Distributed Cache Enabled V2X Networks: Proposals, Research Trends and Challenging Issues

1 Distrib uted Cache Enabled V2X Netw ork, Proposals, Researc h T rends and Challenging I ssues Di Zhang, Member , IEEE Abstract —Nowadays, vehicle to everything (V2X) has been proposed to connect everything on th e road. Howe ver , the current V2X network is motor -vehicle oriented . In contrast, the internet of sharing bicycle (IoSB ) network is vastly and rapidly depl oyed as a feasible solution for t h e last mile problem ( e. g., from station to home/office). T o make full u se of the devices for traffic safety and wireless communications, a comprehensive network model i s needed. In this work, while inv estigating the existing studies, we pr opose a versa tile V2X network with a distributed framewo rk and heterogeneous caching method. In this V2X network, all the devices on th e road (motor - vehicle, non-motor -vehicle, ped estrian, etc.) are connected. W e further introduce a heterogeneous cache method for effective wireless transmission while utilizing the massiv e connected devices. The potential resear ch trends on achieving high speed transmission, deep programming dedi cated network slicing an d big data/ machine learning (ML)/ edge computing b ased image reco gnition and reconstruction are highlighted to provide some in sights for future stu d ies. Fin ally , the challenging i ssues of p ath l oss model, channel model and ultra reliable and low latency communications (URLLC) of th is V2X network are discussed. I . I N T RO D U C T I O N In literatur e, the primar y mo ti vation of co nnecting vehicle is to prevent the collision [3]. W ith this purpo se, IEEE 802. 1 1 p standar d was released in 201 0, an d the ded ic a te d sh ort range commun ications (DSRC) with carrier frequency 5.9 GHz was allocated. In 2 016, 3GPP relea se 14 has initialed the cellular-V2X (C-V2X), the 5G automo ti ve association (5GAA) is established in the same year with the target on C-V2X solutions (i.e. , vehicle to vehicle (V2V), vehicle to pedestrian (V2P), vehicle to infr astructure (V2I), vehicle to network (V2N)) fo r future m o bility and tran sportation services). Based on the C-V2X network, op ening issues o f automatic p iloting, traffic sign al con trol, collision detection , emergency warning and o ptimal r oute p lanning h av e been raised [4]. Howe ver, up to now , it is found that in V2 X studies (DSRC an d C- V2X), existing work mostly is moto r-vehicle oriented , the other elements are playing th e seco ndary ro les f o r the vehicle’ s obstacle d etection, collision p revention, etc. At th e same time, the sha ring bicycle system, with its flexible de p loyment and en vironme n t friendly ch aracters, is becoming a con venie n t solution for the last m ile proble m (e.g. , fr om station to office, home, etc.). I n China, it is believed that the narrow b and IoT (NB-IoT) based sharing bicycle system will bec ome th e first and largest IoT case in the com ing y ears [5]. In ord er to accomplish th e V2X’ s ambitious with conn ecting ev erything , Di Zhang is with the School of Information Enginee ring, Zhengzhou Uni versity , Z hengzhou , 450001, China . fast speed tran smission and in telligent tra ffic service, combi- nation of curren t V2 X, shar ing bicycle a n d other devices is thus inevitable. On the o th er hand, in wireless co mmunicatio ns, conten t caching an d sharin g mech anism (CCSM) has been intensively studied. By in voking CCSM, content requ e st can be satis- fied fr om th e temp orary caches loca te d in the core r outer, base station (BS) and neigh b oring user sides. Th e engaged transceiver equipmen ts, links and resour ces (e.g., carrier fre- quency , consumed energy ) can be greatly reduc ed v ia th is method, which yields an improved system en ergy efficiency (EE) perfo rmance. The previous stud ies o f co ntent caching in vehicle network, h owe ver, were mostly restricted to the motor-vehicle V2V co mmunic a tio ns, for in stance, [6]. Du e to a mu ch smaller n umber of connected motor-vehicles, it is not an effectiv e way to obtain requ e sting contents from the vehicle caches. The merits of con tent cach ing technolo g y are reduce d in this case. In or der to solve the aforemen tio ned pro blems, here in this ar ticle, we p ropose the versatile d istributed cache e n - abled V2X n e tworks. All the devices (i. e ., motor-vehicle, non-m otor-vehicle, we arable smart device an d cell- p hone o f the p edestrian) on -the-ro ad are compr ehensively con n ected for inte llig ent traffic system ( ITS) con trol with regard to public an d traffic safety . Besides, caches are distributed to all conn ected vehicles and devices (mo tor-vehicle, no n-moto r- vehicle, pedestrian, etc.), wh ic h are not limited to the mo tor- vehicles. W e furth er exploit th e massiv e c o nnected vehicles and de vices in the d istributed cache enabled V2X n etworks for wireless transmission with the help of co n tent cac hing technolog y . T o e ffectively use the carrier frequ ency resources, we pro pose an initial c o-existing so lu tion for L TE, DSRC and millime te r wa ve (mmW a ve) fr equencies. W ith the purpo se of redu cing carbon emission, the solar energy is in voked as well by introdu cing the solar en ergy conv e rters to the vehicle bodies. The future resear ch trends and c h allenging issues on ac h iev- ing the distributed cache en abled V2X networks are fo c used afterwards. For the implementatio n, a globa l perspective is needed to re-design the n etwork a r chitecture and infras- tructure. W e discuss the majority futur e research trend s on achieving higher transmission rate, the effectiv e and dedicate intelligent traffic co ntrol n etwork architectu re with networking slicing. T h e big data, mach ine learning (ML) and fog com- puting based im a g e recogn ition and reconstru ction is anothe r topic f or fu ture research . The urban wireless path loss and channel mod els brin g in great c h allenge fo r the quality of service (QoS) of wireless tran smission and intelligence traffic 2 control. T o this en d, we sketch th e path loss model wh ile incorpo rating prior studies. The ultra r eliable co mmunicatio n and lo w latency requ irement are the other challenging issues for its commercial use with regard to the pub lic and traffic safety . I I . T H E C C S M , V 2 X A N D S H A R I N G B I C Y C L E In this section, we discuss the existing work on co ntent caching tech nology , V2X systems and sharing bicycle systems. In wireless comm unication , conten t caching mostly restricts to system EE analy sis, optimal content division and positioning . The V2X and sharing bicycle systems, on the oth er hand , are taken as the promising IoT ap plication scen e s to resha pe the future driving experien ces and vehicle industry . A. The F ea tur es and Benefits of CCSM CCSM was initially o riginated fr om the in formatio n cen tric networking ( ICN). In this paradig m, conten t is distributed in in-network stora ges with a n ame. Th rough mu lticast me c h - anisms, ICN can timely and efficiently deliv er the co ntents to sub sequent users. Nowadays, other than th e upp er layer studies, ICN is introduced to the wireless communicatio n s as well, which is known as the CCSM. As shown by Fig. 1, CCSM assumes that a ll the devices (cor e ro uter, BS, mobile terminal, etc.) can cache and shar e th e tem porary conten ts for subsequen t u sers. In this case, routin g back and o btain th e request contents from the remote center is n ot ne cessary . The benefits of CCSM can be summarized as follows: Alleviated network load : The requ e stin g data can be directly obtained fr om the distributed caches. Compared to routing back and retrieving fr o m the remote content server , the core n etwork load is alleviated. On c o ndition th a t the sub- sequent user gets its requ esting c o ntent from the neighb oring user’ s caches, the BS network load is alleviated. Less enga g ed transceiver components : CCSM need s less engaged transceiver com ponen ts while providing the same amount of data to subseq uent user . F or in stance, if the subse- quent user is satisfied from the neighbo ring user’ s cac h es, the transceiver compon e n ts of BS a nd core router are not needed ; while f rom the BS caches, th e co re network tra n sceiv er compon ents are n ot enga ged; while from the core router caches, th e tran sceiver compon ents of subseq uent rou ters are not necessary . Improv e d EE system performance : The CCSM can reduce the tr ansmission distanc e from tr ansmitter to receiver . Becau se of the shorter distance, emission power is reduced , w h ich brings less system en ergy consump tion. Moreover, CCSM needs le ss eng aged transceiver com p onents. By keep those compon ents into sleep mod e, we may fur ther reduc e the energy consump tion. In o th er words, CCSM con sumes less energy while transmitting th e same amoun t of co n tents. The system EE perfo rmance is the refore im proved. B. The Curr ent V2X and Sharin g Bicycle Networks There are two types of V2 X n etworks, i.e., DSRC and C-V2X. DSRC is to provide active safety commun ications Fig. 1. The CCSM in wireless communications. contributes to safe driving while connec tin g the vehicles to RSU or other vehicles. I n co ntrast, C-V2X aims to conn ect not only the vehicle, pedestrian, but also oth er elem e nts (for instance, tr affic lights, sen sors) on the ro a d. It can provide faster speed and lower latency tr a n smission thro ugh the cel- lular network co mpared to DSRC. A s shown b y Fig. 2. a , vehicles can conn ect to th e roadside units (RSU) v ia V2I , connect to the n etwork via V2 N, con nect amo ng each othe r via V2 V and connect to the ped estrian via V2P . While com- municating , DSRC relies on the wireless local area n etwork (WLAN) ph ysical tran sm ission (PHY) and mediu m access control (MAC). Mean while, C-V2X works o n two mode: 1) d ir ect comm unication mode (DCM), for instance, V2V , V2P; and 2) network based co m munication mo de (NBM), for instance, V2 N , V2I. The ben efits of V2X ne twork can be summarized as follows. Safe a nd intellig ent transport : The centers can broad cast emergency warning message to vehicles via media access control (MAC) broa dcasting to gu ide the vehicles for rescu e and escap e. Wit h the shared real time infor mation, the vehicle can adjust its tr av e l speed to make the green lig ht, take the optimal rou te to av oid traffic congestio n , keep saf e distance to av oid crash, etc. Improv e d in-car entertainment service : The C-V2X net- works enable the vehicle smart device co mmunica tio ns by wireless conn ections working on the unlicensed spectru m (e.g., Bluetooth c o nnection b etween vehicle and cellp hone). For instance, the smart devices can lock o r unlock the vehicle, play video or m usic throu gh the vehicle’ s speakers. Besides, DCM and NBM working on th e licensed spectrum ma kes the access of in-car high - definition me d ia and virtual r eality applicatio ns a reality . Different fro m the C-V2 X network, as shown by Fig. 2. b, the sharing bicycle network, mainly relies on the general packet radio serv ice (GPRS)/Bluetooth/NB-Io T/global positioning system (GPS) technologies for nearest bicycle finding, lo ck/unloc k , infor mation sharing/u ploadin g , location tracking, etc . T h at is, with the help of GPS an d NB-IoT , potential user can find the nearest bicycle and unlock it via GPRS/Bluetooth/NB-IoT . Wh ile ridin g, the real time info rma- 3 Fig. 2. The current V 2X and sharing bicycle systems. tion (loca tio n, speed , acceleration, con sumed e nergy , etc) will be recor ded by the apps or transmitted to the rem ote center . The main benefits of the sharing bicycles network c a n be listed as follows. Last mile problem solutio ns and lower carbon emission : Sharing bicycle network can tackle d own the la st m ile p roblem (e.g., from b us/subway station to offi ce/home) . In this case, more people will p refer the public transpor t, especially when trav elling within the urban area. Carbon emission thus is greatly red uced. Ad ditionally , with less vehicles on th e road , the sharing bicycle n etwork can also alle v iate the traf fic congestion . Personal health diagno sis : Shar ing b icycle network can record the r ider’ s per sonal d ata (e.g ., calo rie consump tion, heart r ate, blo od pressure , ride d istance) with apps o r con - nected wearable sensors. Afterward, with some an alytical tools, the rid e r can e valuate his (h er) physical condition . Th e doctor may also acc ess these data to provid e professiona l health diagno sis under the rider’ s permission. Howe ver, the curr ent V2X ne twork s and sharin g bicycle networks are disconnected. W e might cause a road a c c ident when vehicles and bicycles cannot detect each other . On th e other hand , we canno t provid e the optimal rou ting withou t awareness o f all types of devices on the r oad. It is also worth noting that in V2 X networks, vehicle is p laying a d ominant role. The other elements are playing the secondary roles to assist its safe driving, automatic driving, in- car entertainmen t applications, etc. I I I . T H E P RO P O S E D D I S T R I B U T E D C A C H E E N A B L E D V 2 X N E T W O R K S By incorpo rating the existing works on conten t caching, V2X and sharing bicycle, th e proposed distrib u ted cache enabled V2X networks are elaborated in this section. A. The Pr oposal The distributed cache enab led V2X network s can be shown by Fig. 3. Here in the p roposed networks, CCSM is intro duced to the co re route r, BS and all types of d evices on th e road. In reality , the c a che can be in tegrated into the e ntertainmen t system o f the car, an d located in the intelligent locker system of the bicycle. I n ord er to utilize the clean en ergy , we intro duce   4G/5G BS Router Area center Total center NB-IoT Based locker Solar Panel a nd Battery Solar Pan el and Battery Cache Small cell Fig. 3. The proposed distribute d cache enabled V2X netw orks. All V2X de vices on t he road are connec ted for public and traf fic s afety with the CCSM. In addition, solar panel and battery are equipped in V2X dev ices as well. solar ene rgy converter to the V2X devices 1 (e.g., solar panel on vehicle and d evice b ody). Due to the bicycle’ s limited energy harvest ability , battery is in voked to equip within the basket. That is, the ba tter y u nit is equ ipped to th e bottom of th e basket, the solar energy converter is located over the b attery unit. For the sharin g bicycle within this system, we discard the GPRS/Bluetooth based locker while utilizing the NB-I o T based intelligen t lo cker for fast lock /unlock. Accord in g to the report, lo ck/unlock can be completed within 1 s with NB-I oT , the paym ent process ha s been dropp ed fr om 25 secon ds to less than 5 seconds, while battery life has been pro longed from 1 or 2 month s to more than 2 years [5]. In order to p r ovide fast speed transmission for th e co nnected devices, th e distributed V2X networks can use 5G technolog ies in its NBM transmission. For implemen tation, we can p u t the massi ve multi-inp ut-multi-o utput (MIMO) BS above the building, and vastly dep loy the small cell BSs (femotocell, p ic- ocell) to roadsid e u n its. For the carrier frequen cy , NBM reuses the existing cellular frequ e n cy resour ces for cellular to V2 X devices transmission, an d the m mW ave an d DSRC frequ encies are dedicated to the V2V , V2X, e verything to vehicle (X2V), small cell to vehicle (SC2V) and sma ll cell to pede strian (SC2P) tr a n smissions. Deta iled co m parison between the dis- tributed cache enabled V2 X networks and prior cache enabled vehicle networks, V2X, sharing bicycle are given by T able I. As shown here, the pro posed distributed c ache enabled V2X networks comprehen si vely cover the existing technologies, which ena b les the I TS and wir eless infor mation delivery . The main featur es an d benefits of the pr o posed networks, while comparin g with pr io r work , can be listed as fo llows: Improv e d intellig ent traffic system and wireless trans- mission: In this proposed distributed cache en abled V2X networks, all devices on the road are conn ected f o r public and traffic safety . Th e co nnected V2X devices ar e p laying the equa l roles for wireless commu nication. With the shared upload inform ation f r om all vehicles and d evices on th e road, the co ntrol and service cen te r can mo re effecti vely broadcast 1 While using the wor d devi ces, we mean all de vices besides the vehic le, such as cell -phone, laptop, wearable s m art de vices. 4 T ABLE I C O M PA R I S O N B E T W E E N P R I O R W O R K A N D T H E P RO P O S E D D I S T R I B U T E D C AC H E E NA B L E D V 2 X N E T W O R K S V ehicle Bic ycle Pedestrian Content cachi ng S olar energy DSRC L T E 5G Prior cache enabled vehi cle networks √ × × √ × √ √ × Current V2X system √ × √ × × √ √ √ Sharing bicy cle system × √ × × √ × √ × Our proposal √ √ √ √ √ √ √ √ Fig. 4. The power consumption and system EE performance in dif ferent scenari os while transmitti ng 2G data within 1 second, carrier bandwidth 20MHz, by further considering the machine room (400 W), erbium doped fibre amplifier (EDF A, 243 W ) and other core network elements. In upper two figures, caching in the core router and without cachi ng is compared, where the bott om figures are system po wer consumpti on and EE performance s of cachi ng in the wirele ss secti on. the emergency inform ation and guide the veh icle (speed, accelerate, rou te, e tc. ) on the ro a d. By sharing th e informatio n amongst coop erating vehicles, the auto matics piloting system can be accomp lished even without the help of center . Addi- tionally , by ad opting the 5G an d CCSM, th e pr oposed network can provide mo re effi cient high speed wireless transmissions. Clean energy and more reasonable system model: The V2X devices can use solar energy fo r co m munication , and store it in their batteries. W ith the c lea n solar energy , the distributed V2X networks can r educe the car bon emission an d air po llution. Addition ally , co m pared to prior cache en abled studies, e.g., [7] , solar energy’ s usage of the d istributed V2 X networks br ings hig her en ergy c a pacity to these V2X d evices, which makes CCSM a more reasonable ch oice for V2X commun ications. For instance, with mor e energy , the V2X devices can increase its tran sm ission power f or faster speed transmission or prolon g the tran smission time. Reduced energy consumption a nd enhanced EE per - formance: More energy ca n be sav e d while ob taining the request in formation f rom ca c h es d u e to the sho rter tran s- mission distance and le ss engaged transceiver devices. Th e system EE p erform ance is en hanced as well v ia this meth o d. As shown b y Fig . 4, con tent ca c hing can greatly redu ce the energy consumptio n o f the distributed c ache en a bled V2X networks, especially by cachin g in the wireless section (BS, vehicle, device, etc.). This is mainly du e to th e red u ced core network energy co nsumption . It is worth to note that giving the ba ttery constrain t, caching in th e wireless section is only feasible when they ha ve en ough power . Usage of mixed frequency resources: The D SRC 5 .9 GHz frequen cy as well as mmW a ve freq uency are dedicated for the wireless comm unications with V2V , V2 P , V2I, V2N, SC2V , SC2P and P2 P in the urb an environment. This is mainly due to the high atmospheric atten uation and easily blocked features o f higher freque ncy . By inv ok in g DSRC and mmW ave frequen cies, the user can emp loy a wider ca r rier b andwidth for its large volume inform ation tra nsmission within a limited power u nder the assistance o f conten t caching techno logy . In contract, to reuse th e existing resources, L TE fre quency is in voked fo r the BS to vehic le (BS2V) and BS to p edestrian (BS2P) transmissions. I V . F U T U R E R E S E A R C H T R E N D S A N D C H A L L E N G I N G I S S U E S In th is section, we d iscu ss the futur e researc h tren ds and challengin g issues to acco m plish the proposed cache en- abled V2X networks. Meth ods on ach ieving th e h igh speed transmission, flexible and dedicated network slicing serv ice, big data and ma c hine learnin g b ased imag e recogn ition and reconstruc tion, are main resear ch topics for future research . Additionally , the accu rate urban chann el and pr opagatio n model, ultra reliab le com munication and low laten cy are b ig challenges of the distributed cache enabled V2X networks’ implementatio n. A. Fu tur e Researc h T r end s 1) S olutions for More Than 10 F olds F aster T ransmission Rate: In literature, the massive MIMO is raised as an essential element of 5G. It is proved that with an tenna n umber growing, we can ob tain more channel degree of f reedom (Do F), wh ic h yields faster transmission rate and link r eliability . On the other hand, m mW ave recently is emergin g as a vital element of 5G. Accord ing to Shan non th eory , th e achiev ab le transmission rate can be boosted u p while increasing the car rier frequency bandwidth value. Rece n tly , the n o n-orth ogona l multiple access (NOMA) is also in te n si vely stud ied with regard to 5G’ s sp e c- trum efficiency (SE). It alloca tes th e same car rier f requen cy resource for multiple user’ s informatio n transmission, whereas the encode an d de code procedu r es are executed acco rding to the allocated different power values (power domain NOMA) or codes (code domain NOMA) [8]. Meanwhile, the HetNets technolog y with co-existing macro cells and small c e lls are propo sed. In HetNets, m acro cell 5 are used to p rovide wide coverage area, whereas in the cell edge areas, small cells are utilized to improve the connec- tion an d transmission quality . The car rier ag gregation (CA) and co ordinate multiple p oint transmission ( CoMP) based cloud radio access network ( C-RAN) can further lev erage the transmission rate. Ad d itionally , as talked befo r e, co ntent caching based n etwork evolution is ano th er in teresting topic which attracts inc r easing attentions from both industry and academia. Howev er , accor ding to 5G Summ it in Silicon V alley , scholars claim ed that th e 5G’ s higher transmission rate can n ot be simply ach iev ed with existing techno logies up to n ow , redesignin g the whole network architec tu re, c ombining the existing techn ologies, working o n new air radio technolo g ies are comprehen si vely needed with a joint fo rce fro m both academia and industry . 2) Th e Dedicated Network Slicing Service: In pr ior wire- less gener ations, f or application scene with fast transmission rate, wide coverage area , ultr a reliab le commun ications, low latency communic a tions, mo stly a specialized network ar- chitecture sho uld be established. Th ose specialized network architecture , once established , is hard to accommo date up- dates. Meanwh ile, with large num b er o f accessing vehicles and devices, the m anagemen t of such a large scale network becomes troublesom e [9 ]. T o co pe with the div ersity , dedicated network slicing technolog y [10] can be a feasible choice. Network slicing te c hnolog y is able to vir tu ally divide the network into mu ltiple co-existing sub-networks. It ca n adap- ti vely assign the needed r esources and establish the optim al router for each sub -network ( n etwork slicing service). Since the d ivisions and resource allocations are based o n th e sof t- ware defined deep ly prog ramming , it can reuse the resources and u pdate th e existing sub -network(s) or establish a n ew dedicated sub-n e twork once needed . W ith software defin ed network (SDN) con troller and orch estration, we can co n trol the connected vehicles a n d devices by cr eating a co mprehe n si ve network slice. In ad dition, it is also possible to set u p d ifferent dedicated sub-networks for the h igh speed data transmission and intelligen t traffic control scen a rios, and adaptively adju st the alloc a ted reso u rces accord ing to the update inf o rmation. How to effectively tailor the dedicated n etwork slice with giv en resour ces and flexibly adjust the sub-network to cater to the u pdates with network slicing technolo gy will be an interesting topic for futur e study . 3) B ig Data, Machine Learnin g and F og Computing Based Image Recognition and Recon struction: As known, automatics pilot is calling f or real time u ltra reliable image recog nition and r econstructio n. Du e to th e large scale a nd even fast g row- ing data, previous iterative ML alg o rithm is no t an effecti ve way due to the space a n d time limitations. T o accelerate the recogn itio n spe e d, large scale d e ep learn ing (DL) and high- rank matrix factorizatio n (MF) metho ds are p r oposed with massi ve parameters for imp roved filtering pr oblems [11]. Y et the pr ocessing time a n d req uired resource s a r e still hug e. This is mainly due to the fact that while detectin g the objectives, we need a set of labeled train ing d ata fro m a large scale data volume. Th is is the most time consuming process that called as the groun d tru th labeling [12]. On the contrary , if less parameters are considered , some elements on the road Fig. 5. An explanat ory descripti on of label ing data in image detect ion. T he sharing bicycl es and ordinary bicyc les are all labele d as bicycl e. Additionall y , the chair is not labeled (mark ed with green arro w), which might bring risk to the vehicl e while applyin g in automati c piloting for image dete ction in the follo wing frames. might be omitted, wh ich will bring in r isk for the traffic and public safety . For in stance, wh ile labeling th e data in Fig . 5, the chair is not lab e le d (m arked with green arrow , next to the first tree counting from the bottom left). For th e image detection in the following f rames b a sed on the lab e led data with Fig. 5, once the objective (u nlabeled chair) was loca ted on the motor-way , an accid e n t might be c a used. In add ition, the sha r ing bicycle and ord inary bicycle a re all recog nized as the b icycle with same label to accelerate the recog nition speed in the following steps. Fortunately , Matlab h as released the automated driving system toolbo x to accelerate th e labe lin g process, in a d dition, the Kan a de Lucas T omasi algo rithm (KL T ) can b e inv oked to lab el the objects in the first frame, a n d track them in the following fra m es [12]. Th e specialized GPU (fo r instance, NVIDIA a T esla K40 c ) can be used to speed up th e trainin g process as well. Howev er , with data scale growing, it is still a challenging and tim e-consum ing task. T o this end, a trade-off strategy between the enga ged parameters ( or labels) and real-time recog nition a nd reconstru c tio n req uiremen t is needed. Add itionally , a joint f orce of the off-the-shelf in-car cooper a tive edge compu ting chips and info r mation sharin g strategies am ongst vehicles is n eeded. The developments on ef- fective big data analy sis and DL based recog n ition algorithm s, cooper a tive ed ge and cloud compu ting strategies, specia lize d processing chips are co m prehen si vely needed with a f ull steam ahead for the forthco ming V2X system. B. The Challen ging Issues 1) Th e P ath Lo ss and Chann el Mo del: Th e p ath loss and channel mod el of high e r carrier frequ e ncies with DSCR and mmW ave are vital for cellular coverage and per forman ce es- timation, as well as the pub lic and traffic safe ty . T he enriched buildings around, denser BS dep loyment and e ven higher carrier frequencies are all challenging issues of the rea listic channel and path loss models. In prior studies from UT - Austin, the Manhattan poisson line pr ocess was in troduce d. They assumed the vertical ( N o rth-Sou th direction) and hor izon 6 (East-W est directio n) paths growing infinity with y an d x - axis, b ased on the street canyo ns. Howe ver, in the non light of sight (N L OS) pa th, th e cor ner loss was simplified with a constant factor . T he ITU-R, recently , published th e urban path loss model for the frequ ency r a nges fr om 3 00 MHz to 100 GHz. Howe ver, it is too complex esp ecially for the mmW ave frequen cy . T h e ligh t o f sight (L OS) and NLOS path losses were compreh e n si vely studied with a remed y study in [ 13], the pro posed models were verified in Seou l City . W e inv oke this work here with u rban street configuratio n (street block 100 m, BS h eight 40 m, vehicle and device heig ht 1 m ) , the results ar e given b y Fig. 6. As shown, NL OS p ath loss becomes extremely large after 1-turn wh ile cr o ssing the corner . Thus nor mally , only 1-tu rn NLOS path can be established in urban stree t ca nyo n environment. On the other hand, to the mmW ave path loss m odel, ser ies trials were don e by the New Y or k Univ ersity in the Ne w Y or k urban street. Y et up to n ow , no matured path loss or chann el mo dels ha ve b een arrived. Fig. 6. T he urban path loss s imulation according to [13] with BS height 40 m, vehi cle and de vice height 1 m, street block distance 100 m , carrie r frequenc y 2 GHz. 2) Th e Low Latenc y Commu nication: In the distributed cache en abled V2X networks, less than 1 m s laten cy is a vital issue for automatic pilot, especially while encoun tering som e emergency con ditions. For instanc e, giving a vehicle speed 120 Km/h, with L TE latency 30 m s, the moving distan ce will be around 9 dm. On the contr ary , with 5G’ s 1 ms latency , it will be around 3 cm. Moreover, th e low latency is a natu re feature for real time in- car entertain m ent. As known, the ultr a dense cellular d eployment will r esult in a limited coverage area, low latency will guaran tee the fast respon se for each req uest. The handover time across cellular can be redu ced as well via low latency . The CoMP joint transm ission is c alling fo r low latency processing and precise time synchro nization as well. Albeit 1 m s latency issue h as been app ealed a lo t, in literature, specific tech nology to achieve the less than 1 ms latency is still less. In our distributed cache enab led V2 X networks, althou gh the lo w latency req uiremen t can be par- tially realized via D2D comm unication s, and the NOMA [8] transmission can be in voked f o r simu ltaneous tran smission of multiple user info rmation, it is still n ot enou g h. Genera lly , to achieve this goal, a dedicated network slicing service should b e assigned. The dedicated short packag e emergency informa tio n can be used for fast respon se to further improve its latency p erform a n ce. De veloping on the chip s with r eal time processing ability is also n eeded. T o sum up, th e low latency goal can not be sim p ly ach iev ed, a comp rehensive scope o n cr o ss-layer design and updates from bo th ha r dware and software sides are needed. 3) Th e Ultr a Reliable Communication Issue: The ultra re- liable com munication is an other challengin g issue. The higher transmission speed will be o f less m eaning without small error prob ability . For instan c e , we prefer 100 MBit/s 9 9% reliability rather than 1 GBit/s 50% reliability while accessing the inter n et. Ultra reliable co mmunica tion is critical for vehicle safety and public safety as well. The vehicle cann ot be safely braked without it while en counter in g some em ergencies even the pro c essing time is within 1 ms. Additiona lly , it will cause great damage once the comm u nication link is d amaged o r hacked, in whic h ca se the intelligence control center and automatics pilot cannot successfully manage the traffic. In literature , the ultra re liable commu nication is not a new problem . For example, intensiv e studies have been d one in the core ne twork with backup rou ting an d link con nections across core routers to ensu r e the ro bust connections. For the wireless co m munication , re-tran smission can improve th e transmission succ e ss rate and the completene ss of th e re c e i ved informa tio n. Howev e r, th e massive c o nnected devices bring in new ch allenging issues for the u ltra reliable comm unication, e.g., hacker attack, resou rce c o mpeting , eq uipment failure, un- controllab le interference, missing protoco l, etc. Additionally , the tr a de o ff strategy should be set forth while inv oking the re-transmission metho d with regard to the ultra re liab le and low latency requ irements. T o this end , a lot of work is still needed in future studies. V . C O N C L U S I O N The distributed cache ena b led V2X networks ar e intro duced in this ar ticle . The b asic pr in ciples, merits, futur e r esearch trends, and po tential ch allenging issues are discu ssed. Com- pared with prior stud ies, this proposal can offer more in telli- gent tr affic control while con necting all vehicles and devices on the road. The wireless co mmunicatio ns are leveraged as well by this p roposed system with less en gaged transceiver compon ents, be tter system EE perfo r mance and more effective carrier freque n cy resource usage. More en deav o rs o n the solu- tions for the ev en higher transm ission r ate requirem ent, flexible network slicing serv ice d esign and big data/ML /fog c o mputing image reco gnition and reconstru ction methods are need ed in future studies. Meanwh ile, the acc u rate urban chan nel/path loss mode l, u ltra reliab le c ommun ications and low latency issues are critical factors for its realization. R E F E R E N C E S [1] D. Zhang et al. , “One inte grated energ y effic iency proposal for 5G IoT communicat ions, ” IEEE Internet T hings J. , vol. 3, no. 6, pp. 1346–1354, Dec. 2016. [2] Z . Zhou et al. , “Unlicensed spectrum sharing: From coexiste nce to con ver gence, ” IE EE W ireless Commun. , vol. 24, no. 5, pp. 94–101, Oct. 2017. 7 [3] C. Wu, Z . L iu, D. Zhang, T . Y oshina ka, and Y . Ji, “Spatial intel ligenc e to wards trustwor thy vehi cular IoT (in press), ” IEEE Commun. Mag. , 2018. [4] N. Lu et al. , “Connected v ehicle s : Solutions and challenges, ” IEEE Internet Things J. , vol. 1, no. 4, pp. 289–299, Aug. 2014. [5] Huawei, “The race is on: NB-IoT boosts smart bik e sharing for ofo [online :] http:/ /www .huawei . com/en/c ases/ofo, ” 2017. [6] S. H. Ahmed et al. , “Named data netw orking for software defined vehi cular netwo rks, ” IEEE Commun. Mag . , vol. 55, no. 8, pp. 60–66, Aug. 2017. [7] B. Chen, C. Y ang, and A. F . Molisch, “Cac he-enab led devi ce-to-d e vice communicat ions: Of floading gain and ener gy cost, ” IEEE T rans. W irel ess Commun. , vol . 16, no. 7, pp. 4519–4536, Jul. 2017. [8] D. Zhang et al. , “Performance analysis of non-regenera ti ve Massive - MIMO-NOMA relay systems for 5G, ” IE EE T rans. Commun. , vol . 65, no. 11, pp. 4777–4790, Nov . 2017. [9] J. Huang et al. , “T opology control for buil ding a large-scal e and energy- ef ficient intern et of things, ” IEEE W irele ss Commun. , vol. 24, no. 1, pp. 67–73, Feb . 2017. [10] T . T aleb et al. , “Permit: Network slicing for personal ized 5G mobile telec ommunicati ons, ” IEEE Commun. Mag . , vo l. 55, no. 5, pp. 88–93, May 2017. [11] E. P . Xing et al. , “Petuum: A ne w platform for distribut ed machine learni ng on big data, ” IEEE T rans. Big Data , vol. 1, no. 2, pp. 49–67, Jun. 2015. [12] A. Nehemiah and A. Jayaraman, “Deep learning for automated dri v- ing with m atlab [online:] https://de vblogs.n vidia.com/parallelforall/deep- learni ng-automat ed-driving-matlab/, ” NVIDIA, T echnica l blogs, Jul. 2017. [13] J. L ee et al. , “NLOS path loss model for lo w-height antenn a links in high-rise urban street grid en vironment s, ” Internat ional J . Anten nas Pr opag . , vol . 2015, pp. 1–9, Jan. 2015.