Uplinks Analysis and Optimization of Hybrid Vehicular Networks

Transcription

1 Uplinks Anlysis n Optimiztion of Hybri Vehiulr Networks Shikun Li 1, Zipeng Li 1, Xiohu Ge 1, Yonghui Li 2 1 Shool of Eletroni Informtion n Communitions, Huzhong University of Siene n Tehnology Wuhn, Hubei Chin [e-mil: li_shikun, xhge, zipengli91@mil.hust.eu.n] 2 Shool of Eletril n Informtion Engineering The University of Syney, Syney [e-mil: yonghui.li@syney.eu.u] *Corresponing uthor: Xiohu Ge Abstrt 5G vehiulr ommunition is one of key enblers in next genertion intelligent trnsporttion system (ITS), tht require ultr-relible n low lteny ommunition (URLLC). To meet this requirement, new hybri vehiulr network struture whih supports both entrlize network struture n istribute struture is propose in this pper. Bse on the propose network struture, new vehiulr network utility moel onsiering the lteny n relibility in vehiulr networks is evelope bse on Eulien norm theory. Builing on the Preto improvement theory in eonomis, vehiulr network uplink optimiztion lgorithm is propose to optimize the uplink utility of vehiles on the ros. Simultion results show tht the propose sheme n signifintly improve the uplink vehiulr network utility in vehiulr networks to meet the URLLC requirements. Keywors: Hybri vehiulr network, Ultr-relible n low lteny ommunition (URLLC), Vehiulr network utility moel, Preto optimiztion

2 1. Introution In reent yers, the intelligent trnsporttion system hs beome reserh hotspot whih is one of the most promising pplition senrios in the fifth genertion (5G) mobile ommunition systems [1-2]. Different from the tritionl mobile ommunition networks, the ritil informtion generte by the vehiles nee to be trnsmitte with very low lteny n extremely high relibility [3-4]. It is iffiult for tritionl vehiulr networks to relize ultr-relible n low lteny ommunition (URLLC) [5]. How to esign ynmil vehiulr network to hieve URLLC presents gret hllenge. Currently, there re minly two types of network strutures use to relize vehiulr networks in the existing stuies, entrlize network struture bse on ellulr networks [6-8] n istribute vehiulr network struture, lso known s the vehiulr -ho network (VANET) [9-11]. The vntge of LTE bse entrlize vehiulr networks ws isusse in [6]. In [7], TD-LTE bse vehiulr network struture ws introue n the trnsmission relibility uner ifferent trnsmission istnes ws stuie. The nlysis results showe tht LTE is promising tehnology to support the long rnge vehile to infrstruture (V2I) ommunitions. However, the high lteny remins s key issue in LTE bse vehiulr networks. Sine the low trnsmission relibility of messges in vehiulr networks le to the lrge resen time whih oes not gurntee the short hnling time for seurity messges when there is lrge istne between vehile n ro sie unit. In ft, s one of the most importnt performne riteri in vehiulr networks, the lteny of trnsmitting ritil informtion, suh s the sfety-relte messges is one of key metris to evlute performne of vehiulr network [8]. In istribute VANET, vehiles on the ro n ommunite with eh other vi vehile to vehile (V2V) links n vehile lusters re forme to provie flexible network struture. In [9], n ess relibility moel in VANET ws propose to nlyze the vehiulr ess relibility with fixe lteny onstrint. Connetivity probbility uner ifferent trffi ensity for V2V ommunition senrios in one- n two-wy pltoon-bse VANETs ws investigte in [10]. In [11], one-imensionl multi-hop brosting moel onsiering the ifferent sfe istnes between vehiles ws propose to nlyze the brosting suess probbility n multi-hop lteny in VANETs. However, either entrlize or istribute vehiulr network struture nnot meet the URLLC requirements of future vehiulr networks n n only stisfy prts of the vehiulr network requirements [12]. To tkle this problem, in this pper, hybri vehiulr network moel whih n just network struture ynmilly is propose to hieve URLLC, s Preto improvement strtegy is eploye in the hybri vehiulr network whih n le to n optiml vehiulr network utility uner ifferent istnes between vehiles n the ssoite RSU. The ontributions of this pper re summrize s follows. 1) To nlyze the lteny n relibility of messges in vehiulr networks, novel vehiulr network utility moel bse on Eulien norm theory is propose for the hybri vehiulr networks. 2) By using the Preto optimiztion, n uplink optimiztion lgorithm is propose to improve the vehiulr network utility of uplinks in the hybri vehiulr networks. 3) Simultion results show tht the propose uplink optimiztion lgorithm n signifintly improve the uplink relibility n reue the uplink lteny. The vehiulr network utility of the Preto improvement strtegy is improve up to 15%

3 ompre with rnom rely strtegy n ellulr ess strtegy in hybri vehiulr networks. 2. System Moel Zoom in V in V g V out Inp RSU Centrl re Vehile luster Vehile R2I link V2-ho link V2R link Fig. 1. System moel We onsier typil urbn vehiulr network shown in Fig. 1. In this network, ro sie units (RSUs) re istribute uniformly long the ros. These RSUs n provie vehiulr network ess servies for vehiles through the vehile to RSU (V2R) wireless uplinks. Due to the limite omputtion pbility of RSUs, in most pplition senrios, RSUs re only responsible for the messge forwring, n the omplex lultion servie is hr to be one in RSUs. Therefore, n infrstruture provier (Inp) is eploye to gther informtion from RSUs n provie lultion servies for vehile pplitions. Messges or requests generte by vehiles re first trnsmitte to RSUs n these messges n requests will be forwre to Inp through RSU to Inp (R2I) links. In this pper we propose hybri vehiulr network frmework, integrting both entrlize n istribute vehiulr network strutures. In the entrlize struture, vehiles iretly ess to RSUs. Vehiles on the ro n iretly ommunite with RSU through V2R links. In the istribute vehiulr network struture jent vehiles ommunite with eh other vi vehile to vehile (V2V) links n form luster. Messges n be store n forwre in eh rely vehile. The vehile with the best link to RSU in the luster is selete s the gtewy vehile. Any vehiles in the luster n trnsmit messges to the gtewy vehile vi multi-hop V2V links. The gtewy vehile forwrs this messges to the ssoite RSU. Compre with istribute vehiulr network struture, the entrlize vehiulr network struture n effetively reue the trnsmission lteny of messges when the trnsmission istne is short beuse there is no unneessry store n forwr lteny. However, the link relibility of the entrlize vehiulr network struture epens on the istne between the vehile n the ssoite RSU. When the istne between the vehile n ssoite RSU is very long, the link relibility between the vehile n RSU beomes very low. In this se, the

4 istribute vehiulr network struture n be opte to gurntee the link relibility vi multi-hop V2V links. Sine the istne between jent vehiles is shorter thn or equl to the istne between the soure vehile n the ssoite RSU in multi-hop links, the relibility of the multi-hop links n be signifintly improve. However, suh multi-hop trnsmission introues long ely. Therefore, the totl lteny of the multi-hop V2V links beomes very high when the number of rely vehiles is lrge. To meet ifferent requirements of vehiulr networks, in this pper hybri vehiulr network frmework is propose by effetively integrting these two type of vehiulr network strutures s shown in the right sie of Fig. 1. Sine the Inp n gther informtion from RSUs n vehiles, the Inp is serve s the entrl ontroller of this frmework n hs ess to the following informtion: 1) lotions of RSUs n vehiles; 2) stte informtion of V2R n V2V links. A istne threshol bse on whih the ontroller etermines whih type of D thre network struture is selete for the hybri vehiulr network frmework. When the istne between the vehile n ssoite RSU is less thn or equl to, the entrlize vehiulr network struture is opte for messges trnsmission. In this se, the vehile ommunites with the ssoite RSU iretly. When the istne between the vehile n ssoite RSU is lrger thn, the istribute vehiulr network struture is opte for D thre messges trnsmission. In this se, the vehile to the gtewy vehile gtewy vehile V g V g V out D thre V in trnsmits messges vi V2-ho links whih is losest to the ssoite RSU in the vehile luster. The forwrs these messges to the ssoite RSU. 3. Vehiulr Network Utility Funtion To evlute the performne of hybri vehiulr network frmework, unifie vehiulr network utility funtion is efine s ef w xi xi 1 xi, (1) where w x i is the lteny utility funtion, x i is the relibility utility funtion, 1 is n initor funtion, x i is the istne between the vehile n the ssoite RSU, n the symbol is the set of ll possible istnes between the vehile n the ssoite RSU. In (1) the vehiulr network utility funtion is efine s n itive funtionl form of the eterministi effet of the lteny utility funtion n relibility funtion. By utilizing the Eulien norm with weights, the effet of lteny n relibility on vehiulr network performne n be evlute in the vehiulr network utility funtion spe. As mentione in Setion II, istne threshol D is use for the ontroller to etermine whih type of network strutures is selete for the hybri vehiulr network frmework. Therefore, the vehiulr network utility funtion with the entrlize or istribute vehiulr network struture n be lssifie oring to the istne between the vehile n the ssoite RSU. When the istne between the vehile istne threshol D thre x i thre n the ssoite RSU x i is less thn the, high relibility of links n be ensure. In this se, the vehiulr

5 network utility epens on the lteny utility. Sine there is no store-n-forwr lteny in entrlize vehiulr network struture ompre with istribute vehiulr network struture, entrlize vehiulr network struture is opte in this se. The vehiulr network utility funtion with entrlize vehiulr network struture n be expresse s ef C w x x 1 x, x Dt hre C i i i i w where the step is the Eulien norm opertion, vehile to the ssoite RSU, n w 2, (2) enotes the wireless uplink from the re the lteny utility funtion n relibility utility funtion in entrlize vehiulr network struture, respetively. is the weight ftor of lteny utility whih inites the weight of lteny utility in the vehiulr network utility funtion. Similrly, is the weight ftor of relibility utility. The uplink in the hybri vehiulr network is onfigure s wireless link between the vehile n the Inp. Messges of pplitions re proue in vehiles n nee to be proesse in Inp, so wireless link is onfigure to uplo these messges. Consiering the URLLC in vehiulr networks, these messges nee to be trnsmitte with short lteny n high suess probbility. Therefore, the uplink utility funtion is introue to evlute the network performne uner ifferent uplinks with ifferent trnsmission istnes. When the istne between the vehile V n ssoite RSU is lrger thn the istne threshol D thre, to overome the negtive impt of long istne link on the relibility of vehiulr network, the istribute vehiulr network struture is opte. The vehiulr network utility funtion uner istribute vehiulr network struture is expresse s ef w x x 1 x, x Dt hre i i i i b where w n w , (3) represent the lteny utility funtion n the relibility utility funtion uner istribute vehiulr network struture, respetively. The weight ftor of lteny utility n relibility is efine s n. To simplify the nottion of the lteny n relibility utility funtion, we set, n then the vehiulr network lteny utility funtion w n relibility utility funtion is extene s w Treq T Treq e, (4) Preq P P req e, (5)

6 where T is the uplink trnsmission lteny when the uplink is, P is the probbility of suessful messge trnsmission, i.e., the trnsmission relibility. The lteny n relibility requirements re enote s T req n P req, respetively. Bse on (4) n (5), the vlues of these two utility funtions re lwys lrger thn 0. When the lteny n relibility requirements re both stisfie in vehiulr networks, i.e., T T req n P P req, the vlues of lteny n relibility utility funtions re lwys lrge thn or equl to 1. Otherwise, the vlues of lteny n relibility utility funtions re lwys lrge thn 0 n less thn 1 if the lteny n relibility requirements re both not stisfie, i.e., T Treq n P Preq. 4. Uplinks Anlysis of Hybri Vehiulr Networks In this setion the uplinks of hybri vehiulr networks, i.e., the wireless link from the vehile to the ssoite RSU n the wire link from the RSU to the Inp re investigte. 3.1 Uplink Lteny in Hybri Vehiulr Networks In this se, the whole uplink lteny in hybri vehiulr networks TV I onsists of two prts: the first prt omes from wireless V2R uplinks, enote s the wireless trnsmission lteny ; the seon prt is the trnsmission lteny in wire links between RSUs n Inp, TV R enote s expresse s TR I. Therefore, the uplinks lteny V I V R RI TV I in hybri vehiulr networks is T T T. (6) Assume tht millimeter wve trnsmissions re use for V2n V2R ommunitions. In etil, 72 GHz millimeter wve pth loss moel is use to nlyze the pth loss of wireless signl trnsmissions [13-14]. Bse on the results in [14], when the istne between the trnsmission noe n reeiving noe is, the pth loss is expresse s PL[ B]( ) log( ), ~ 0, 2, (7) where is the show fing oeffiient, n is the stnr evition of show fing. Sine the bem forming tehnology is wiely opte in millimeter wve trnsmissions, the urbn senrio in this pper is regre s noise-limite senrio [15], the signl to noise rtio (SNR) in the reeiving noe is erive s SNR B P B PL B N W B, (8) tx 0 mmwve where P tx is the trnsmission power of vehiles, N 0 is the Gussin white noise power spetrl ensity n W mmwve is the bnwith of millimeter wve. When the SNR t the

7 reeiving noe is lrger thn the SNR threshol, the messges n be reeive suessfully. Therefore, the trnsmission suess probbility in the uplink with the single hop is erive s hop tx 0 mmwve tx 0 mmwve P P SNR B B P P B PL B N W B B P PL B P B B N W B P P B B N W B log tx 1 1 erf mmwve 10 0 mmwve 10 P hop, (9) P B B N W B log, (9b) tx where erf () is the Gussin error funtion. Assume tht messges re enpsulte in pkets for wireless trnsmissions in hybri vehiulr networks. The trnsmitting slot of pket is enote s. Therefore, the trnsmission lteny in the single hop is expresse s t slot T hop T hop t slot P hop 2tslot. (10) 1 erf 2 When entrlize vehiulr network struture is opte in hybri vehiulr networks, V2R uplink is one hop wireless link. Assume the istne of uplink is. The wireless trnsmission lteny uner entrlize vehiulr network struture TV R Centrlize is expresse s 2tslot TV RCentrlize Thop. (11) 1 erf 2 When istribute vehiulr network struture is opte in hybri vehiulr networks, V2R uplink is the multi-hop wireless link. In this se, the multi-hop wireless uplink is expresse s D1 D2 D y N =,,, 1,2,,, (12) y hop

8 where Let t pro D y istne of is the y th hop of multi-hop uplink n N hop enotes the store-n-forwr lteny in eh rely vehile n D y enotes the number of the hops. y enotes the. The wireless trnsmission lteny uner istribute vehiulr network struture TV R Distribute is erive s T T t. (13) N 1 V RDistribute hop y hop pro Dy In this pper RSUs re only responsible for the trnsmission of messges n Inps provie t storge n lultion servies. Different from the wireless trnsmission lteny, empiril mesurements show tht there is long-til effet on the lteny of t pkets in wire networks [16]. Therefore, exponentil istribution moels re wiely use to nlyze the lteny of t pket in wire links. Without loss of generlity, R2I wire uplink lteny is expresse s [16] TV I 1 T r e r, (14) R 2 2 I r RI w 0 I wri I 2 where r is the intermeite vrible, is the ensity of RSU, is sling ftor tht inites the wire hnnel stte. R I is the ensity of Inp, w Assume tht the istne between two jent RSUs is n the number of RSUs ssoite with Inp is N. Therefore, the ensity of RSU in this re is expresse s RSU R I RSU L N. (15) Let Inp enote typil vehiulr network re whih inlues Inp n Bse on the results in [17-18], the PDF of the re of Inp is expresse s N RSU RSUs. f Inp I b 1 bi, (16) e t e 0 is the Gmm funtion, where 1 n is the intermeite vrible, is the shpe prmeter n b I is the inverse sle prmeter for Gmm istribution [19-20]. The ro length in the overge of Inp is erive s I b 1 bi L Inp 0 ro e, (17)

9 where ro is the ro ensity in urbn environments. Assume tht RSUs re governe by uniform istribution. The expettion of N RSU is erive s I 1 b 1 bi 0 ro NRSU e L. (18) Aoring to (14), (15) n (18), the R2I wire uplink lteny TR I in hybri vehiulr network is erive s 2 R 2 I r TR I w 2 0 I r e r I w R I N 12 RSU w I I 1 2 b 1 w I 0 ro 1 bi e 2L 3.2 Relibility of Uplinks in Hybri Vehiulr Networks. (19) Sine the link outge problem in hybri vehiulr networks is minly use by the wireless links, the probbility of suessful messge trnsmission in wireless links P is use to represents the relibility of hybri vehiulr networks. When entrlize vehiulr network struture is opte in hybri vehiulr networks, the P equls to the one-hop trnsmission probbility n is uplink suess probbility expresse s 1 1 P Phop erf. (20) 2 2 When istribute vehiulr network struture is opte in hybri vehiulr networks, the P equls to the prout of suess probbility uplink suess probbility of the uplinks of eh single hop n is expresse s hop y Dy P P. (21) Let 5. Uplinks Optimiztion of Hybri Vehiulr Networks enotes the set of ll possible uplinks for vehile in hybri vehiulr networks. Obviously, set ontins not only the uplinks of entrlize vehiulr network struture but lso the uplinks of istribute vehiulr network struture. To hieve high network utility,

10 the vehile nees to hoose suitble uplink with low lteny n high relibility from. Compre with multi-hop links in istribute vehiulr network struture, the store-n-forwr lteny n be voie in entrlize vehiulr network struture. However, the relibility of istribute vehiulr network struture n be improve by shorting the wireless trnsmission istne ompre to the entrlize vehiulr network struture. Therefore, there is treoff between lteny n relibility for optimizing uplinks in the hybri vehiulr networks. In orer to mximize, moel lle Preto improvement whih is wiely use in fiel of eonomis is introue to optimize the seletion of uplinks from set. In the fiel of eonomis, the onept of Preto improvement is to mke ny iniviul or preferene riterion better without mking t lest one iniviul or preferene riterion worse [21]. In this se, the uplink lteny utility n relibility utility n be trete s two iniviuls in Preto improvement moel n the optimiztion objetive funtion is in hybri vehiulr networks. Bse on the Preto improvement, new uplink in set will be selete if les to shorter lteny without reuing the relibility or les to higher relibility without inresing the lteny ompre with the urrent link, i.e., T T &P P P P &T T. or Therefore, the optimiztion problem is to fin n optiml uplink opt whih mximizes the network utility bse on Preto improvement. The optimiztion problem is formulte s rgmx opt subjet to opt opt y 0. (22) Dy opt e1 D D req opt e2 P P & T T e3 &, T T & P P req opt opt opt where the onstrint e 1 inites tht if the uplink is multi-hop link, the sum of istnes in ll hops shoul be lrger thn or equls to whih is the istne between the vehile n RSU in the iret V2R link. The onstrint e 2 gurntees mximum lteny n minimum relibility requirements tht n be tolerte in vehiulr networks. The onstrint e 3 is the onition of Preto improvement. In orer to solve the optimiztion problem (22), n itertive lgorithm, i.e., the uplink preto improvement lgorithm for hybri vehiulr network is evelope s follows. Algorithm 1: Uplink Preto Improvement Algorithm for Hybri Vehiulr Network D 0

11 Begin 1: Initilize opt, D 0, T req, P req T,, opt P opt ; 2: Get number of vehiles between trnsmission vehile n it s losest RSU s ; ll vehiles position s j 1,2,, N j v N v, get 3: If Nv 0 4: T T D ; opt hop 0 5: P P D 6: else ; opt hop 0 7: For n 1:1: Nv o 8: For k 1:1: n 9: Initilize T, P, ; D0 10: Set trget positions k k ; n 1 11: For i 1:1: Nv 12: Fin the losest vehile j 1,2,, N, bs bs ; j v j k j k 13: k j; 14: En for 15: 1 V i T T Thop k k ; P P Phop k k 1 ; 16: 17: Get by eqution (3); 18: En For 19:,,, k k : If 21: opt ; s the kth rely vehile whih stisfy opt & T T opt & P P opt

12 22: opt ; 23: En if 24: En For 25: En if The Preto Optimlity link is opt with the best opt. 6. Numeril Results n Disussion To evlute the performne of the propose lgorithm, the onfigurtion prmeters for the numeril simultions re summrize in Tble 1. Tble 1. onfigurtion prmeters Prmeter Vlue The ensity of Inp ~ 3 10 per squre meter[15-16] The vehile ensity on the ro 0.08~0.24 vehile per meter [22] I The trnsmission power of vehiles N 0 v P tx The noise power ensity The stnr evition of show fing The urtion of slot t slot The SNR threshol The sling ftor of wire links between RSU n Inp The ro ensity uner urbn environment b The uplink lteny requirement The uplink relibility requirement w T req 30 Bm -174 Bm/Hz [14, 24] 5 50 miroseons [15] 5 B [16] ro meter per squre meter [23] 3.61 [19] 3.57 [19] 1 milliseon [11] P req 0.9,,, 0.5 Fig. 2 shows the uplink suess probbility P with respet to the istne D 0 between vehile n the ssoite RSU onsiering ifferent vehiulr network strutures n vehile ensities on the ro. When the vehiulr network struture is fixe, it is shown tht V uplink suess probbility ereses s the between vehile n the ssoite RSU inreses. When the istne between vehile n the ssoite RSU is fixe, the uplink suess probbility of entrlize vehiulr network struture is less thn the uplink suess probbility of istribute vehiulr network strutures. Furthermore, when the istne between vehile n the ssoite RSU is fixe, the uplink suess probbility of istribute vehiulr network struture inreses with the vehile ensity on the ro.

13 Fig. 2. Uplink suess probbility with respet to the istne between vehile RSU n the ssoite Fig. 3. Wireless uplink V2R lteny (ms) with respet to the istne between vehile V n the

14 ssoite RSU Consiering entrlize vehiulr network struture n istribute vehiulr network struture, the impt of istne between vehile n the ssoite RSU on the wireless uplink V2R lteny TV R is investigte in Fig. 3. When the vehiulr network struture is fixe, the wireless uplink V2R lteny inreses with the istne between vehile ssoite RSU. When the istne between vehile n the n the ssoite RSU is fixe, the wireless uplink V2R lteny of entrlize vehiulr network struture is less thn the wireless uplink V2R lteny of istribute vehiulr network struture. Fig. 4 illustrtes the wire uplink R2I lteny TR I with respet to the Inp ensity onsiering ifferent istnes between jent RSUs. When the istne between jent RSUs is fixe, the wire uplink R2I lteny ereses with Inp ensity. When the Inp ensity is fixe, the wire uplink R2I lteny ereses s istne L between jent RSUs inreses. When the Inp ensity is fixe, the expettion of istne between Inp n vehile is fixe. When the istne L between jent RSUs is inrese, the istne between vehiles n the ssoite RSU is erese. As onsequene, the wire uplink R2I lteny is inrese. Fig. 4. Wire uplink R2I lteny (ms) with respet to the Inp ensity n istne between jent RSUs Fig. 5 shows the vehiulr network utility with respet to the istne between vehile n the ssoite RSU onsiering ifferent vehiulr network strutures. When the vehiulr network struture is fixe, the vehiulr network utility is eresing funtion of istne between vehile n RSU. When the istne between vehile V n the ssoite RSU is

15 less thn 27 meters, the vehiulr network utility of entrlize vehiulr network struture is lrger thn tht of istribute vehiulr network struture. This inites tht vehiulr network utility is minly ffete by the lteny when the istne between vehile n the ssoite RSU is short, i.e., the istne between vehile thn 27 meters. When the istne between vehile n the ssoite RSU is less n the ssoite RSU is lrger thn or equl to 28.3 meters, the vehiulr network utility of entrlize vehiulr network struture is less thn tht of istribute vehiulr network struture. This result implies tht vehiulr network utility is minly ffete by relibility when the istne between vehile V n the ssoite RSU is long, i.e., the istne between vehile n RSU is lrger thn or equl to 28.3 meters. Fig. 5. Vehiulr network utility with respet to the istne between vehile V n the ssoite RSU onsiering ifferent vehiulr network strutures In Fig. 6() n Fig. 6(b), the impt of istne between vehile V n the ssoite RSU n network requirements on the vehiulr network utility is investigte. When the network requirements re fixe, the utility of entrlize vehiulr network struture is higher thn tht of istribute vehiulr network struture in the se tht istne between vehile n the ssoite RSU is less thn threshol. When the network requirement is more stringent, i.e., the soli lines inites stringent network requirement ompre with otte lines, the threshol will inrese.

16 () Different lteny requirements (b) Different relibility requirements Fig. 6. Vehiulr network utility with respet to the istne between vehile V n the ssoite RSU onsiering ifferent network requirements

17 Fig. 7. Uplink suess probbility with respet to the istne between vehile n the ssoite RSU onsiering ifferent network ess strtegies Fig. 8. Uplink V2I lteny (ms) with respet to the istne between vehile n the ssoite RSU onsiering ifferent network ess strtegies

18 Fig. 7 ompres the Uplink suess probbility with respet to the istne between vehile n the ssoite RSU onsiering three types of network ess strtegies. When the istne between vehile n the ssoite RSU is less thn or equl to 25 meters, the uplink suess probbility of Preto improvement strtegy is equl to the uplink suess probbility of ellulr ess strtegy in vehiulr networks. When the istne between vehile n the ssoite RSU is lrger thn 25 meters, the uplink suess probbility of Preto improvement strtegy is lose to the uplink suess probbility of rnom rely strtegy. Fig. 8 ompres the uplink V2I lteny with respet to the istne between vehile n the ssoite RSU onsiering three types of network ess strtegies. When the istne between vehile n the ssoite RSU is less thn 25 meters, the propose Preto improvement strtegy hs the sme performne with the ellulr ess strtegy. When the istne between vehile n the ssoite RSU is lrger thn 25 meters, the uplink V2I lteny of Preto improvement strtegy is less thn the uplink V2I lteny of rnom rely strtegy n lrger thn the uplink V2I lteny of ellulr ess strtegy. Fig. 9. Vehiulr network utility with respet to the istne between vehile n the ssoite RSU onsiering ifferent network ess strtegies Fig. 9 shows the vehiulr network utility omprison mong three ifferent network ess strtegies with respet to the istne between vehile n the ssoite RSU. The ellulr ess strtegy is the entrlize network struture in vehiulr networks. The rnom rely strtegy mens tht eh noe hooses network noe rnomly within its ommunition rnge s the rely noe. The Preto improvement strtegy refers to the propose uplink Preto improvement lgorithm is opte in vehiulr networks. When the network ess strtegy is fixe, the vehiulr network utility ereses with inrese of istne between vehile n RSU. When the istne between vehile n the ssoite RSU is lrger thn 25 meters, the vehiulr network utility of Preto improvement strtegy lwys outperform the vehiulr network utility of ellulr ess strtegy n rnom rely strtegy in vehiulr networks. When the istne between vehile n the ssoite RSU is less thn or equl to 25 meters,

19 the vehiulr network utility of Preto improvement strtegy is equl to the vehiulr network utility of ellulr ess strtegy in vehiulr networks. The Preto improvement strtegy is bse on the Preto improvement sheme whih n le to Preto optimlity on this problem, so this lgorithm n lwys give n optimum utility funtion uner ifferent istnes between vehiles n the ssoite RSU. 7. Conlusion In this pper, novel hybri vehiulr network frmework ombining entrlize n istribute vehiulr network strutures ws propose. Bse on this network frmework, new vehiulr network utility moel bse on Eulien norm theory ws propose to evlute the lteny n relibility utility of vehiulr networks. Moreover, vehiulr network uplink optimiztion lgorithm bse on Preto improvement ws propose to optimize the vehiulr network utility. Simultion results inite tht the propose uplink optimiztion lgorithm n improve the uplink relibility n reue the uplink lteny n the vehiulr network utility is improve up to 15%. Referenes [1] B. Pen, M. Cp, S. Z. Yong, et l., A survey of motion plnning n ontrol tehniques for self-riving urbn vehiles, IEEE Trnstions on Intelligent Vehiles, vol. 1, no. 1, pp , Mrh, Artile (CrossRef Link) [2] X. Ge, Z. Li n S. Li, 5G softwre efine vehiulr networks, IEEE Communitions Mgzine, vol. 55, no. 7, pp , July, Artile (CrossRef Link) [3] A. Osseirn, F. Bori, V. Brun et l., Senrios for 5G mobile n wireless ommunitions: the vision of the METIS projet, IEEE Communitions Mgzine, vol. 52, no. 5, pp , My, Artile (CrossRef Link) [4] A. Festg, Coopertive intelligent trnsport systems stnrs in europe, IEEE Communitions Mgzine, vol. 52, no. 12, pp , Deember, Artile (CrossRef Link) [5] G. Krginnis, O. Altints, E. Ekii, et l., Vehiulr networking: survey n tutoril on requirements, rhitetures, hllenges, stnrs n solutions, IEEE Communitions Surveys & Tutorils, vol. 13, no. 4, pp , July, Artile (CrossRef Link) [6] G. Arniti, C. Cmpolo, M. Conolui, et l., LTE for vehiulr networking: survey, IEEE Communitions Mgzine, vol. 51, no. 5, pp , My, Artile (CrossRef Link) [7] S. Chen, J. Hu, Y. Shi, et l., LTE-V: TD-LTE-bse V2X Solution for future vehiulr network, IEEE Internet of Things Journl, vol. 3, no. 6, pp , Deember, Artile (CrossRef Link) [8] S. Chen, J. Hu, Y. Shi, et l., Vehile-to-everything (v2x) servies supporte by LTE-bse systems n 5G, IEEE Communitions Stnrs Mgzine, vol. 1, no. 2, pp , July, Artile (CrossRef Link) [9] S. I. Sou n O. K. Tonguz, Enhning VANET onnetivity through rosie units on highwys, IEEE Trnstions on Vehiulr Tehnology, vol. 60, no. 8, pp , Otober, Artile (CrossRef Link) [10] C. Sho, S. Leng, Y. Zhng, et l., Performne nlysis of onnetivity probbility n onnetivity-wre MAC protool esign for pltoon-bse VANETs, IEEE Trnstions on Vehiulr Tehnology, vol. 50, no. 12, pp , Deember, Artile (CrossRef Link) [11] X. Li, B. j. Hu, H. Chen, et l., Multi-hop ely reution for sfety-relte messge brosting in vehile-to-vehile ommunitions, IET Communitions, vol. 9, no. 3, pp , Februry, Artile (CrossRef Link) [12] X. Ge, S. Tu, G. Mo, et l., 5G Ultr-Dense Cellulr Networks, IEEE Wireless Communitions, vol. 23, no. 1, pp , Februry, Artile (CrossRef Link)

20 [13] S. Li, Z. Li, X. Ge, et l., Multi-hop links qulity nlysis of 5G enble vehiulr networks, in Pro. of 9th Interntionl Conferene on Wireless Communitions n Signl Proessing, pp. 1-6, Otober 11-13, Artile (CrossRef Link) [14] A. Ghosh, T. A. Thoms, MC. Cuk, et l., Millimeter-wve enhne lol re systems: high-t-rte pproh for future wireless networks, IEEE Journl on Selete Ares in Communitions, vol. 32, no. 6, pp , June, Artile (CrossRef Link) [15] G. Zhng, T. Q. S. Quek, A. Hung, et l., Dely moeling for heterogeneous bkhul tehnologies, in Pro. of 2015 IEEE 82n Vehiulr Tehnology Conferene, pp. 1-6, September 6-9, 2015.Artile (CrossRef Link) [16] D. C. Chen, T. Q. S. Quek n M. Kountouris, Bkhuling in heterogeneous ellulr networks: moeling n treoffs, IEEE Trnstions on Wireless Communitions, vol. 14, no. 6, pp , June, Artile (CrossRef Link) [17] T. Mttfelt, Stohsti geometry n its pplitions, Journl of Mirosopy, vol. 183, no. 3, pp , September, Artile (CrossRef Link) [18] J. S. Feren, n Z. N, On the size istribution of Poisson Voronoi ells, Physi A Sttistil Mehnis & Its Applitions, vol. 385, no. 2, pp , June, Artile (CrossRef Link) [19] L. Xing, X. Ge, C. X. Wng, et l., Energy effiieny evlution of ellulr networks bse on sptil istributions of trffi lo n power onsumption, IEEE Trnstions on Wireless Communitions, vol. 12, no.3, pp , Mrh, Artile (CrossRef Link) [20] X. Ge, H. Chen, G. Mo, et l., Vehiulr Communitions for 5G Coopertive Smll Cell Networks, IEEE Trnstions on Vehiulr Tehnology, vol. 65, no.10, pp , Otober, Artile (CrossRef Link) [21] Y. Liu n L. Dong, Spetrum shring in MIMO ognitive rio networks bse on oopertive gme theory, IEEE Trnstions on Wireless Communitions, vol. 13, no. 9, pp , September, Artile (CrossRef Link) [22] W. Zhng, Y. Chen, Y. Yng, et l., Multi-hop onnetivity probbility in infrstruture-bse vehiulr networks, IEEE Journl on Selete Ares in Communitions, vol. 30, no. 4, pp , My, Artile (CrossRef Link) [23] H. Fng, C. Ren, L. Lin, et l., A ro ensity prtition metho for the evlution of ro seletion, in Pro. of 23r Interntionl Conferene on Geoinformtis, pp. 1-6, June 19-21, Artile (CrossRef Link) [24] X. Ge, Y. Sun, H. Ghrvi n J. Thompson, Joint Optimiztion of Computtion n Communition Power in Multi-user Mssive MIMO Systems, IEEE Trnstions on Wireless Communitions, Vol. 17, No. 6, pp , June Artile (CrossRef Link) Shikun Li reeive his Bhelor s egree in ommunition n informtion systems from Huzhong University of Siene n Tehnology (HUST), Wuhn, Chin in in 2016, where he is urrently working towr his Ms- ter s egree. His reserh interests inlue Hybri vehiulr network n 5G ommunition systems.

21 Zipengli Li reeive the B.E. egree in Teleommunition Engineering n M.S. egree in Communition n Informtion System from Huzhong University of Siene n Tehnology (HUST), Wuhn, Chin in 2011 n 2014, respetively. He is urrently working towr the Ph.D. egree in HUST. His reserh interests inlue vehiulr networks n 5G mobile ommunition systems. Xiohu Ge is urrently full Professor with the Shool of Eletroni Informtion n Communitions t Huzhong University of Siene n Tehnology (HUST), Chin. He is n junt professor with with the Fulty of Engineering n Informtion Tehnology t University of Tehnology Syney (UTS), Austrli. He reeive his PhD egree in Communition n Informtion Engineering from HUST in He hs worke t HUST sine Nov Prior to tht, he worke s reserher t Ajou University (Kore) n Politenio Di Torino (Itly) from Jn to Ot His reserh interests re in the re of mobile ommunitions, trffi moeling in wireless networks, green ommunitions, n interferene moeling in wireless ommunitions. He hs publishe more thn 200 ppers in referee journls n onferene proeeings n hs been grnte bout 25 ptents in Chin. He reeive the Best Pper Awrs from IEEE Globeom Dr. Ge serve s the generl Chir for the 2015 IEEE Interntionl Conferene on Green Computing n Communitions (IEEE GreenCom 2015). He serves s n ssoite eitor for IEEE Wireless Communitions, IEEE Trnstions on Vehiulr Tehnology n IEEE ACCESS, et. Yonghui Li reeive his PhD egree in November 2002 from Beijing University of Aeronutis n Astronutis. From , he ws ffilite with Linkir Communition In, where he hel position of projet mnger with responsibility for the esign of physil lyer solutions for the LAS-CDMA system. Sine 2003, he hs been with the Centre of Exellene in Teleommunitions, the University of Syney, Austrli. He is now Professor in Shool of Eletril n Informtion Engineering, University of Syney. He is the reipient of the Austrlin Queen Elizbeth II Fellowship in 2008 n the Austrlin Future Fellowship in His urrent reserh interests re in the re of wireless ommunitions, with prtiulr fous on MIMO, millimeter wve ommunitions, mhine to mhine ommunitions, oing tehniques n oopertive ommunitions. He hols number of ptents grnte n pening in these fiels. He is now n eitor for IEEE trnstions on ommunitions n IEEE trnstions on vehiulr tehnology. He lso serve s guest eitor for severl speil issues of IEEE journls, suh s IEEE JSAC speil issue on Millimeter Wve Communitions. He reeive the best pper wrs from IEEE Interntionl Conferene on Communitions (ICC) 2014, IEEE PIMRC 2017 n IEEE Wireless Dys Conferenes (WD) 2014.