**5.2 Routing**

462 Wireless Communications and Networks – Recent Advances

In the data fusion process the main focus is on object and situation refinement levels, which refer to the state estimation of objects and the relations among them, correspondingly. The discrimination between these levels is also made by using the terms low and high level fusion instead of object and situation refinement. The different levels of the JDL model are

 *Level 0*: Preprocessing of sensor measurements (pixel/signal-level processing). *Level 1*: Estimation and prediction of entity states on the basis of inferences from

*Level 2*: Estimation and prediction of entity states on the basis of inferred relations

*Level 4*: Adaptive data acquisition and processing related to resource management and

In the past decade the advances in autonomous sensor technologies and the major objective of the European Union to reduce to a half road accidents and fatalities by 2010, led to the development of advanced driver assistance systems (ADAS). The fusion of data coming from different advanced in-vehicle sensors was initially in the centre of this attempt.

the sensor systems cannot perform well in all environments (the urban roads comprise

in several cases the system is not able to perceive the situation in time in order to warn

the cost of the sensor systems is too high and so their installation is feasible only at

the perception environment of the vehicle cannot go beyond the sensing range,

*Level 3*: Estimation and prediction of effects on situations of planned or

However, this approach suffers from serious limitations. Specifically:

estimated/predicted actions by the participants.

the driver and suggest a corrective action,

Fig. 10. Joint Directors of Laboratories (JDL) model.

summarized below:

observations.

among entities.

process refinement.

a major challenge),

luxurious vehicles.

Routing is the process of finding a path from a source node to a destination node. In this section the word "node" will be used interchangeably with the word "vehicle" because a vehicle is actually a node of a vehicular network. Since each node has limited transmission range, messages often need to be forwarded by other nodes in the network to reach their final destination (i.e. multi-hop communication).

Despite the fact that there are already some routing protocols available, which are mainly derived from the Mobile Ad-hoc Network (MANET) domain, it is an intensive scientific research area due to the highly dynamic nature of vehicular networks.

The routing protocols designed specifically for co-operative systems (Lee et al., 2010; Li & Wang, 2007) can be divided into two broad categories: *topology-based* routing and *locationbased* routing. The former use information about the existing links of the network to forward the relevant messages. In the latter forwarding decisions are based on the location of the nodes. Moreover, position based routing protocols can be further divided into *proactive* and *reactive*.

Proactive algorithms are using classical routing strategies such as distance-vector routing or link-state routing. Proactive algorithms maintain routing information about the available paths in the network even if these paths are not currently used. The main disadvantage of this approach is that the maintenance of unused paths occupies a significant part of the available bandwidth if the network topology changes frequently.

In response to the problem of maintaining the paths of proactive protocols, reactive routing protocols were created. Reactive protocols maintain only routes that are in use, thereby reducing the load on the network when only a small subset of available paths are used.

In location-based routing, forwarding decisions are based on the location of the node that forwards the message according to the location of the source and destination nodes. In contrast to pure ad hoc approaches which are based on topology-based routing, here it is not necessary to setup or maintain a path since packets are forwarded directly. Location-based routing protocols consist of location services and geographical forwarding.

Geographical forwarding takes advantage of a topological assumption which works well for wireless ad hoc networks: nodes that are physically close are likely to be close in the network topology too. Each node is aware of its location using technologies such as GPS and periodically broadcasts its presence, location and speed to its neighbors. Thus, each node maintains a table with the identities and locations of its current neighbors. When one node needs to forward a packet it includes the identifier of the destination-node and its geographical location into the header of the packet. Each node along the forwarding path consults its list of neighbors and forwards the packet to the neighbor closest to the destination in terms of physical location, until it reaches its final destination.

Although the geographical forwarding works well for networks where nodes are uniformly distributed, perhaps cannot find a route to a packet's destination when the packet has to travel around a topology "hole" - that is, when an intermediate forwarding node has no neighbors who are closer than itself to the destination of the packet.

An overview of some *topology-based* routing algorithms is given below:


A brief description of some *location-based* routing algorithms is given below:

 **Connectivity-Aware Routing (CAR)** is a routing algorithm which derives from the work performed by the Preferred Group Broadcast (PGB) to reduce the broadcasted packets during the discovery of the AODV route taking also into account the mobility of the nodes. CAR uses the route discovery of AODV to find routes with reduced broadcasting from PGB. However, the nodes forming the route record neither their previous node from the backward learning nor their previous node which forwards the response route packet from the destination. Only anchor points, which are nodes near an intersection or a curve of the road, are recorded in the route discovery packet. A node defines itself as an anchor point if its velocity vector is not parallel to the velocity vector of the previous node in the packet. The destination may receive multiple route discovery packets. If this happens it chooses the path that provides the best connectivity and the shortest delays. More details about CAR can be found in (Naumov & Gross, 2007).

 **Geographic Source Routing (GSR)** is based on the availability of a map. It calculates the shortest Dijkstra path of the cascading graph where vertices are intersection nodes and edges are the roads connecting these vertices. The sequence of intersections is setting up the route to the destination. Then the packets are greedily forwarded between intersections. GSR does not take into account the connectivity between two intersections, so the route might not be fully connected. In case such a situation occurs a recovery with greedy forwarding takes place. The most significant difference between the GSR and CAR is that CAR does not use a map and uses proactive discovery of anchor points that indicate a turn at an intersection. More details about GSR can be found in (Lochert et al., 2003).
