A Cognitive Self-Organising Clustering Algorithm for Urban Scenarios

Cooperative communications based on data sharing and relaying have been gaining huge interest lately, due to the increase in the number of mobile devices and the advancement in their capabilities. Research on green communications, location based services and mobile social networking have fueled research on this topic. Vehicular technology have also fostered this cooperative approach as a means to provide scalability and privacy preserving mechanisms. In these scenarios, a commonly suggested approach to benefit from cooperation is the formation of virtual groups of mobile terminals, usually referred to as clusters. Mobility-aware clustering algorithms are commonly proposed to form such clusters based on the mobility characteristics of the mobile devices. However, these solutions are limited by the unpredictable nature of mobility behavior that leads to frequent disconnections of nodes from clusters; hence reducing the time availability of cooperative relationships. In this paper, we go beyond existing research on clustering by including a cognitive perspective. We propose data mining and cooperative optimization in order to deduce mobility pattern information in conjunction with the clustering process. We propose a low complexity algorithm that can dynamically adapt to different mobility characteristics of an urban scenario, more importantly without the need for previous configuration/information. The proposed technique achieves considerable gains in terms of stability in urban scenarios. Additionally, the paper presents a comprehensive analytical evaluation of the problem and the proposed solution, and provides extended simulation results in both matlab and ns2. Results show an outstanding gain up to 150 % in cluster lifetime and 250 % in residence time of nodes within clusters and reduces the overhead for clustering maintenance in 70 %.