Modeling of mobility and groups in inter-vehicular MANET-based networks Juan-Carlos Cano, Carlos Calafate, Pietro Manzoni Member, IEEE, C-K. Toh Senior Member, IEEE. Abstract— This paper evaluates the behavior of mobile ad hoc networks when group mobility is involved. We concentrate on group mobility because road vehicles tend to form groups or fleets whose behavior depends on the behavior of the close-by vehicles. We propose four different group mobility models and present a mobility pattern generator, called grcmob, that we designed to be used with the ns-2 simulator. We evaluate the performance of vehicular ad hoc networks under a wide number of scenarios using both TCP and UDP data traffic. We show that the number of groups is more important than the number of nodes and that the impact of area size is almost negligible. We also reveal that the mixture of inter- and intra- group communications has the strongest impact on performance. Finally, we show that the presence of groups forces the network topology to be more sparse and therefore the probability of network partitions and node disconnections grows, affecting performance especially as inter-group traffic increases. Index Terms— Mobility models, ad hoc networks, optimization, inter-vehicle communication. I. I NTRODUCTION If cars were able to communicate with other surrounding cars, it would be possible to avoid crash situations thus saving lives on roads. Cars could form groups whose membership is unpredictable and whose structure is variable with time. A Mobile ad hoc network (MANETs) is an example of a spontaneous wireless network that do not require any fixed infrastructure. Ad hoc networks are being adopted in situations where communication is required but is not possible to create a fixed network infrastructure. Ad hoc networking is regarded as an adequate solution to cooperative driving between com- municating cars on the road. The Internet Engineering Task Force (IETF) MANET work- ing group has proposed various routing protocols for ad hoc networks over the past few years. The evaluation of most of these proposals has been performed with the aid of various network simulators. Most of these tools, such as the ns-2 or the GloMoSim, make use of synthetic models for mobility and data patterns. Our work shows the impact of group mobility on the perfor- mance of a MANET and presents the critical factors that affect its behavior. We compare our results with the classic random waypoint model [1] commonly used by other researchers. We concentrate on group mobility because vehicles tend to form groups or fleets whose behavior is dependent on the behavior of close-by vehicles [2]. For example, one particular case of study on inter-vehicle coordination is at a 4-way stop. A 4-way stop is a typical road junction, with four approaching roads of J.C. Cano, C. Calafate, and P. Manzoni, are with the Polytechnic University of Valencia. C-K. Toh is with the Hong-Kong University. equal importance, meeting at a single point. The problem of a 4-way stop is in determining which vehicle should be allowed to cross the junction at any one time. In this paper we propose four different group mobility mod- els and present a mobility pattern generator called grcmob that we have designed and developed for use with the ns-2 simulator. In our analysis, we select a Dynamic Source Routing (DSR) based MANET but other routing protocols could also be considered. Our paper is organized as follows: Section II describes related work dedicated to the analysis of the impact of group mobility on MANETs. Section III describes our proposed mobility models and the designed software tool and discusses problems associated with group mobility. Section IV presents the performance evaluation of a DSR-based MANET with our proposed four mobility models. Finally, Section V concludes our work. II. RELATED WORK Mobile ad hoc networks require an efficient routing protocol to cope with the dynamics in network topology. Based on the study of mobility behavior of mobile users [3], existing mobil- ity models try to closely represent the movement behavior of mobile users. These models, along with appropriate MANETs scenarios, provide a suitable environment for the simulation and evaluation of ad hoc communication performance. The general problem of modeling the behavior of nodes belonging to a mobile network does not have a unique or straightforward solution. Mobility patterns are dependent on various factors, such as the physical environment, the user’s objectives, and the user’s inter-dependencies. Hong et al., [4] showed that these models can greatly affect the results of the simulation, and thus, the evaluation of these protocols. The most widely used individual-based mobility model is the random waypoint model where motion is characterized by two parameters: (a) the maximum speed and (b) the pause time. Each node starts moving from its initial position to a random targeted position residing within the simulation area. The node speed is uniformly distributed between 0 and the maximum speed. When a node reaches the targeted position, it waits for a pause period, then selects another random targeted location and moves again. Several papers have highlighted the weaknesses of the random waypoint model (e.g., [4], [5]). Some authors consider the random waypoint model unrealistic, as Tian et al., pointed out in [6]: “Imagine a world where people constantly try to pass through walls and cars suddenly leave the roads and drive into rivers”. In [7] the authors show that the random waypoint model fails to provide a 1