help from BS, t brngs a lost n global awareness. For nstance, once a devce s pared wth one relay devce, t can not get served by other potental relay devces though they mght be better choce.
3) Last but not least. presence of cellphones n a deep-ndoor scenaro and rural area can not always be expected. Instead of usng cellphones as relays, one of the MTC devces s apponted as an aggregator n [12], where full knowledge of Channel Qualty Indcators (CQIs) s necessary to be avalable n BS to set up D2D pars properly. Thus, f a cluster conssts of M devces, then BS needs to collect M (M 1) 2 CQIs nformaton wth each CQI representng the channel stuaton between two devces. Ths requrement leads to a very cumbersome sgnalng procedure wth hgh power consumpton and sgnalng overload. In ths work, we nspect on how to mprove the avalablty and battery lfe of MTC UEs by explotng D2D communcaton. MTC UEs are assumed to be statcally dstrbuted nsde buldngs. In the proposed approach, MTC UEs located deepn-door have the opportunty to set up D2D lnks wth relay MTC UEs. Compared wth the exstng work n lterature, our proposal contrbutes to the followng aspects. Sgnalng schemes wth low overhead are provded to support the explotaton of D2D communcaton n the consdered scenaro. The sgnalng schemes enable an effcent collecton of context nformaton. Wth the help from the collected context nformaton, BS can optmze the transmsson mode (TM) of each user and mprove the system performance. Our work s organzed as follows. At begnnng, a cluster based transmsson mode selecton (TMS) scheme s proposed n Sect. II. After that, correspondng sgnalng schemes are provded n Sect. III to support the proposed D2D communcaton wth a low sgnalng load. In Sect. IV, methodology used to evaluate the system performance of MTC servces s stated wth detals. Smulaton results are gven n Sect. V where a large gan n terms of avalablty and battery lfe can be seen. Fnally, we conclude ths work n Sect. VI. II. TRANSMISSION MODE SELECTION BASED ON VIRTUAL SECTORS As mentoned before, D2D communcaton s exploted n ths work to facltate the uplnk reports from remote UEs. In ths scheme, three dfferent transmsson modes exst, as followng: cellular transmsson mode, n whch the devces upload ther reports to BS wth cellular lnks; relay transmsson mode, n whch the devces are confgured by network to relay the reports from remote UEs and meanwhle transmt ther own reports to BS; D2D transmsson mode, n whch the remote UEs transmt ther reports to relay UEs. In order to adapt to any system changes n real tme, MTC UEs are dynamcally confgured wth ther transmsson modes by BS. Fg. 1 provdes a graphcal descrpton of ths scenaro. As t can be seen, sensor 4 and sensor 5 experence bad channel condtons for ther cellular lnks and thus are referred as remote UEs. Meanwhle, sensor 2 s seen by BS as an optmal relay node for sensor 4 and 5. Thus, D2D connectons are establshed between sensor 2 and sensor 4, also between sensor 2 and sensor 5. After that, uplnk data of sensors 4 and 5 are transmtted to BS through sensor 2. Besdes beng a relay node for remote UEs, sensor 2 also transmts ts own packet to BS. Moreover, sensors 1 and 3 are confgured as normal cellular UEs and they are only responsble for transmssons of ther own packets. From effcency pont of vew, D2D communcaton should be appled n cases where transmtter and recever are nearby each other. Thus, a clusterng approach s requred at BS to make sure that a relay UE only serves remote UEs n ts proxmty. Afterwards, BS needs to select proper transmsson mode for each UE, takng nto account of the context nformaton (e.g., channel state nformaton (CSI) between BS and the UE, locaton and battery level nformaton). Therefore, the proposed context-aware D2D communcaton can be dvded nto two steps: 1) clusterng devces nto dfferent groups; 2) selecton of transmsson mode for each UE. A. Vrtual clusterng In ths secton, four dfferent methods wth dfferent consderatons are ntroduced to group devces. Moreover, for devces locatng near the BS, ther propagaton losses are relatvely low and thus D2D communcaton s not appled for these UEs. The area wthout applcaton of D2D communcaton can be represented by a cycle wth a radus of R n. 1) Geometrcal clusterng: In ths method, the coverage area of one BS s sectorzed geometrcally, as shown n Fg. 2. The area covered by the -th cluster can be represented by a radus of R and an angle of ϕ, as Fg. 1. scenaro descrpton R start < R R end, (1)
the operaton of ths algorthm. These are the two dfference of ths scheme compared wth the K-means clusterng method. 4) Dstance plus CSI based clusterng: In ths scheme, not only the locaton nformaton of UEs are consdered, but also the CSI between each UE and the BS. The dfference compared wth dstance based clusterng method s that, the K centrods are selected from devces whch have cellular SNR values hgher than a pre-defned threshold. B. Transmsson mode selecton The task of TMS s to confgure devces n each cluster so that each devce s aware of the transmsson mode t should apply for the uplnk report. Multple context nformaton are taken nto account n ths work to acheve an effcent TMS algorthm. As mentoned before, the battery lfe requrement of MTC devces can be up to 10 years [4]. For UEs who cannot meet the battery lfe requrement, D2D communcaton s exploted. The equaton below descrbes the condton of remote devces whose battery lfe requrement cannot be met by cellular lnks: Fg. 2. geometrcal clusterng ϕ start < ϕ ϕ end. (2) where R start and R end represent the dstances from the orgn (.e. the BS) to nner and outer crcles of the -th cluster. Moreover, ϕ start and ϕ end represent the reference angles between whch the -th cluster covers. The number of clusters covered by one BS s functon of A sector, whch s the area of one cluster. 2) K-means clusterng: K-means s one well-known clusterng algorthm and ts basc steps are lsted below. (1) Intally, we select K devces whch are placed as far away as possble from each other and these K devces are consdered as centrods of the K clusters. (2) Then, another devce s selected and assocated to the cluster, the centrod of whch has the shortest dstance from the selected devce. (3) Calculate the mean coordnate of the new formed cluster and select the nearest devce to the mean coordnate as the new centrod. (4) Repeat step (2) and (3) untl all devces are assocated to a cluster. 3) Dstance based clusterng: Ths scheme s smlar to the K-means clusterng algorthm, wth followng steps. (1) K devces are randomly selected as centrods of K clusters. (2) Take one another devce and assocate t to the cluster whch has the shortest dstance from ts centrod to ths devce. (3) Repeat step (2) untl all devces are assocated to a cluster. It can be notced that, the centrods are selected randomly from the data set and these centrods are not updated durng BC (, j) EC (, j) < BL threshold. (3) The requrement of battery lfe s denoted by BL threshold. BC (, j) refers to the battery capacty of user-j n cluster- and EC (, j) s the energy consumpton by usng cellular lnk for a tme unt of t. t s the tme dfference between any two successvely TMS update commands for user-j. Wth a smaller value of t, the transmsson modes of UEs are updated wth a hgher frequency and network can respond to condton changes n a more tmely manner. On the other hand, a smaller value of t also ntroduces a hgher sgnalng load. Thus, n order to acheve a compromse n between the effcency and the sgnalng load, t can have a value rangng from several hours to several days, dependng on traffc models of UEs. Moreover, users whch cannot reach BS wth cellular lnks can be assumed to have an nfnte value of energy consumpton for t. Thus, these users also fulfll the nequalty n Eq. (3) and D2D communcaton s also appled to mprove ther avalablty. If some UEs are classfed as remote UEs n a cluster, BS checks whether some UEs n the same cluster fulfll the followng condtons for beng relays: BC (, j) EC (, j) > BL threshold, (4) SNR cellular (, j) SNR threshold. (5) Eq. (4) represents the condton that user- j n cluster- can meet the battery lfe requrement by usng cellular transmsson. In other words, ths user has enough battery capacty to serve as a relay for other remote UEs n cluster-. In Eq. (5), SNR cellular s the SNR value of the cellular lnk (, j) and SNR threshold s a threshold value to check whether the channel condton of the cellular lnk s good enough. For example, a value of 3 db s used n Sect. V as SNR threshold.
'() %& &'& #$ '$%%! " ' '% Fg. 3. Sgnalng scheme for D2D cluster formaton and TMS update Wth a hgher value for ths parameter, UEs wth better cellular lnk CSI are consdered as relay UEs and thus a hgher spectral effcency can be acheved for the relay lnk. However, a hgher value of SNR threshold means less feasble relay UEs and there s a hgher rsk that no relay UE exsts n a cluster. Once the BS obtans the lst of feasble relay UEs n one cluster, BS pcks up one relay UE and sends the D2D setup command to both the relay UE and remote UE(s). Upon recevng the D2D setup command, channel condtons between the relay and remote UEs are estmated to nspect f the D2D communcaton can contrbute to a better servce avalablty and energy effcency. If a D2D setup procedure s successful, the establshed D2D lnk s exploted for uplnk transmsson of packets from the remote UE. The correspondng sgnalng schemes are detaled n Sect. III. III. RADIO LINK ENABLERS In ths secton, we ntroduce the sgnalng schemes to support the proposed context-aware D2D communcaton. A. TMS and D2D cluster formaton In case ether a UE s ntally attached to the network or BS conducts the TMS update procedure, the sgnalng procedure shown n Fg. 3 wll be performed to confgure the correspondng UEs. Detals of ths sgnalng scheme are gven below, organzed by the steps shown n the fgure. 1) BS collects the context nformaton, e.g., locaton, battery level, traffc type and receved sgnal reference power (RSRP) of ts served MTC UEs. Besdes, the contextaware clusterng algorthm and TMS procedure are also performed. 2) BS confgures UEs wth D2D lnk system nformaton blocks (SIBs) for drect D2D dscovery, e.g. resource used for D2D dscovery. Moreover, UEs wll also be nformed by downlnk control nformaton (DCI) for the user specfc control nformaton, e.g. ID of the cluster to whch the UE belongs, transmsson mode of the UE, condtons whch D2D lnk should fulfll and so on. In case f a UE s confgured as ether a relay or a remote UE, nformaton of the other end of ths D2D lnk are also provded. 3) For relay UE, t send D2D dscovery announcement to the target remote UEs, wth ts own ID and IDs of the target UEs beng conveyed n the announcement message. In ths message, reference sgnals for D2D lnk channel estmaton are also carred. 4) Upon recevng the D2D dscovery announcement, the remote UE determnes whether the request s accepted or not, based on the estmated D2D lnk RSRPs. And a response message of acknowledgement/nonacknowledgement s sent back to the relay UE. 5) If the relay UE receves an D2D acknowledgement message, a securty assocaton could be establshed by exchangng messages between the relay and remote UEs wth securty algorthms. 6) The response of the D2D cluster formaton s further transmtted from the relay UE to the servng BS. In case the D2D request s not accepted, the servng BS should be nformed and thus t can avod to par the prevous selected relay UE wth ths remote UE n future. If a UE s confgured by BS to have cellular transmsson mode, t s not nvolved n the D2D dscovery procedure. It s also to be notced that, when the UE s ntally deployed by people, the locaton nformaton, traffc type and cellular lnk RSRP of one MTC UE mentoned n step 1) can be collected. Snce MTC UEs are assumed to be statc n ths work, these nformaton can be foreseen as unchanged. Besdes, the actual battery level of one UE can be transmtted to BS together wth the data transmsson n uplnk. Last but not least, f a D2D lnk s successfully establshed n step. 5), confguraton nformaton of ths D2D lnk should be stored at both the relay and remote UEs. B. D2D communcaton Once a D2D lnk s establshed accordng to Sect. III-A, the D2D lnk s used for uplnk packet transmsson of the remote UE. The correspondng sgnalng scheme s shown n Fg. 4 wth detals gven below. 1) Relay and remote UEs nvolved n D2D communcaton are confgured by the SIBs and DCI. These nformaton are confgured and stored at both relay and remote UEs when the D2D dscovery procedure s accepted. 2) In moble termnated case (e.g., UE s paged by network for ts report), one remote UE or a group of remote UEs wll be paged by BS for data report, or 3) In moble orgnated case (e.g., data s avalable n the buffer of the UE and wats for beng transmtted), a remote UE generates a data packet and tres to transmt ths packet to the relay UE. Ths step also ncludes the random access procedure, D2D lnk connecton setup procedure between the relay UE and remote UE, also D2D retransmsson f an error occurs n transmsson. In ths step, both D2D ends should be aware of the tme and frequency resource for D2D transmsson. 4) After successful recevng packets from remote UE(s), relay UE reples wth acknowledgment message(s) to the remote UE(s).
# & $! $!%$ '()!*! "! " where a Madrd grd model s appled [14]. The proposed envronment model algns well wth the realty to generate meanngful and precse results. In ths model, an urban envronment s depcted wth 3D vsualzaton where each grd composes of one park and 15 buldngs wth dfferent dmensons. The dmenson of one Madrd grd s 387 meters n west-east drecton and 552 meters n north-south drecton. In order to acheve a cell radus of 866 meters, multple replcas of Madrd grd are placed n the system level smulator. Moreover, buldng heghts n the Madrd grds are unformly dstrbuted between 8 and 15 floors wth a heght of 3.5 meters per floor. Fg. 4. D2D communcaton procedure Macro antenna 387 meters (west-east) Fg. 5. Envronment and deployment model 5) Relay UE further forwards the successfully receved packets to the servng BS. Ths process can be performed as a normal cellular uplnk transmsson where a control plane (CP) connecton needs to be establshed. Another alternatve s that the receved packets wll be buffered n the relay UE and then transmtted together wth ts own packet. In ths case, an advantage n power savng for relay UE can be ntroduced, snce the relay UE only needs to wake up and perform the CP connecton establshment procedure for once. 6) Upon successfully recevng packets from the relay UE, BS sends an acknowledgment message to the relay UE. IV. EVALUATION METHODOLOGY In order to evaluate the proposed technology, a system level smulator s mplemented n ths work and algned tghtly wth the real world. In ths secton, models used n our smulator are stated wth detals. Please note that, only the dfference compared wth ITU-R performance evaluaton document [13] are gven here. For other parameters not mentoned here, they are algned wth the ITU-R document [13]. A. Envronment model In ths work, system performance of the MTC servces s nvestgated n dense urban envronment as shown n Fg. 5, 552 meters (north-south) B. Deployment Scenaro A sngle macro BS wth a cell radus of 886 meters s deployed n the Madrd grds, n order to acheve an nter ste dstance (ISD) of 1732m as defned n 3GPP [4] [15]. The poston of macro antennas s also plotted n Fg. 5. For the macro staton, t operates n three cell sectors wth a carrer frequency of 900 MHz and drectonal antennas are postoned wth 120 degree dfference from each other n the horzontal plane. C. User deployment and traffc model In the coverage of the BS, 20 thousand sensor devces are randomly dstrbuted nsde buldngs and are assumed to be statc. An sotropc antenna s nstalled on each devce at 1.5 meter heght wth a maxmal transmsson power of 23 dbm. Moreover, a report packet of 2000 bts wth a perodcty of 24 reports per day s exploted n ths work as traffc model for sensor devces. D. Channel model A 3D channel model proposed by 3GPP [4] s appled here, n whch the penetraton loss through buldng floors and walls s taken nto account. To characterze the channel n between the two ends of one D2D communcaton, channel models proposed n [16] are appled. In [16], channel characters are captured n three dfferent scenaros for ndoor UEs,.e., two D2D ends are on the same floor n the same buldng; two D2D ends are on the dfferent floors n the same buldng; two D2D ends are n dfferent buldngs. E. User devce power consumpton model In order to evaluaton the power consumpton of MTC devces, power consumpton related parameters [4] [17] are lsted n Tab. I. Please notce that, a new UE state called connected-nactve state s proposed to serve for 5G [18], and thus the duraton of control plane establshment s calculated based on ths new technology. V. NUMERICAL RESULTS In ths work, LTE technology s used n our work for modelng rado lnks and the mappng between SNR value and lnk capacty s performed by explotng the results provded
TABLE I DEVICE POWER CONSUMPTION PARAMETERS Parameter Descrpton Value Tme duraton f applcable P tx transmsson power 45% PA effcency plus 60 MCS and packet mw/s for other crcutry sze related P rx power to receve packets from remote UEs 100 mw/s MCS and packet sze related P pagng power to receve pagng command 100 mw/s 10 ms P clock clock to obtan synchronzaton 100 mw/s 10 ms P cp power consumpton durng thecontrol plane establshment procedure 200 mw/s 10 ms P sleep power consumpton n sleepng mode 0.01 mw/s D rx UE wakes up to lsten to pagng 4 tmes/day C battery capacty 6500 J tme of UE stayng n sleepng mode n [19]. Thus, f a rado lnk experences a SNR value lower than -7 db, no data transmsson s possble on ths lnk. To reflect ths aspect, avalablty s consdered as the metrc to be evaluated and ts mathematcal defnton s gven as: avalablty = number of UEs can be served. (6) total number of UEs In other words, avalablty shows the rato of users that can upload ther packets to BS. Addtonally, battery lfe of each UE s also nspected to calculate the rato of users who can meet the battery lfe requrement of MTC servces. Here, the target battery lfe s set to be ten years. In the left hand of Fg. 6, the cumulatve dstrbuton functon (CDF) of battery lfe of UEs s plotted, w.r.t. dfferent clusterng algorthms, as gven n Sect. II. As a baselne scheme, the performance of LTE system s also drawn. As an nput for geometrcal clusterng method, the area of a cluster s 40000 square meters. As an output from geometrcal clusterng method, number of clusters s further fed to other three clusterng algorthms, n order to acheve a far comparson among dfferent algorthms. Moreover, a D2D pathloss value lower than 136dB s set to be the crteron for D2D connecton setup, as stated n Sect. III-A. Part of the CDF plot s zoomed n and shown at the rght hand sde of ths fgure, snce t s the most nterestng part for our nspecton. As t can be seen from the fgure, 14% of MTC UEs can not transmt uplnk reports to BS by usng LTE technology, whle only 2% of UEs can not connect to BS f D2D communcaton s exploted. Thus, the avalablty s mproved from 86% to 98%. Moreover, 80.5% of UEs can meet the battery lfe requrement of ten years (3650 days) n LTE system whle ths value can be mproved to 90% by explotng D2D communcaton. In Fg. 7, the same settngs as n Fg. 6 are appled, except that 2500 square meters s consdered as the coverage area of one cluster for geometrcal clusterng method. Due to ths fact, a larger number of clusters exst n the same coverage area of the BS and thus each cluster comprses a smaller number of UEs. In ths sense, there s a lower probablty of havng feasble relay UE n each cluster. Ths s the reason why the avalablty values of three clusterng algorthms are decreased to approxmately 96%, compared wth the values shown n Fg. 6. As an excepton, the avalablty of the Dstance+CSI based clusterng algorthm s mproved to 99%. Ths s due to the fact that UEs wth good channel condtons are selected as centrods n the ntal step of the clusterng algorthm. Moreover, by havng more clusters, the coverage area of each cluster s reduced and the pathloss value between two D2D ends s smaller. Thus, the battery lves of UEs can be mproved for all the appled D2D clusterng schemes. For the Dstance+CSI based clusterng algorthm, up to 95% of UEs can meet the battery lfe requrement of ten years. VI. CONCLUSION As shown n ths paper, a context-aware D2D communcaton can be appled to enhance the avalablty and mprove devce power consumpton performance for MTC applcatons. Sgnalng schemes for D2D cluster formaton and D2D communcaton are also provded to support the proposed context-aware D2D formaton and TMS scheme. At the same tme, the desgned sgnalng schemes have also the advantages of less extra sgnalng overload compared wth other schemes proposed n the lterature. Moreover, the proposed concept s evaluated by a system level smulator and a large performance gan can be obtaned by explotng the proposed D2D communcaton. ACKNOWLEDGMENT Part of ths work has been performed n the framework of H2020 project METIS-II, whch s funded by the European Unon. The vews expressed are those of the authors and do not necessarly represent the project. The consortum s not lable for any use that may be made of any of the nformaton contaned theren. REFERENCES [1] Len, S.Y.; Chen, K.C.; Ln,Y. Toward ubqutous massve accesses n 3GPP machne-to-machne communcatons, Communcaton Magazne. IEEE 2011, 49, 66-74. [2] 3GPP Techncal Report 36.888. Study on provson of low-cost Machne- Type Communcatons (MTC) User Equpments (UEs) based on LTE (Release 12), June, 2013.
Fg. 6. system performance w.r.t devce power consumpton, Asector = 40000m2 Fg. 7. system performance w.r.t devce power consumpton, Asector = 2500m2 [3] Lang, J.M.; Chen J.J.; Cheng, H.H.; Tseng, Y.C. An Energy-Effcent Sleep Schedulng Wth QoS Consderatonn 3GPP LTE-Advanced Networks for Internet of Thngs, Emergng and Selected Topcs n Crcuts and Systems, IEEE Journal on 2013, 3, 13-22. [4] 3GPP Techncal Report 45.820. Cellular system support for ultra-low complexty and low throughput Internet of Thngs (CIoT) (Release 13), November, 2015. [5] Farhan Ahmad, Safdar Nawaz Khan Marwat, Yasr Zak, Yasr Mehmood, Carmelta Goerg. Machne-to-machne Sensor Data Multplexng usng LTE-Advanced Relay Node for Logstcs, https://fles.nyu.edu/yz48/publc/ Fnal_LDIC_Farhan.pdf [6] J Langha, A. Klen, N. Kuruvatt, H. D. Schotten, System Capacty Optmzaton Algorthm for D2D Underlay Operaton, n Proceedngs of Workshop on 5G Technologes at IEEE Internatonal Conference on Communcatons (ICC), Sydney, Australa, June 2014. [7] J Langha, A. Klen, N. Kuruvatt, R. Sattraju, H. D. Schotten, Dynamc Context-aware Optmzaton of D2D Communcatons, n Proceedngs of 2nd Internatonal Workshop on 5G Moble and Wreless Communcaton System for 2020 and Beyond at IEEE 79th Vehcular Technology Conference (VTC-Sprng), Seoul, Republc of Korea, May 2014 [8] M. J, G. Care and A. F. Molsch. Wreless Devce-to-Devce Cachng Networks: Basc Prncples and System Performance, n IEEE Journal on Selected Areas n Communcatons, vol. 34, no. 1, pp. 176-189, Jan. 2016. [9] N.K. Pratas and P. Popovsk. Underlay of low-rate machne-type D2D lnks on downlnk cellular lnks, 2014 IEEE Internatonal Conference on Communcatons Workshops (ICC), Sydney, NSW, 2014, pp. 423-428. [10] N. K. Pratas and P. Popovsk. Low-Rate Machne-Type Communcaton va Wreless Devce-to-Devce (D2D) Lnks, arxv preprnt arxv:1305.6783. [11] 3GPP meetng, RP-151948. New WI Proposal: D2D based MTC, December 2015. [12] Orsno, A., Arant, G., Mltano, L., Alonso-Zarate, J., Molnaro, A., Iera, A. (2016). Energy Effcent IoT Data Collecton n Smart Ctes Explotng D2D Communcatons, Sensors (Basel, Swtzerland), 16(6), 836. http://do.org/10.3390/s16060836 [13] ITU-R M.2135. Gudelnes for evaluaton of rado nterface technologes for IMT-Advanced, 2008. [14] METIS, Delverable 6.1, Smulaton Gudelnes, October, 2013. [15] 3GPP Techncal Report 38.913. Study on Scenaros and Requrements for Next Generaton Access Technologes, October, 2016. [16] 3GPP, R1-132030. Channel models for D2D performance evaluaton, May, 2013. [17] Tuomas Trronen, Anna Larmo, Joachm Sachs, Bengt Lndoff and Nclas Wberg. Machne-to-machne communcaton wth long term evoluton wth reduced devce energy consumpton, Trans. Emergng Tel. Tech., 2013. [18] METIS-II, Delverable 6.1, Draft asynchronous control functons and overall control plane desgn, June, 2016. [19] Jar, M. and Fettwes, G. Throughput Maxmzaton for LTE Uplnk va Resource Allocaton, Internatonal Symposum on Wreless Communcaton Systems (ISWCS), 2012.