Improving Device-to-Device Communication Performance in 5G Networks through Joint Power Control and Resource Allocation

Transcription

1 Internatonal Journal of Communcatons htt:// Imrovng Devce-to-Devce Communcaton Performance n 5G Networks through Jont Power Control and Resource Allocaton HWANG-CHENG WANG 1, KUN-KAI CHEN, FANG-CHANG KUO 1, CHIH-CHENG TSENG 1 and JENNIFER POUNJEBA Natonal I-Lan Unversty 26047, Tawan 1 {hcwang, kfc, tsengcc}@nu.edu.tw {r , r }@ems.nu.edu.tw Abstract: - The rad growth n user oulaton and the rolferaton of dverse systems and servces lace a heavy burden on moble communcaton networks. Devce-to-Devce (D2D) communcaton technology s consdered one of the key technologes n 5G because of ts ablty to allevate ths burden and mrove system erformance. Most current research focuses on how to enhance transmsson effcency but roosed solutons tend to be hghly comlex. Therefore, the base staton works for a long duraton on ths rocess, ncreasng the cost of communcaton equment. To tackle ths ssue, we roose a low-comlexty method that makes the use of ower control and sutable resource allocaton. A centralzed network archtecture s adoted to smlfy ths rocess and mrove erformance. Smulaton results show that the roosed aroach can sgnfcantly decrease nterference, reduce ower consumton, and ncrease throughut. Key-Words: - 5G, D2D communcaton, ower control, resource allocaton, RB reuse, decrease nterference 1 Introducton Devce-to Devce (D2D) communcaton technology s the latest nnovaton that can reduce the loads the base statons. It allows two devces wthn a certan dstance to communcate drectly among themselves wthout the hel of evolved Node B (enb) hardware. D2D communcaton utlzes the cellular sectrum (lcense band) and s aded by cellular network nfrastructure [1]. Current network archtecture s dvded nto two tyes. Frst s centralzed management knows as, the Resource Block (RB), whch s allocated drectly from the base staton to the desgnated D2D user. Ths archtecture rovdes a better allocaton of resources and s often used n densely oulated areas. Second s decentralzed management, n whch the base staton nforms D2D users whch RBs they can use, and D2D users then choose RBs. Ths archtecture s more lkely to cause collsons unlke the frst archtecture, thus ts effcency s relatvely oor. Hence, ths archtecture can be used n suburbs and other areas where there are fewer Cellular User Equment (CUEs). In ths aer, we consder a centralzed management envronment, where all oeratons are erformed through enb (Evolved Node B) because of ts hgh comutatonal ablty. By usng ower control [2][5], we can reduce the nterference caused by the user, and mrove resource utlzaton by reusng RBs that are used by CUEs [6]. We can also mrove the system s transmsson effcency by allocatng resources to the D2D ar accordng to ther requests. 1.1 Related Research In [7], the authors resented several new challenges for the D2D related technology. One of the bggest challenges s how Devce User Equment (DUE) determnes the resence of other DUE. D2D s a technology that allow devces to communcate drectly wth the one another, thus devces need to ar before communcatng. There are two man methods for D2D arng. The frst method s searchng for nearby devces. Fndng more devces n less tme s more conducve to establsh a connecton for transmsson. Another method s to drectly secfyng the devce to be ared wth. Ths tycally occurs after the revous search has been done, and the devce then secfes the other devce and successfully ars. Ths aer s based on the remse that devce exloraton has been comleted and that all the D2D ars are ared already. ISSN: Volume 2, 2017

2 Internatonal Journal of Communcatons htt:// Hwang-Cheng Wang et al. The next challenge s that of resource allocaton. There s an exclusve resource allocaton method, n whch resources are gven drectly to the devces. In ths case, there s no nterference between the D2D ars and and D2D ars are evenly dstrbuted n the coverage area. D2D ars are at least 35 meters away from the base staton [12]. In ths context, D2D ars want to communcate drectly. CUEs, but the sectrum effcency wll be lower. Hence, D2D ars and CUEs can use the same resources to ncrease sectrum effcency,.e., D2D devces can reuse the resources used by the CUEs, thus ncreasng the sectrum effcency [13]. In [8], a D2D resource allocaton method s roosed where nterference between the devces because of reduced reuse of RBs. However, ths s a one-to-one mang roblem, so wreless resources can only be reused once. Another study [9] suggests assgnng several users to a grou/coalton to reduce the comlexty of rado resource allocaton. Ths coalton formng method can sgnfcantly reduce nterference comared to allocatng rado resources to the users one by one. In [10], the authors used centralzed and decentralzed ower control algorthms to mrove the erformance of D2D communcaton n a sngle-cell random network model. In the centralzed ower control algorthm, D2D ars can communcate wth a secfc transmt ower that can maxmze coverage. On the other hand, the decentralzed ower control algorthm used an on-off ower control to manage nterference between D2D ars. Fg. 1 System Model. 2.1 System Descrton and Assumtons A detaled descrton of the envronment consdered n ths aer are as follows: 1. D2D ars only reuse CUEs ustream resource. 2. All D2D ars are already ared. 3. The enb knows the coordnates of all users. 4. There are N D2D ars and N c CUEs. Hence n ths aer, we roose a ower control and resource management scheme that would hel us Reduce nterference caused by other DUE ars. Reduce ower consumton of D2D ars. Reduce resource requrements. Imrove the average sgnal-to-nterferencelus-nose rato (SINR) of each D2D ar. 5. N RB, req s the number of RB requred for D2D ar. The CUE has comleted ts ower control and the number of RBs beng used s a unform random varable between 1 and 10. We have made some assumtons to smlfy the smulaton and analyss, whch are as follows: 1. D2D ars can send the requred number of RBs through the Physcal Random-Access Channel (PRACH). 2. D2D ars reuse the CUE s ulnk sectrum. 3. enb allocates ulnk resources to the CUEs based on sem-ersstent schedulng. 4. Interference occurrng outsde of the nterferental coverage can be gnored. 5. The channel qualty of all RBs s the same for a D2D ar. 6. The number of RBs requred for a D2D ar after ower control s only related to SINR. The rest of ths aer s organzed as follows. In Secton 2 we ntroduce the system model consdered n ths aer. Secton 3 exlans ower control and resource management methods. Secton 4 contans smulaton results and analyss. Secton 5 concludes the aer. 2 System Model The system model an outdoor urban macrocell model wth only one base staton s shown n Fg.1. The coverage of the base staton s 500 meters and CUEs ISSN: Volume 2, 2017

3 Internatonal Journal of Communcatons htt:// 2.2 Symbol Defnton Ths aer wll use some symbols to facltate our descrton, the followng symbols used to defne: 1. N : The number of D2D ars n the envronment. 2. N c : The number of CUEs n the envronment. 3. PL : Path loss of a D2D ar, between two DUEs. RB, req 4. N : The number of RBs requred for a D2D ar 5. P 0 : The transmt ower of a D2D ar n envronment. 6. P : The transmt ower of D2D ar after ower control. 2.3 Overvew of Proosed Method In ths aer, we use ower control method and resource allocaton to mrove the overall erformance of the D2D network [14]. For ower control, we assume that nterference outsde of the nterferental coverage s neglgble. If there s another D2D ar wthn the nterferental coverage area of a D2D ar, that ar should erform ower control at ts maxmum, so that we can enhance the chances to of allowng the D2D ars to share RBs wth other D2D ars durng RB allocaton. Otherwse, the D2D ar need not erform ower control, but nstead use the default transmt ower of 23 dbm. Because the default transmt ower s 23 dbm, the maxmum nterferental coverage dstance of DUE s 153 m. To smlfy, we choose 150m as the nterferental coverage area of DUE before ower control. Durng resource allocaton, users can reuse a CUE s RB rovded they do not nterfere wth each other. Ths wll mrove the utlty of resources, and can reduce the total number of requred RBs. The stes of our roosed method are as follows: 1. enb calculates the shortest dstance d, from D2D ar to D2D ar where = 1, 2, 3, N,. 2. Calculate the nterference value of each D2D n accordance wth other D2D ars. 3. Start ower control wth the D2D ar wth the hghest nterference value, and then recalculate the nterference of each D2D ar caused by the resence of other D2D ars. Reeat the above stes untl all D2D ars are checked at least once. 4. Durng RB allocaton, D2D ars that stll have nterference wth other ars are gven dfferent RBs to communcate. 3 Power Control The followng s a detaled descrton of the ower control method. 3.1 Calculaton of the Mnmum Power Before carryng out ower control, we must defne the dstance between the DUEs n a D2D ar. As shown n Fg. 2, d and, d, are the dstances between the transmtters and recevers of the D2D ar and D2D ar, resectvely. d, s the shortest dstance between D2D ar and D2D ar. Fg. 2 The dstance between D2D ar. After erformng ower control, the mnmum ower that allows D2D ar to stll communcate s denoted by P and calculated as mn mn D P ( SINRmn 10log I (mw) (mw) ) G (1) where SINR mn s -6.7 db. Ths s the lowest SINR value at whch the D2D ar can stll communcate [12] I s the nterference on the D2D ar caused by the other D2D ars. The formula for the nterference s I N ( P0 PL, )/10 E (2) (mw) 10 1, ISSN: Volume 2, 2017

4 Internatonal Journal of Communcatons htt:// where E s an ndcator functon. Because we assume that nterference outsde of the nterferental coverage area can be gnored, we use ths functon to determne whether nterference exsts. Thus E can be exressed as: E 1, d, 150 m 0, d, 150 m (3) nterference to D2D ar after erformng ower control. Ths s ndcated by the sold arc n Fg. 3. The value of s calculated as follows: d, (db) 40log( ) (7) d, s thermal nose. Wth a bandwdth of 20 MHz, = D -101 dbm. G s the channel gan of D2D ar, whch s the recrocal of the ath loss of D2D ar. mn P s the mnmum transmt ower requred for D2D ar to communcate, whch can be used to check whether the transmt ower generated after ower control s suffcent. The ath loss model between D2D ar and D2D ar s: PL, (db) Max[20log( d, ) 38.4, 22.7log( d, ) 33.02], d, m Max[20log( d, ) 38.4, 40log( d, )+11.73], d, m (4) Fg. 3. Calculaton of the transmt ower for D2D ar. Ths formula s modfed by formula (4) because the value s dfferent only when the dstance s greater than or equal to 22 meters. It s also ossble to elmnate the mn constant term because RSS s ncluded. Therefore, we choose the maxmum value,and hence the formula wll be as shown n (5). 3.2 Calculaton of Transmt Power To calculate the transmt ower after erformng ower control, we use Eq. (5) to calculate the mnmum mn Receved Sgnal Strength (RSS), denoted by RSS, mn for D2D ar and use RSS to calculate the maxmum transmt ower that can be used after ower control. SINR mn, I and are the same as those used n mn P. mn RSS SINR mn 10log[ I(mW) (mw)] (5) 3.3 Resource Allocaton Next, we must calculate the transmt ower after erformng ower control as, P RSS (6) mn where P s the requred transmt ower of D2D ar mn whch s the sum of RSS and. As shown n Fg.3, mn RSS s the mnmum RSS needed to ensure when the two devces n a ar when searated by a dstance d, can communcate wth each other. Ths s, ndcated by the dotted arc n the fg.3. The urose of addng s to allow the transmt ower of D2D ar to reach the maxmum value wthout causng Fg. 4 Scenaro dectng the reuse of RBs. For resource allocaton, we must fnd how the RBs can be reused, takng nterference nto account. As llustrated n Fg. 4, there are three D2D ars and two CUEs. D2D ar 1 s n the nterferental coverage area of CUE1, hence D2D ar 1 can only share RBs wth CUE2. Smlarly, D2D ar 2 s n the nterferental ISSN: Volume 2, 2017

5 Internatonal Journal of Communcatons htt:// coverage area of CUE2 and can only share RBs wth CUE1. However, D2D ar 3 s not n the nterferental coverage area of ether CUE1 or CUE2 therefore, t can share RBs wth both CUEs. 3.4 Greedy Algorthm 4.1 Smulaton Setu Table 2 Smulaton arameters. The throughut obtaned usng ower control was comared wth the outut obtaned usng a greedy algorthm whch s as follows: 1. A D2D ar and CUE share the same ulnk RB. Also consderng the ower control technque, we assume that each CUE can share only one RB wth a D2D ar. 2. The enb allocates the RB to the D2D ar based on ts rorty. 3. Greedy algorthm s used for resource allocaton and to cancel the nterferences [15]. The throughut obtaned usng a greedy algorthm s result of greedy result wth ower control - result of greedy 100% 4 Smulaton Results and Analyss (8) Table 2 summarzes the arameters used n the smulaton. There are 20 D2D ars and 10 CUEs dstrbuted evenly n the 500-meter coverage. DUEs must be at least 35 meters away from the base staton [12]. In ths scenaro, we can erform D2D communcaton and verfy erformance wth our roosed ower control and resource allocaton methods. Table 1 Examle of CUE arameters. 4.2 Smulaton Results Table 1 shows an examle of the key arameters of 10 CUEs. The frst column reresents the number (ID) of each CUE. The second column s the number of RBs used by each CUE allocated by enb. The last column s the nterferental coverage area of each CUE. enb wll frst determne whether a D2D ar s n the nterferental coverage range of a CUE, and then select the CUE wth the largest number of RBs to share RBs wth the D2D ar. Fg. 5 Total ower consumton of D2D ars before and after ower control. We smulate the scenaro u to 100 tmes, consderng the dstrbuton of DUEs and CUEs. The number of D2D ars nsde the enb coverage starts at 20 ars and ncreases to 25 ars, 30 ars, and 35 ars. The results are obtaned usng our roosed ower and resource allocaton methods and are comared wth results obtaned wthout ower control. The ISSN: Volume 2, 2017

6 Internatonal Journal of Communcatons htt:// comaratve results are shown n a bar chart. We also analysed the cause of erformance mrovement. At N = 20, the total ower consumton before ower control was about 4000 mw, and after ower control was done, the average total ower consumton droed to about 1173 mw, gvng a ower savngs rato of 70.6%. At N = 25, the total ower consumton before erformng ower control ncreased wth the number of D2D ars, reachng to about 4988 mw, and after erformng ower control, the average total ower consumton droed to about 1192 mw, yeldng a 76.09% ower savng rato. At N = 30, the total ower consumton before erformng ower control was about 5986 mw, and the average total ower consumton droed to about 1120 mw after ower control, for a ower savngs rato of 81.29%. At N = 35, the total ower consumton before erformng ower control was about 6983 mw,and ower control droed the average total ower consumton to 1238 mw, gvng a ower savngs rato of 82.27%. As N ncreased, so dd the total ower consumton before erformng ower control. But after carryng out ower control, total ower consumton decreased, thus the roorton of ower savngs wll ncrease. From ths result, we can see that when the system has many D2D ars, we can save more ower through our roosed method. was 161 before ower control and 28 after ower control leadng to an 85.71% reducton n requred RBs. Smlarly, the average number of RBs for N = 30, was 172 before ower control and 25 after ower control, yeldng a RB reducton rato of 85.46%. At N = 35, the average number of RBs requred before ower control was 201, and the average number of RBs requred after ower control was 33, gvng a reducton rato of 83.58%. As a result, as the number of N ncreases, D2D ars usng the ower control scheme also ncrease, leadng to mutual nterference between the D2D ar. Hence, the number of D2D ars that reuse the RBs wll also decrease. As a result, the ratos of RB demand when N = 30 and N = 35 are lower than that of s stll mressve. N = 25. Nonetheless, the reducton Fg. 7 Average SINR of each D2D ar before and after ower control. Fg. 6 Average total number of RBs requred by D2D ars before and after ower control. Fgure. 6 shows the average total number of RBs requred by D2D ars before and after erformng ower control. For N = 20, the total number of requred RBs before ower control was 109 and the total number of RBs after ower control was 23. Therefore, the rato of reducton n requred RBs was 78.89%. the total number of requred RBs for N = 25, Fgure 7 shows the average SINR of each D2D ar before and after erformng ower control. At N = 20, the average SINR of each D2D ar before and after ower control s done was 20 db and 33 db, resectvely, a reducton of u to 13 db. At N = 25, the nterference ncreased wth N, resultng n the average SINR of each D2D ar fallng u to 18 db. After erformng ower control, the transmsson ower of the D2D ars also droed. But because of the hyothess of ths aer, nterference out of nterferental coverage area s gnorable. Thus, the nterference s drastcally reduced or even elmnated. In ths case, the average SINR for each ar of D2D ar was greatly ncreased, reachng about 40 db, an ncrease of about 22 db. At N = 30, the average ISSN: Volume 2, 2017

7 Internatonal Journal of Communcatons htt:// SINR was about 12 db before erformng ower control and 41 db afterward, an ncrease of u to 28 db. At N = 35, the average SINR before ower control was about 11 db, and t ncreased u to 52 db after ower control, resultng n a 41 db dfference. After comarson, we can fnd that the nterference of each D2D ar s obvously reduced. When N = 20, the throughut obtaned by drectly allocatng RBs wthout ower control s about 0.44 Gbs, and the result obtaned erformng usng ower control s about 0.78 Gbs. After ower control, the average SINR of the overall D2D ar can be mroved, whch makes the communcaton more sgnfcant. When the number of N s a hgher number, the lft rato wll be more ronounced. In Fg. 8 we comare, the throughut of three dfferent algorthms. Only lmted number of RBs were reused. But after erformng ower control, most of the D2D ars wll not nterfere wth each other and are more lkely to share ther RBs. Usng the roosed ower control method allows us to acheve a more than 400% ncrease n throughut comared to a greedy algorthm. Fg. 8 Throughut of three dfferent algorthms. 5 Concluson In ths aer, we reduce the nterference between D2D ars by means of a roosed roer ower control scheme and by allocatng allocate resources, accordngly so that the D2D ars do not nterference wth each other. Smulaton results show that erformance has been mroved sgnfcantly comared to D2D communcaton when no ower control was emloyed. Secfcally, the roosed method ontly consders ower control and resource allocaton and exhbts sueror erformance n ower consumton, SINR, throughut, and number of RBs requred to meet the demand of D2D ars. Hence through the formaton of coaltons and ower control, the average SINR of each D2D ar s sgnfcantly ncreased. The rato of reuse of RBs was also hgh. Future research wll focus on mrovng the SINR of D2D ars by consderng nterference outsde the nterference coverage area to acheve better erformance. Comarsons wth other methods such as greedy resource allocaton wll also be nvestgated. References: [1] Shahd Mumtaz, Kaz Mohammed Sadul Huq and Jonathan Rodrguez, Drect Moble-to-Moble Communcaton: Paradgm for 5G, IEEE Wreless Communcatons, October [2] htt:// %8E%87%E6%8E%A7%E5%88%B6 [3] 3GPP, Evolved Unversal Terrestral Rado Access (E-UTRA); User Equment (UE) rado transmsson and receton, Release 13, TS V13.5.0, [4] 3GPP, Evolved Unversal Terrestral Rado Access (E-UTRA); Physcal layer rocedures, Release 13, TS V13.3.0, [5] 3GPP, Evolved Unversal Terrestral Rado Access (E-UTRA) and Evolved Unversal Terrestral Rado Access Network (E-UTRAN); Overall descrton; Stage 2, Release 13, TS V13.5.0, [6] D. Feng, L. Lu, Y. Y. Wu, G. L, S. L, and G. Feng, Devce-to-devce communcatons n cellular networks, IEEE Communcatons Magazne, vol. 52, ssue 4, , [7] D. Panatool, C. Mouton, B. Lecroart, Y. Lar, and P. Delahaye, Recent advances n 3GPP Rel-12 standardzaton related to D2D and ublc safety communcatons, arxv rernt arxv: , [8] N. Chen, H. Tan, and Z. Wang, Resource allocaton for ntra-cluster D2D communcatons based on Kuhn-Munkres algorthm, IEEE Vehcular Technology Conference,. 1 5, [9] J.-Y. Shh and C.-C. Tseng, Resource Allocaton for D2D Communcatons Usng Two-stage Coalton Formaton and Nash Barganng Soluton n LTE Networks, Natonal Ilan Unversty, Deartment of Electrcal Engneerng,2016. [10] A. Memm, Z. Rezk, and M. Aloun, Power Control for D2D Underlay Cellular Networks wth ISSN: Volume 2, 2017

8 Internatonal Journal of Communcatons htt:// Imerfect CSI, IEEE Globecom Workshos (GC Wkshs),. 4 8, [11] 3GPP, 3rd Generaton Partnersh Proect; Techncal Secfcaton Grou Rado Access Network; Study on LTE devce to devce roxmty servces; Rado asects (Release 12), TR V12.0.0, Mar [12] A. Ghosh and R. Raeeat, Essentals of LTE and LTE-A, Cambrdge Unversty Press, [13] S-An Cou, Jung-Chun Kao, Chung Y Lee, and Kuo-Y Chen, Mult-Sharng Resource Allocaton for Devce-to-Devce Communcaton Underlayng 5G Moble networks, 2015 IEEE 25th Internatonal Symosum on Personal, Indoor and Moble Rado Communcatons - (PIMRC): Moble and Wreless Networks [14] Mukheree M., Shu L., Zhang Y., Zhou Z., Wang K. (2015) Jont Power and Reduced Sectral Leakage-Based Resource Allocaton for D2D Communcatons n 5G. In: Wang G., Zomaya A., Martnez G., L K. (eds) Algorthms and Archtectures for Parallel Processng. Lecture Notes n Comuter Scence, vol Srnger, Cham [15] Yu Tao, Jun Sun, Shxang Shao, Rado Resource Allocaton Based on Greedy Algorthm and Successve Interference Cancellaton n Devce- To-Devce (D2D) Communcaton, IET Internatonal Conference on Informaton and Communcaton technologes (IETICT 2013) ISSN: Volume 2, 2017