[attachment=8759]
ABSTRACT
Inter-vehicle ad-hoc (IVC) networks are highly volatile due tohigh mobility rates of vehicles and varying driver behavior. Asignificant challenge in IVC networks would be theimplementation of suitable routing protocols so as to ensuresuccessful data packet delivery and lower packet delivery delay.IVC networks are subject to several limitations such asfragmentation of the network due to low deployment of ITS invehiclesystems and low vehicle density. The rapidly changingtopology would cause links between vehicles to be less reliablethan those in traditional mobile ad hoc networks. This paperframes the issues and then surveys recent work done in this field,including routing protocol studies and research on networkarchitectures making use of fixed roadside gateways
1. INTRODUCTION
intelligent Transportation Systems (ITS) is receiving increasedemphasis due to its role in traffic safety and in ensuring betterquality travel. ITS provides a variety of user services, such astravelers trip planning and driver control/vehicle safetyinformation, through the use of a wide range of communicationstechnology and applications. Effective implementation of the ITSuser services would help manage traffic flow and provide essentialinformation to drivers that can in tum, save lives, time and money[I].Implementation of the ITS user services would require datadissemination in a timely and accurate manner. Five basic types ofmessages that are exchanged between vehicles are [2]:I) Basic Safety Messages: contains data describing thesender characteristics and driving conditions (e.g. senderposition, speed, vehicle brake and steering conditions)2) Warning Messages: contains critical warning onemergency situations that has occurred or could possiblyoccur within the traffic.3) Infotainment Messages: contains data about services andresources available and offered by other vehicles andinformation of general interest (e.g. traffic conditions ormeteorological data).4) Routing Messages: contains data used by routingprotocols.5) Inter-Personal Messages: contains different profiles ofthe drivers and of the passengers situated in the vehicles.These data are used when employing interpersonalapplicationsComplementing hardware such as advanced radars, videoprocessing systems, as well as in-vehicle Geographic InformationSystems (GIS) [2], inter-vehicle (IVC) and road-vehicle (RVC)communication technology play an essential role in allowing data exchange among vehicles where velocity and location areconstantly and rapidly changing. A wireless ad-hoc network needsto be spontaneously created and re-configured in a variety of trafficsituations to accommodate the volatile nature of inter-vehiclecommunication.
2. INTER-VEHICLE COMMUNICATIONS
The successful deployment of Intelligent Transportation Systems(ITS) relies heavily upon inter-vehicle communication for effectivedissemination of information. Wireless inter-vehicle ad-hocnetworks (IVC networks) can be classified as a special case oftraditional mobile ad-hoc networks (MANET).Two problems characterize the problem space of IVC (FigureI). Firstly, the high mobility rates of nodes lead to rapid topologychanges. However, unlike typical ad hoc networks, the effective adhoc network membership itself changes rapidly, as cars move frombeing far apart to close together and vice versa. What do we meanby effective ad hoc network membership? Most of thecommunications that would need to be supported between vehiclesare communications between vehicles in close geographicalproximity (within a few kilometers at most, depending on theapplication). Hence, at any given time, we can define the effectivead hoc network of a vehicle as those set of other vehicles it mayreasonably expect to communicate with. In many other ad hocnetworks, this set is relatively fixed, even if the nodes are movingrapidly.The second problem is what we call the scatteredcommunications effect. Each vehicle that participates in intervehiclecommunication must be equipped with special in-vehiclesystems (wireless transceiver). However, due to slow introductionand adoption of ITS, deployment rates of such systems areconsiderably low [3]. This results in fragmentation of the network,whereby there are groups of vehicles (scattemets), which cancommunicate, but no means exists to communicate between thosegroups.A similar sort of fragmentation occurs in cases where shortrangeradio communication is used (e.g. the transmission range ofeach vehicle is about I km under line of sight conditions). Also,when vehicle density is low the vehicles may be spread far apartcausing them to be out of each others' transmission range.Frequent topology changes that occur in IVC networks also causelinks between vehicles to be broken.In section 3, we survey work done on suitable routingprotocols that enable multi-hop data exchange and forwardingbetween cars and stationary gateways in a rapidly changingtopology, where the effective ad hoc network of each participantmay change relatively quickly. In section 4, we survey work doneto solve the scattered communications effect. In Section 5, wehighlight a research issue that arises from use of fixed roadsidegateways for forwarding data packets.
3. ROUTING PROTOCOLS
Mauve et al [4] classified available routing protocols into twodistinct groups, namely topology-based and position-basedrouting.
3.1. Topology-Based RoutingTopology
-based routing protocols [4], use information about tbelinks in the network to perform packet forwarding. It can be furtberdivided into proactive, reactive and hybrid approaches. Theproactive approach (e.g. DSDV) maintains routing informationabout all available patbs in tbe network. A significant part of tbecommunications bandwidtb would be used up to maintain unusedpatbs if tbe topology of tbe network changes frequently. The keyadvantage of DSDV over traditional distance vector protocols istbat it guarantees loop freedom [7], [10].To address the flaws of tbe proactive approach, reactiverouting protocols (e.g. AODV, DSR) were introduced where onlyroutes tbat were active are maintained. Since routes are onlymaintained while in use, it is typically required to perform a routediscovery before packets can be exchanged between a new pair ofsource and destination. This leads to a delay for tbe first packet tobe transmitted and a significant amount of network traffic when tbenetwork changes frequently. Also, packets en-route to tbedestination are likely to be lost if the route to the destinationchanges [4].AODV is an improvement on DSDV because it typicallyminimizes tbe number of required broadcasts by creating routes ona demand basis, as opposed to maintaining a complete list of routesas in tbe DSDV algoritbm. AODV employs a route discoveryprocedure similar to DSR. Overhead of DSR is potentially largertban that of AODV since each DSR packet must carry full routinginformation, whereas in AODV packets need only contain tbedestination address. The memory overhead may be slightly greaterin DSR because of the need to remember full routes, as opposed toonly next hop information in AODV [6].The hybrid approach (e.g. ZRP) combines proactive andreactive routing strategies. However, at least tbose network patbstbat are currently in use must be maintained, limiting tbe amountof topological changes tbat can be tolerated within a given amountof time. In [9], Brealdi & Baldoni propose a new cachemechanism, based on tbe notion of caching zone, whichproactively removes stale information from tbe caches of all tbenodes. The additional overhead due to tbe maintenance procedurefor cached data is highly compensated by tbe reduction in tbenumber of patb discoveries, so tbe overall protocol efficiencyincreases, getting an application level performance improvement.
3.2. Position-Based RoutingPosition-
based routing [4] alleviates some of tbe limitations oftopology-based routing. It requires tbat each node determine itsown position tbrough tbe use of Global Positioning System (GPS).Position-based routing thus does not require tbe establishment ormaintenance of routes but routing decisions at each node is basedon the destination's position contained in tbe packet header. Whenattempting to send a packet, a location service (e.g. DREAM,Quorum-based, GLS, Homezone) is used to determine the positionof the destination. Each location server may maintain tbe positionof some specific or all nodes in a network.In DREAM, tbe position information is flooded within thenetwork. Quorum-based position discovery requires identifyingoverlapping groups of participants. The GLS approach works byhashing the ID of a node on the IDs of so-called location servers.The Homezone algorithm requires tbat tbe ID of a node be hashedon a position.Three main packet forwarding strategies for position-basedrouting are greedy forwarding (e.g. GPSR), restricted directionalforwarding (e.g. LAR), and hierarchical approaches (e.g.Terminodes) in which nodes attempt to forward a given packet toone or more on-hop neighbors tbat are located closer to tbedestination tban tbe forwarding node itself. Each node updates tbemost accurate position of tbe destination before forwarding it.Periodic one-hop broadcasts are sent to learn tbe position of nodesneighbors. These forwarding strategies may fail if there are no onehopneighbors' which are closer to tbe destination tban theforwarding node itself. However, tbis kind of failure is resolvedby the use of recovery strategies.Forwarding packets based on position information isseparated into three distinct areas. Greedy routing works byforwarding packets in the direction of the destination. GPSR'sperimeter routing can be used to avoid dropping tbe packet. Inrestricted directional flooding, LAR, tbe packets are broadcasted intbe general direction of the destination. LAR differs from DREAMin that it uses position information only to set up a route in anefficient manner. The actual data packets are routed witb aposition-independent protocol. In tbe Terminodes approach,routing is done hierarchically by means of a position-independentprotocol at tbe local level and a greedy variant at tbe long-distancelevel [4], [12], [13].