A Cluster-on-demand Algorithm With Load Balancing For VANET Urban Environment

The network topology of Vehicular Ad-hoc Network(VANET)changes frequently due to the fast-moving of nodes.Though clustering of nodes is widely used as an effective method to control the network topology,it is still a challenge to improve the network stability while to reduce data access delay simultaneously.Therefore,two novel clustering algorithms for urban VANET are proposed in this thesis.Firstly,a cluster-on-demand clustering algorithm(CDVC)with load balancing is proposed based on the classification of communication requirements and the principle of neural network.In the algorithm,the nodes contained different types of messages are divided into several sub-clusters in order to reduce the access delay of data caused by different types of message.Self-Organizing Map(SOM)is for the sub-clustering to enhance the network stability,which can control the number of clusters and members of each cluster according to the similar properties among nodes such as location,velocity and so on.Secondly,a cluster-on-demand clustering algorithm based on fuzzy neural network and load balancing is proposed according to the characteristic of VANET.In this algorithm,a learning mechanism by combining the neural network,fuzzy logic and the behavior predictipn of drivers is established so that the variation of acceleration and deviation angle can be self-learned for dynamic and adaptive clustering.Finally,the effectiveness of the proposed algorithms is testified by simulation.MATLAB and SUMO are used to simulate the cluster performance metrics such as duration time of cluster head,number of cluster and distribution of cluster members,and the communication metrics such as throughput and access delay.Moreover,we compare the performances of the proposed algorithm with that of Mobility Based on Clustering(MOBIC)and Lowest ID(LID).The simulation results show that the proposed algorithms are superior to other two methods in terms of the mentioned metrics.