|Keywords:||Computer Science and Software Engineering|
|Full text PDF:||http://hdl.handle.net/10415/193|
A knowledge of real time traffic density on different roads has many applications such as real time navigation for driver, designing efficient vehicular routing protocol and building fully autonomous vehicles. Forming clusters of vehicle is the first step towards achieving these goals. Once clusters are formed, distributed servers could be built which would collect and store all the information. Querying the distributed servers would give density information at various roads. We propose a clustering protocol, RSDCP which adapts dynamically to high mobility of vehicles. Our protocol is able to form stable clusters by choosing the cluster head based on relative speed. To demonstrate this, we compare it with two other protocols, one which has same clustering mechanism as that of our protocol but is based on id instead of relative speed (IDDCP) and the second is a simple clustering protocol (IDS) where cluster head is elected periodically based on id, similar to the one proposed by Gerla et al. . We did the analysis from three different perspectives, in terms of time, clustering, and network packets. For evaluating in terms of time, time spent as part of cluster was measured. From clustering perspective, number of clusters, and number of vehicles per cluster were measured. To estimate the network performance, number of protocol packets, and application packets transmitted were measured. Results show that in RSDCP, vehicles are part of cluster 30% longer than IDS and nearly 5% longer than IDDCP. RSDCP on almost all the test scenarios has higher average number of clusters as well as higher number of vehicles per cluster. Comparing the overall packet overhead, we notice that IDS has nearly 50% higher overhead than RSDCP while IDDCP has nearly 10% higher overhead than RSDCP.