An Integrated Study on Mobility Models and Scalable Routing Protocols in VANETs Wenjing Wang, Fei Xie and Mainak Chatterjee School of Electrical Engineering and Computer Science University of Central Florida Orlando, Florida 32816 Email: {wenjing, xiefei, mainak}@eecs.ucf.edu Abstract—A vehicular ad hoc network (VANET) is a high mobility wireless ad hoc network that is targeted to support vehicular safety, traffic monitoring, and other commercial ap- plications. Mobility models used in traditional mobile ad hoc networks cannot be directly applied to VANETs since real world factors such as road layouts and traffic regulations are not considered. In this paper, we propose a vehicular mobility model that reflects real world vehicle movement on the road. Based on the mobility, we study the performance of existing Mobile Ad Hoc network (MANET) routing protocols, i.e., AODV and GPSR. We observe the drawbacks of the MANET protocols and argue the inappropriateness of directly applying those MANET protocols to VANETs. We also propose simple modifications to these protocols which make them more suitable for small scale VANETs. When investigating the large scale VANETs, we introduce a two phase routing protocol that incorporates map information. The proposed protocol defines an overlay graph with roads of high vehicular density and forwards packet along the pre-calculated path in the overlay. The access, which is the rest areas relies on our modified small scale routing protocols to send packets to overlay . Both small and large scale routing protocols are validated with simulation. We generate vehicular mobility traces for different road layouts in Orlando by making vehicles follow the mobility model. We feed the traces to network simulators to study the routing behavior. Simulation results show the performance and effectiveness of our modified and proposed routing protocols for VANET scenarios. I. I NTRODUCTION The approval of the 75 MHz of spectrum at 5.9 GHz for dedicated short range communications (DSRC) [6] by the Federal Communications Commission (FCC) and the successful deployments of WLAN technologies are making vehicular ad hoc network (VANET) a reality [3], [21]. In recent years, VANET has emerged as a research area that has received increased attention from the community [23], [22]. However, due to the cost and difficulty associated with the implementation of VANETs in real world, computer simula- tions remain as one of the primary techniques to investigate networking characteristics of VANETs. In this regard, it is very important to adopt realistic vehicular mobility models, and design network protocols that are capable of delivering good end-to-end performance in such a highly mobile environment. It is widely accepted that the underlying mobility models greatly affect ad hoc network performance [1], [4], [26]. Many studies on mobile ad hoc networks (MANETs) have used the random way point (RWP) or the Manhattan model to simulate the node movement in an open field. Though such mobility models work well in certain scenarios, they are not suitable for VANETs simply because the movement of vehicles are constrained by the layouts of the roads. More- over, traffic regulations (e.g., speed limits, traffic lights) and driver behaviors (e.g., overtaking or following) make realistic vehicular mobility far different from the commonly used ones in MANETs. The traffic simulator framework introduced by Saha and Johnson [18] makes it convenient to use the real world map of USA and convert it to a graph model where the edges represent the roads and the vertices represent the intersections. The (map) input to the simulator is the data as provided by the US Census Bureaus’ TIGER project [20]. In their framework, each vehicle randomly chooses a source and a destination on the map, and moves along the route which is calculated by Dijkstra’s single source shortest path algorithm. The output of the framework is the trace of the vehicles on the map, the data format of which is compatible with the network simulator ns-2 [16]. However, the framework only uses simple mobility models (similar to RWP [12]) and does not include detailed realistic vehicle mobility models. Using the aforementioned simulator framework, we are motivated to investigate the performance of existing routing protocols in VANETs and design new ones. Due to the high mobility of the hosts (vehicles), proactive routing protocols are not suitable because the administrative overhead would be large. We focus on AODV [17] which is a reactive routing protocol and evaluate its performance in terms of average data packet delay, data packet delivery ratio, and routing overhead. We suggest modifications to the protocol that make it more suitable in a small scale VANET environment. We also consider a location-based routing protocol called GPSR [10] that utilizes the geographic information. We extend GPSR enabling it to handle highly dense and congested VANETs. We use two real geographic locations that yield different road topologies. Extensive simulations are conducted on the ns-2 platform that demonstrate the performance of the modified routing protocols and reveal that the proposed techniques are well-suited in a small scale VANET scenario. However, when applied to a large scale VANET (tens of miles scale), neither protocols maintain fair performance. The reason is quite intuitive; the large scale VANET is largely This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE INFOCOM 2008 proceedings. 978-1-4244-2026-1/08/$25.00 © 2008 IEEE