Routage Intelligent dans les réseaux de capteurs à grande échelle

Abstract

Reducing energy consumption and scalability are key requirements in wireless sensor networks (WSNs), as these networks are generally composed of a large number of sensors under energy constraint. Therefore, energy efficiency in this type of networks is considered a critical problem. One way to achieve this goal is to minimize the amount of redundant data sent to the base station through the clustering approach which is one of the best approaches in terms of energy efficiency in large-scale RCSFs. In this thesis, we have proposed energy-efficient solutions for large-scale RCSFs. These solutions are based on an improvement of the unsupervised learning approach (K-Means) and imply methods to determine the appropriate number of clusters (Silhouette, Elbow and "Rule of Tumb"). In the first contribution, we evaluated each of these methods in order to know the most suitable approach for determining the number of clusters. In the second contribution, we proposed a routing scheme based on an improved version of K-Means. The third contribution is a routing scheme based on dynamic clustering and the fourth contribution is a routing scheme which involves “Rule of Thumb” to determine the number of CHs, K-Means to organize the network into clusters and an improved genetic algorithm to establish the paths between each CH and the base station. The proposed routing schemes were developed over Matlab. Simulation results have shown the benefits of our solutions in terms of power consumption, lifetime and scalability compared to other routing schemes.