Lte for vehicular networking

1,125 views
999 views

Published on

Published in: Technology, Business
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
1,125
On SlideShare
0
From Embeds
0
Number of Embeds
118
Actions
Shares
0
Downloads
0
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Lte for vehicular networking

  1. 1. IEEE Communications Magazine • May 2013148 0163-6804/13/$25.00 © 2013 IEEEINTRODUCTIONEnabling wireless connectivity on wheels is theaim of several players, driven by the social andeconomic benefits expected from intelligent trans-portation systems (ITS) applications, supportingroad safety and traffic efficiency through vehicle-to-vehicle (V2V) and vehicle-to-infrastructure(V2I) communications. Safety applications rely onshort-message broadcasting in a vehicle’s neigh-borhood to reduce fatalities on the road; trafficefficiency applications need the support of road-side units (RSUs) with communication capabili-ties to send periodic updates to remote trafficcontrol centers. These applications exhibit someunique features, in terms of generation patterns,delivery requirements, communication primitives,and spatial and temporal scope, which challengeexisting wireless networking solutions.IEEE 802.11p [1] is the standard that sup-ports ITS applications in vehicular ad hoc net-works (VANETs). Easy deployment, low cost,mature technology, and the capability to nativelysupport V2V communications in ad hoc modeare among its advantages. Nonetheless, this tech-nology suffers from scalability issues, unboundeddelays, and lack of deterministic quality of ser-vice (QoS) guarantees [2]. Furthermore, due toits limited radio range and without a pervasiveroadside communication infrastructure, 802.11pcan only offer intermittent and short-lived V2Iconnectivity. The above mentioned concernsmotivate the recent increasing interest in LongTerm Evolution (LTE) [3] as a potential accesstechnology to support communications in vehic-ular environments.LTE is the most promising wireless broadbandtechnology that provides high data rate and lowlatency to mobile users. Like all cellular systems,it can benefit from a large coverage area, highpenetration rate, and high-speed terminal sup-port. Extending its use to also support vehicularapplications would open new market opportuni-ties to telco operators and service providers.Vehicles are indeed the third place, after homesand offices, where citizens spend more time daily.According to the U.S. Department of Transporta-tion and Safety Administration, commuters spend500 million hours per week in the car. Alcatel-Lucent’s 2009 study showed that over 50 percentof interviewed consumers found the idea of a con-nected car highly appealing, and 22 percent wouldbe willing to pay $30–65 per month for value-added connectivity services while on the road.Indeed, LTE particularly fits the high-bandwidthdemands and QoS-sensitive requirements of a cate-gory of vehicular applications known as infotainment(information and entertainment), which includes tra-ditional and emerging Internet applications (e.g., con-tent download, media streaming, VoIP, webbrowsing, social networking, blog uploading, gam-ing, cloud access). In any case, its capability to sup-port applications specifically conceived for thevehicular environment to provide road safety andtraffic efficiency services is still an open issue.The main concern comes from the centralizedLTE architecture: communications always crossinfrastructure nodes, even though all that isrequired is a localized V2V data exchange, as forsafety-critical applications, with negative conse-quences on message latency. In addition, in densetraffic areas, the heavy load generated by periodicmessage transmissions from several vehiclesstrongly challenges LTE capacity and potentiallypenalizes the delivery of traditional applications.These topics are under investigation by spe-cialized groups of standardization and govern-ment bodies. The European TelecommunicationsStandards Institute (ETSI), the InternationalStandards Organization (ISO), and the U.S.Department of Transportation (DOT) are cur-rently investigating the complementary roles ofIEEE 802.11p, LTE, and other cellular technolo-gies in supporting cooperative ITS applications[4–6]. Early works evaluating LTE’s effectivenessABSTRACTA wide variety of applications for road safetyand traffic efficiency are intended to answer theurgent call for smarter, greener, and safer mobil-ity. Although IEEE 802.11p is considered the defacto standard for on-the-road communications,stakeholders have recently started to investigatethe usability of LTE to support vehicular appli-cations. In this article, related work and runningstandardization activities are scanned and criti-cally discussed; strengths and weaknesses of LTEas an enabling technology for vehicular commu-nications are analyzed; and open issues and criti-cal design choices are highlighted to serve asguidelines for future research in this hot topic.TOPICS IN AUTOMOTIVE NETWORKINGAND APPLICATIONSGiuseppe Araniti, Claudia Campolo, Massimo Condoluci, Antonio Iera, and Antonella Molinaro,University Mediterranea of Reggio CalabriaLTE for Vehicular Networking: A SurveyCAMPOLO2 LAYOUT_Layout 1 4/29/13 1:10 PM Page 148
  2. 2. IEEE Communications Magazine • May 2013 149for communications involving vehicles can alsobe found in the scientific literature [7–15].The objective of this article is to provide acritical assessment of the LTE capabilities tosupport the unique set of vehicular applications.First, the state of the art in LTE use for vehicu-lar environments is inferred from the scientificliterature and standard documents. Then openchallenges are discussed, and predictions aboutthe possible role of LTE in providing services onthe road are formulated. To the best of theauthors’ knowledge, this is the first work thatanalyzes the cited topic in a systematic way.VEHICULAR APPLICATIONS ANDENABLING TECHNOLOGIESBesides infotainment, a set of unique applica-tions have been conceived for users on wheelsand classified based on their targets as activeroad safety and traffic efficiency.Active road safety applications aim at reducingthe risk of car accidents, and have timeliness andreliability as the major requirements. Two maintypes of safety messages have been standardized,the transmissions of which can be periodic orevent-triggered. In ETSI documents [16] they arerespectively referred to as cooperative awarenessmessages (CAMs) and decentralized environmen-tal notification messages (DENMs); basic safetymessages (BSMs) are the terminology used in [6]for both periodic and event-triggered messages.CAMs (a.k.a. beacons or heartbeat messages)are short messages periodically broadcast fromeach vehicle to its neighbors to provide informa-tion of presence, position, kinematics, and basicstatus. DENMs are event-triggered short mes-sages broadcast to alert road users of a haz-ardous event. The main requirements of CAMsand DENMs are reported in Table 1, togetherwith the relevant use cases identified by ETSI.Both CAM and DENM messages are deliv-ered to vehicles in a particular geographic region:the immediate neighborhood (awareness range)for CAMs, and the area (relevance area) poten-tially affected by the notified event (congestion,hazard warning, etc.), even spanning several hun-dred meters, for DENMs. The capability of trans-mitting a message to nodes satisfying a set ofgeographical criteria is called geocast and repre-sents, together with reliability and low-latencydelivery, a crucial requirement of the typical tem-poral- and spatial-relevant vehicular applications.Traffic efficiency applications aim to optimizeflows of vehicles by reducing travel time and traf-fic congestion. These applications have no strictdelay and reliability requirements, but their qual-ity gracefully degrades with increases in packetloss and delay. In this class, the decentralizedfloating car data (FCD) [17] service requiresperiodic transmissions of information collectedby vehicles, from internal and external sensors(e.g., CAN bus, in-vehicle camera, environmentalmonitoring sensors), to remote managementservers. They process collected data, monitor andpredict traffic congestion, and send up-to-datetraffic information back to the vehicle’s naviga-tion system by also suggesting alternative routes.Several wireless technologies have been ana-lyzed as candidates to support the mentionedapplications through V2V and V2I communica-tions. In agreement with the CommunicationsAccess for Land Mobiles (CALM) guidelines, theITS station reference architecture proposed inETSI specifications [16] leverages the complemen-tary strengths of distributed short-range networks(e.g., IEEE 802.11p and its European counterpartITS-G5, Wi-Fi) and centralized cellular technolo-gies, among which LTE is the most promising one.Early signs of this trend toward heterogeneousnetworking in the complex vehicular environmentcan be found in the United States as well [6].The main candidate access technologies havedifferent characteristics and can match the vehic-ular applications’ requirements, more or lesseffectively, as summarized in Table 2 and dis-cussed in the next sections.Table 1. Safety message requirements and use cases.Cooperative awarenessmessage (CAM)Periodic time-triggeredposition messages–Frequency: 1–10 Hz–Max latency: 100 ms–Length: up to 800 bytes (securi-ty overhead included) dependingon the type of applicationUse cases–Emergency vehicle warning–Slow vehicle indication–Intersection collision warning–Motorcycle approaching indication–Collision risk warning–Speed limits notification–Traffic light optimal speed advisoryDecentralized environ-mental notificationmessage (DENM)Event-driven hazard warnings–Max latency: 100 ms–Length: typically shorter thanCAMsUse cases:–Emergency electronic brake light–Wrong way driving warning–Stationary vehicle accident–Stationary vehicle-vehicle problem–Traffic condition warning–Signal violation warning–Road-work warning–Collision risk warning–Hazardous location–Precipitation, wind–Road adhesion–VisibilityActive road safetyapplications aim atreducing the risk ofcar accidents, andhave timeliness andreliability as themajor requirements.Two main types ofsafety messages havebeen standardized,the transmissions ofwhich can beperiodic or event-triggered.CAMPOLO2 LAYOUT_Layout 1 4/29/13 1:10 PM Page 149
  3. 3. IEEE Communications Magazine • May 2013150LTE IN A NUTSHELLLTE represents the new generation of mobileradio networks defined by the Third GenerationPartnership Project (3GPP) [3].The LTE system is characterized by a flat all-IParchitecture (Fig. 1) with a reduced number of net-work devices. IP-based data, voice, and signalingtransmissions allow for greater deployment feasibil-ity and extendibility with respect to previous cellularnetworks. Thanks to its simplified architecture,LTE can provide a round-trip time theoreticallylower than 10 ms, and transfer latency in the radioaccess up to 100 ms. This is especially beneficial fordelay-sensitive vehicular applications. The accessnetwork is composed of eNodeBs, which manageradio resources and handover events, whereas thecore network is composed of three main units: themobility management entity (MME), responsiblefor control procedures, such as authentication andsecurity, and storing of users’ position information;the S-GW, responsible for routing, data forwarding,and charging by coupling with the policy and charg-ing rules function (PCRF); and the P-GW, the out-going entity that allows communication with IP andcircuit-switched networks.The LTE air interface has the flexibility tosupport time-division duplexing (TDD), frequen-cy-division duplexing (FDD), and half-duplexFDD schemes; it also provides scalable channelwidth (1.4–20 MHz). Concerning the access tech-nology, orthogonal frequency-division multipleaccess (OFDMA) is used in the downlink andsingle-carrier frequency-division multiple access(SC-FDMA) in the uplink, both providing highflexibility in the frequency-domain scheduling.Multiple-input multiple-output (MIMO) tech-niques improve the spectral efficiency by a factorof 3 to 4 compared to 3.5 generation (3.5G) sys-tems even at high terminal speeds, making LTEparticularly efficient in challenging and dynamicpropagation environments like the vehicular one.Radio resources are centrally managed by aneNodeB at every transmission time interval (1 msduration), with the aim of satisfying QoS con-straints while increasing channel utilization. Akey role is played by the packet scheduler at theeNodeB. It selects the traffic flow to serve, basedon the related QoS requirements (as specified bythe QoS class identifier, QCI), and decides theappropriate modulation and coding schemebased on feedback from the mobile terminals onthe channel quality. QCI refers to a set of packetforwarding treatments, for example, resourcetype (guaranteed or not guaranteed bit rate), pri-ority, packet loss rate, and delay budget.LTE also supports high-quality multicast andbroadcast transmissions through the evolvedmultimedia broadcast multicast service(eMBMS) [18, 19] in the core and in the radioaccess network. It offers the possibility of send-ing the data only once to a set of users registeredto the offered service, instead of sending it toevery node separately.The standardization of LTE-Advanced (LTE-A) is ongoing in 3GPP (Rel. 11) as a majorenhancement of LTE in terms of bit rate, capac-ity, and spectral efficiency, mainly through theFigure 1. LTE architecture: access network (eUTRAN) and core network (EPC) entities.S1-UX2S1-MMEMMEMobilitymanagemententityS-GWServinggatewayP-GWPacket datanetwork gatewayHSSHome subscriberserverPCRFPolicy and chargingrules functionsE-UTRANEPCeNBeNBPCRFHSSP-GWMMEInternetControl planeUser planeS-GWTraffic efficiencyapplications aim tooptimize flows ofvehicles by reducingtravel time and traf-fic congestion. Theseapplications have nostrict delay and relia-bility requirements,but their qualitygracefully degradeswith increasesin packet lossand delay.CAMPOLO2 LAYOUT_Layout 1 4/29/13 1:10 PM Page 150
  4. 4. IEEE Communications Magazine • May 2013 151support of advanced MIMO techniques, carrieraggregation, and relay nodes. With LTE-A stillin an early stage, the focus of the related workreported in this article is on LTE, but LTE-Apotentialities are discussed.LTE AS A SOLUTION TO SUPPORTVEHICULAR APPLICATIONS:MOTIVATIONS AND CONCERNSThere are several reasons for LTE applicabilityin vehicular environments; the major issues arediscussed in the following.Coverage and mobility: LTE will rely on acapillary deployment of eNodeBs organized in acellular network infrastructure offering wide areacoverage. This would solve the 802.11p issue ofpoor, intermittent, and short-lived connectivity,and would particularly indicate LTE for V2Icommunications even at high node speeds. TheLTE infrastructure exploitation would also repre-sent a viable solution to bridge the network frag-mentation and extend the connectivity in thosescenarios where direct V2V communications can-not be supported due to low car density (off-peakhours, rural scenarios, etc.) or to challengingpropagation conditions (e.g., corner effect due tobuilding obstructions at road intersections).Market penetration: A higher penetrationrate is expected to be achieved by LTE com-pared to 802.11p. The LTE network interfacewill be integrated in common user devices likesmart phones, so passengers would be accus-tomed to being connected to the Internetthrough these devices while on the road as well.Capacity: LTE offers high downlink anduplink capacity (up to 300 and 75 Mb/s, respec-tively, in Rel. 8, and up to 1 Gb/s for LTE-A inRel. 11), which can potentially support severalvehicles per cell. Such values are higher than802.11p, which offers a data rate up to 27 Mb/s.On the other hand, some critical issues alsoraise concerns about the applicability of LTE forsupporting the reference vehicular applicationscenarios.Centralized architecture: The main concerncomes from LTE’s centralized architecture,which would not natively support V2V commu-nications, but would instead require passingthrough infrastructure nodes in the core networkthat should intercept uplink traffic before redis-tributing it to the concerned vehicles. A myopicmessage broadcasting over the entire cell mayreach vehicles that are not interested (e.g., carsmoving in the opposite direction to the roadhazard advertised by a DENM, or vehicles out-side the awareness range of a CAM). Therefore,specialized network entities (e.g., back-endservers) and other core network elements shouldbe involved and wise policies designed for coop-erative ITS messages dissemination.Channels and transport modes: The down-link transport mode (unicast or broadcast) andthe selected uplink/downlink transport channelsTable 2. Main candidate wireless technologies for on-the-road communications.Feature Wi-Fi 802.11p UMTS LTE LTE-AChannel width 20 MHz 10 MHz 5 MHz1.4, 3, 5, 10, 15, 20MHzUp to 100 MHzFrequency band(s) 2.4 GHz, 5.2 GHz 5.86–5.92 GHz 700–2600 MHz 700–2690 MHz 450 MHz–4.99 GHzBit rate 6-54 Mb/s 3–27 Mb/s 2 Mb/s Up to 300 Mb/s Up to 1 Gb/sRange Up to 100 m Up to 1 km Up to 10 km Up to 30 km Up to 30 kmCapacity Medium Medium Low High Very HighCoverage Intermittent Intermittent Ubiquitous Ubiquitous UbiquitousMobility support Low Medium HighVery high (up to350 km/h)Very high (up to350 km/h)QoS supportEnhancedDistributed ChannelAccess (EDCA)EnhancedDistributed ChannelAccess (EDCA)QoS classes andbearer selectionQCI and bearerselectionQCI and bearerselectionBroadcast/multicastsupportNative broadcast Native broadcast Through MBMS Through eMBMS Through eMBMSV2I support Yes Yes Yes Yes YesV2V support Native (ad hoc) Native (ad hoc) No NoPotentially, throughD2DMarket penetration High Low High Potentially high Potentially highCAMPOLO2 LAYOUT_Layout 1 4/29/13 1:10 PM Page 151
  5. 5. IEEE Communications Magazine • May 2013152(e.g., dedicated or common channel) have aneffect on the delay and capacity in terms of themaximum number of vehicles per cell.Status mode of the device: Latency is alsoinfluenced by the status mode of the mobile termi-nal. In order to save resources, cellular networksare configured to keep non-active terminals in idlemode, but the connection setup necessary toswitch to connected mode before sending data maytake a longer time than the simple transmissiondelay.1 Vehicles should be in connected mode tosend periodic CAMs, whereas the transmission ofan event-triggered DENM could require a vehicleto switch from idle to connected mode.LTE APPLICABILITY TO VEHICULARSCENARIOS: PRELIMINARY STUDIESThe above mentioned critical issues motivatedthe recent investigation in the scientific litera-ture and standardization groups in order toassess the applicability of LTE to support vehic-ular applications. The related work surveyed inthis section focuses on safety and traffic efficien-cy applications that have mainly attracted theinterest of the researchers in the field.SAFETY TRAFFIC: THE CAM ANDDENM CASESSafety applications require periodic V2V dataexchanges in a vehicle’s neighborhood (this is thecase of CAMs) or event-triggered V2V and V2Icommunications (this is the case of DENMs).ETSI and ISO are currently investigating LTE’sability to support these cooperative applications;preliminary results are reported in [4].CAM and DENM exchanges in LTE involvetransmissions from vehicles to infrastructurenodes, and successive traffic distribution to theconcerned vehicles.Regarding the transportation modes, unicast isalways used for uplink transmission, while unicastand broadcast modes can be used on the down-link by leveraging MBMS capabilities. In theuplink case, the problem is to select the mostappropriate channel type without congestionrisks. The random access channel (RACH) is acommon uplink transport channel usually selectedfor signaling and to transmit small data amounts,such as CAMs and DENMs. In the downlinkcase, broadcast mode is more resource-efficientthan unicast mode, although it could imply longerdelays due to the MBMS session setup.In both cases, ETSI specifications foresee thepresence of a special-purpose back-end serverthat supports geocasting, by intercepting trafficfrom vehicles and processing it before redis-tributing it only to the concerned vehicle(s) in agiven geographical area [4].In order to identify the concerned vehicles in agiven area and act as a reflector [7], the back-endserver has to know the list of geographical areas,their coordinates, the cars in any area at all times,and their IP addresses and positions. Accordingto the ETSI specifications [4], each time vehiclescross over to a new area, the server informs themabout the coordinates of their current geographi-cal area. The area size can vary from applicationto application, thus affecting the signaling over-head. Then, whatever the server location, data isdistributed to concerned vehicles through MBMSor via multiple unicast connections.Different approaches to server deploymenthave dissimilar impacts on the signaling proce-dures, as discussed in [7]. If the server is installedin the mobile operator’s core network, then itmay exchange location information with theMME module in the LTE architecture in Fig. 1,which regularly receives location updates fromthe connected vehicles. If the server is located inthe Internet and, thus, decoupled from themobile operator network, each vehicle maintainsa direct connection to the server and regularlysends position updates to it.Figure 2 reports the example of DENM dis-tribution procedures augmented with the back-end server. In the case of unicast distribution(left), vehicles are addressed individually, so thatthe same message is separately transmitted to allconcerned vehicles. In the broadcast/multicastcase (right), all vehicles in the relevance area arecollectively addressed, through geo-addressingcapabilities leveraging the geographical positionof nodes, and a message transmission is per-formed once by relying on MBMS features(dashed lines in Fig. 2). In both cases, latency1 Tests performed on 3.5Gnetworks have demon-strated that idle-to-con-nected mode switchingrequires 2 to 2.5 s, whichis intolerable for vehicularapplications. Lower delayvalues are expected forLTE, although still notinvestigated in compre-hensive testbeds [3].Figure 2. Unicast (left) and multicast (right) DENM delivery in LTE. Only vehicles within the relevance area (rectangular red area) areaddressed.D!eNBCell rangeBackend serverCore networkinfrastructureRelevance area!BCFEEAABCABCD!eNBCell rangeBackend serverCore networkinfrastructureRelevance area!BCFEEA {A, B, C}{A, B, C}CAMPOLO2 LAYOUT_Layout 1 4/29/13 1:10 PM Page 152
  6. 6. IEEE Communications Magazine • May 2013 153could become an issue, especially for localizedsafety-critical V2V communications.Even in the case of CAMs, messages have tocross the infrastructure for multicast distribution.In Fig. 3 the back-end server collectively address-es all vehicles in the awareness range of the send-ing vehicles (A and B). On the contrary, in Fig. 4,when an IEEE 802.11p network is available, asingle broadcast transmission can be used to dis-tribute the message from a vehicle in its aware-ness range (in the case of CAMs) or within therelevance area (in the case of DENMs).CAMSThe main challenge in supporting CAMs is toavoid system overload due to the heavy trafficbroadcasted by a high number of vehicles fre-quently (typically every 100 ms). This is especiallycritical in dense areas, like city centers, or duringpeak hours.Analytical results in [9] show that LTE isunable to satisfy the CAM delivery requirementswhen the eNodeB retransmits all received CAMsto every vehicle in the cell in unicast mode. Simi-lar results are achieved when the eNodeB uni-casts CAMs to every vehicle in the one-hopneighborhood. Improvements can be obtainedthrough CAM broadcasting in the cell.In [10] the authors enhance the unicast CAMdownlink transport with a filtering scheme inorder to reduce the load and meet the CAMdelay requirements. Filtering relies on the ideathat not all vehicles in a cell need to receive allCAMs. Accordingly, based on the received vehi-cles’ location information, the back-end serverselects a subset of vehicles that receive CAMs onunicast links. Results attained in urban and ruralscenarios show that the selective unicast of aggre-gated CAMs may also easily overload LTE, and ahigher number of vehicles per cell can be sup-ported when decreasing the CAM rate down to 2packets/s. The conducted study contributed tothe architecture and the results reported in [4].The authors finally suggest the use of MBMSas a means to increase the downlink capacity, asalso claimed in [11], where the authors advocatethe complementary use of cellular systems and802.11p to successively broadcast the receivedCAMs on the downlink at road intersections,where 802.11p may suffer from non-line-of-sightconditions due to buildings.The main assumption in the above mentionedstudies is that the LTE capacity is exclusivelyused for CAMs, without accounting for otherbackground traffic with different QoS require-ments, such as voice or video, traditionally trans-mitted over LTE. Further investigation isrequired to analyze:• The reciprocal interference between CAMsand other traffic types• The effect of the LTE QoS class selected tocarry CAMs• The effectiveness of scheduling techniquesdeployed at eNodeBsDENMSDENMs generate a lower traffic load comparedto CAMs; thus, the cell capacity is only tem-porarily and partially used. In fact, DENMs,generated as a reaction to a hazard, have a limit-ed lifetime, and the number of senders is typical-ly significantly lower compared to CAMs.The main challenge is related to simultaneouswarning transmission attempts by all vehiclesdetecting a specific hazard (e.g., slippery roads;vehicle collision events may be detected andnotified by every vehicle passing the area). Inthis case, again, the back-end server plays theFigure 3. Multicast CAM delivery in LTE. The awareness range of the vehicles does not coincide with thecell range.CDE{C, D}{E, F, G}{C, D}{E, F, G}eNBBackend serverCore networkinfrastructureFAwareness rangeof vehicle AAwareness rangeof vehicle BCell rangeGI’m B in (XB, YB), SBI’m A in (XA, YA), SAThe main challengein supporting CAMsis to avoid systemoverload due to theheavy traffic broad-casted by a highnumber of vehiclesfrequently (typicallyevery 100 ms). Thisis especially critical indense areas, like citycenters, or duringpeak hours.CAMPOLO2 LAYOUT_Layout 1 4/29/13 1:10 PM Page 153
  7. 7. IEEE Communications Magazine • May 2013154crucial roles of reflector and aggregator. It canfilter the multiple uplink notifications of theevent according to the location, time stamp, andheading field of the received messages, and sendout only one consolidated message [4]. This lat-ter feature allows the server to infer a betterglobal view of the road conditions [8]. Suchadded remote intelligence, which tracks events,can only be offered in a centralized architecture.In addition, the detecting vehicle receives animplicit acknowledged notification of the sameevent on the downlink, so it has no need to repeatthe same DENM transmission several times. Sys-tem scalability is thus improved, channel resourcessaved, and uplink congestion avoided. As an addi-tional benefit, the wide cellular coverage guaran-tees the event dissemination also when there is nonearby vehicle to relay the message. Therefore,DENM over LTE results in a much more reliablesolution as demonstrated in [4, 10] in an emptysystem, where only DENMs are generated. In[12] the traffic is generated from a single vehicletransmitting a DENM to the base station, whichrepeatedly rebroadcasts it to all vehicles in thecell through MBMS. Different downlink schedul-ing schemes are compared, showing that QoS-aware schemes meet the DENM delay constraints.TRAFFIC EFFICIENCY: THEFLOATING CAR DATA CASEThe high deployment costs of a ubiquitous802.11p roadside infrastructure implies that onlya few RSUs will be installed to offer wirelessconnectivity to vehicles in the near future. SparseRSU settling and vehicle mobility would causeintermittent and short-lived Internet connectivitywith adverse effects on the performance of V2Iapplications. On the other hand, applications likeFCD relying on V2I communications can benefitfrom the LTE capillary network infrastructure.Although the latency and reliability require-ments are not as strict as for cooperative safetymessages, FCD generated by several vehicles canheavily load the uplink channel, thereby threat-ening the delivery of traditional human-to-human(H2H) traffic. In [13] a channel-aware transmis-sion scheme is proposed according to which vehi-cles probabilistically transmit FCD based on themeasured signal-to-noise ratio (SNR): the higherthe SNR, the higher the FCD transmission prob-ability. In doing so, the channel load is reduced,but the distribution of active FCD devices isinhomogeneous, opposite to what is required forreliable and accurate traffic forecasts.To reduce the traffic load generated by FCDtraffic, a hybrid clustering framework is proposedin [14] that relies on LTE and 802.11p. TheeNodeB is responsible for forming clusters amongnodes and selecting one vehicle in each cluster asthe cluster head (CH). The CH transmits theFCD from its cluster’s members to the eNodeB.The proposed technique achieves significantadvantages in terms of bandwidth usage andpacket delivery, compared to the case of directvehicle-to-eNodeB FCD transmission, thanks tothe aggregation performed by the CH. Intra-clus-ter communications leverage time-slotted accesson top of 802.11p that shows low overhead andreliability but requires strict synchronization. Nodetails about the LTE scheduling policy and theFCD impact on other traffic types are reported.LTE IN VEHICULAR SCENARIOS:LESSONS LEARNED ANDOPEN CHALLENGESThe surveyed literature provides some prelimi-nary results, limited to the illustrated cases ofCAMs, DENMs, and FCD support, and mostlyunder the simplistic assumptions of no othertraffic types in the system and no specificscheduling policy at eNodeB. In summary, wehave learned that:•Regarding DENMs, LTE can augment theability (i) to consolidate the numerous eventnotifications originated from all the vehicles in agiven area, and (ii) to disseminate only usefulinformation in a specific area, with positiveeffects on system scalability, congestion avoid-ance, and delivery reliability.•CAM delivery through LTE may suffer frompoor uplink performance in terms of messagelatency and potential congestion; however, LTEprovides advantages in terms of coverage in spe-cific hostile areas such as road intersections,where obstacles like buildings can obstruct theline of sight among all vehicles. In summary,LTE offers limited support to CAMs, providedthat it can control the CAM generation rule toavoid congestion.•Considerations on the CAM transmissionsFigure 4. CAMs and DENMs delivery in IEEE 802.11p. Messages are locally broadcast through V2V com-munications.Relevance areaAwareness range!!I’m A in (XA, YA), SADENMs generate alower traffic loadcompared to CAMs;thus, the cellcapacity is onlytemporarily and par-tially used. In fact,DENMs, generatedas a reaction to ahazard, have a limit-ed lifetime, and thenumber of senders istypically significantlylower comparedto CAMs.CAMPOLO2 LAYOUT_Layout 1 4/29/13 1:10 PM Page 154
  8. 8. IEEE Communications Magazine • May 2013 155also hold for FCD on the LTE uplink; they couldeasily overload the network due to periodic trans-missions. However, unlike CAMs, FCD must notbe transmitted by all vehicles: studies havedemonstrated that the collected traffic informa-tion is reliable even if only a small percentage ofthe vehicles periodically transmit FCD.•Unicast CAM delivery is less resource effi-cient than MBMS delivery but it shows advan-tages in terms of delays, since multicast setupprocedures can be avoided that are especiallycumbersome under heavy traffic load.•The backend server plays a key role in V2Vcommunications. The vehicle-to-server and in-network signaling load, which is also server-loca-tion-dependent, and the required intelligence atthe server vary with the vehicular application.Besides reflecting or aggregating messages, theserver may also take care of repeating a messageas long as the notified event persists so thatinformation can be refreshed for newly arrivedvehicles [8] in the relevant area.Many other challenges relevant to the LTEcapability to support the described vehicular appli-cation scenarios definitely require deeper analysis.Some of them uniquely arise for vehicular net-working, whereas others are exacerbated in thevehicular environment. The most relevant issuesare addressed below and summarized in Table 3.MULTICAST/BROADCAST SUPPORTMBMS is a promising solution to the localizeddistribution of cooperative safety applications.However, for MBMS to support geocasting, aback-end server with geomessaging capabilities isrequired, whose role, tasks, and deploymentissues should be specified in the LTE architec-ture. When MBMS operates in synergy withsuch a back-end server, only the intendedreceivers are addressed without unnecessarilyloading the network and burdening uninterestedvehicles with processing load. The drawbackcould be the signaling overhead due to subscribeand join procedures to the multicast service thatare performed on a per-user basis. Henceforth,lightweight procedures should be designed tobetter fit the delay requirements of vehicularapplications, especially when several, dynamic,and large multicast groups are to be served. Adedicated MBMS carrier for downlink-onlytransmissions is advocated in [4], which wouldeliminate the control overhead for unicast.NATIVE V2V SUPPORTV2V communications are not natively supportedin LTE; therefore, infrastructure nodes must beinvolved to distribute messages among vehicles.However, research is ongoing to enable directdevice-to-device (D2D) communications in LTE-A. In D2D mode, terminals in close proximitymay communicate directly and offload eNodeBresources. D2D would be an appealing solutionfor local data exchange among vehicles, but sev-eral issues should be addressed for D2D to bereally effective [20] in vehicular environments.Radio resource management policies shouldcontrol the interference between cellular andD2D communications by considering the highTable 3. A summary of the main deployment issues to support vehicular applications delivery over LTE.EnablingfeaturesKey issues Expected benefitsMBMS• Design of lightweight joining/leaving procedures fordynamic groups of vehicles• Backend server role, task and deployment modedefinition to support geo-addressingEfficient CAMs/DENMs disseminationScheduling• Proper mapping of vehicular traffic patterns to exist-ing QCI and/or new QCI definition• Cross-layer scheduling algorithms accounting formobility and vehicular communication patternsQoS support and differentiationNo penalization for non-vehicular applicationsD2D• Radio resource management policies to minimizeinterference in mobility conditions• Mode selection for D2D communicationLocalized V2V communications (e.g., CAMs) supporteNodeB offloadingMTC• Efficient transmission of small amounts of data withminimal network impactEasier management of some ITS applications (like FCD)Enhanced device• Powered with the vehicle’s battery• Multi-interface platformsBattery savingFlexibility offered by multi-technology communicationsBusiness models• Incentive-based approaches• Value-added services provisioningLarger subscribers basinHigher return-on-investmentsStandardization• Harmonized MTC and ITS standardization activities• LTE role in ITS reference architectureEnhanced functionalities/architecturesNew use cases and synergic solutionsCAMPOLO2 LAYOUT_Layout 1 4/29/13 1:10 PM Page 155
  9. 9. IEEE Communications Magazine • May 2013156mobility of devices. Moreover, the decisionabout the vehicles’ communication mode (cellu-lar or D2D) should account for the feasible D2Drange under different vehicle-eNodeB distances:• To not cause harmful interference to nearbynodes• To guarantee cooperative road messagesdissemination over an area even rangingseveral hundred metersPACKET SCHEDULING AND QOS SUPPORTResearch has focused on the design of LTE pack-et schedulers that satisfy the often conflictingobjectives of high spectrum efficiency, throughput,and fairness. However, the scheduling techniquesdesigned for H2H communications cannot bestraightforwardly applied to vehicular applications.In this case, the design of efficient schedulers isespecially crucial for the uplink channel, whichcould be a bottleneck in densely populated net-works. On the downlink, instead, the effort is toprovide efficient and reliable broadcasting thatcoexists with the conventional unicast mode.Closely linked to scheduling issues is the mappingof vehicular applications onto LTE QoS classes.There is a wide consensus on the assumption thatDENMs should be handled at the highest priority,but no QCI mapping is suggested. This is mainlybecause, in the surveyed studies, vehicular applica-tions are assumed to be deployed in an “empty”LTE system to assess the best case system capaci-ty. All in all, cross-layer scheduling techniques thataccount for node mobility and traffic generationpatterns, and new QoS classes could be consid-ered to match the vehicular application require-ments without penalizing H2H communications.AMENDMENTS TO THESTANDARD DOCUMENTS AND ARCHITECTURESIn order to enable LTE to support road safety andtraffic efficiency applications, some amendmentsare necessary to the current standard documentsand architectures. For example, in the ITS ETSIstation reference model, details should be addedon the manner of interfacing the LTE accesstechnology. The introduction of LTE as an addi-tional candidate access technology would requiresome changes in the specification of use cases inTable 1. For example, the emergency vehicle warn-ing service, currently based on CAMs, can beenhanced by using DENMs in cellular networks.In this case, the position of the emergency vehiclecould be used by the back-end server to sendDENMs to cars close to the emergency vehicle,but beyond the CAM awareness range, henceallowing the emergency vehicle faster movement.MACHINE-TYPE COMMUNICATION FORSUPPORT OF ITS APPLICATIONS3GPP is working on evolving LTE-A to accom-modate the requirements of machine-type com-munications (MTC), potentially involving a verylarge number of communication devicesautonomously (i.e., without human intervention)exchanging small amounts of data traffic. It isworth analyzing their relationship with ITS stan-dardization activities. As a matter of fact, severalvehicular applications, like FCD, vehicle diagno-sis, and fleet management, that imply data collec-tion from in-vehicle sensors and their transmis-sion to a remote server, are considered as MTCin [21]. Solutions under study in 3GPP for effi-cient transmission of small amounts of data withminimal network impact (e.g., signaling over-head, network resources, delay for real location)also show promising benefits for supporting thementioned ITS applications over LTE-A.DEVICE DEPLOYMENTLTE connectivity can easily be provided throughcommon user devices like smart phones. Althoughearly tests demonstrated the leading role of smartphones and mobile apps in the support of vehicu-lar applications [17], their pervasive use for thispurpose is questionable. The major concerns areraised by a possible cause of distraction for thedriver, the battery-powered nature of these devicesthat would require specific care in designing ener-gy-saving protocols and circuits, and the non-per-manent availability of these devices (e.g., if theyare switched off or out of battery, or if they arebusy in a traditional voice communication). As analternative solution, dedicated hardware could bedeployed, that is, an onboard unit powered by thevehicle’s battery system, and endowed with one ormultiple radio interfaces (e.g., IEEE 802.11p,LTE, positioning systems, in compliance with theITS ETSI station). Despite preliminary attempts,the automotive industry does not see the value ofadding such an expensive converged networkingplatform into vehicles unless a stable and conve-nient business model is conceived.CONNECTION COSTS AND BUSINESS MODELSBesides the discussed technical issues, economicissues should also be handled. Since LTE operatesin licensed spectrum, vehicles’ owners may becharged communication costs for data exchange.The costs could be not negligible when the data traf-fic is heavy and frequent, as is case with FCD andCAMs. Despite the diffusion of always-on Internetconnectivity encouraged by flat-rate subscriptions,users could be reluctant to bear the communicationcosts, unless attractive value-added services are pro-vided. The market value associated with on-the-roadservice provisioning could be huge; thus, new busi-ness models should be conceived between theinvolved parties: telco operators, road transportauthorities, service providers, and users.CONCLUSIONSIn this article we provide a survey on the state ofthe art of LTE in the view of assessing its capabil-ity to support cooperative ITS and vehicularapplications. There is a wide consensus on lever-aging the strengths of LTE (high capacity, widecoverage, high penetration) to face the well-known drawbacks of 802.11p (poor scalability, lowcapacity, intermittent connectivity). The conduct-ed analysis qualitatively captures the main fea-tures, strengths, and weaknesses of the standardguidelines and solutions under development.In the initial deployment phase of vehicularnetworks, LTE is expected to play a critical role inovercoming situations where no 802.11p-equippedvehicle is within the transmission range. Thiscould be the case in rural areas where the car den-LTE connectivity caneasily be providedthrough commonuser devices likesmart phones.Although early testsdemonstrated theleading role of smartphones and mobileapps in the supportof vehicularapplications,their pervasive usefor this purpose isquestionable.CAMPOLO2 LAYOUT_Layout 1 4/29/13 1:10 PM Page 156

×