Decentralized clustering in VANET using adaptive resonance theory

Nowadays VANETs are becoming a dominating technology in automotive industries where vehicles communicate with each other to deliver safety messages or any type of information to other vehicles. However, the increasing numbers of vehicles on the roads poses a challenge on designing an efficient communication protocol for VANETs. The scalability issue in VANETs has a deteriorating effect on latency and on the stability of the network. Clustering is one technique used for solving this issue. In this work, we propose a clustering technique that creates mini clusters that are in the same communication range of the vehicles with the help of Adaptive resonance theory (ART). These mini clusters are created based on speed where it categorizes the vehicle in one of three levels: high, medium or low speeds. ART is an unsupervised neural network model that classifies inputs based on the degree of similarities of the input. By carefully tuning ART, three clusters are always obtained corresponding to the above speed classifications. The proposed work was simulated and compared against traditional clustering methods where our work presented a 50% advantage over these techniques.