Déploiement optimal des nœuds de capteurs employant le clustering K-means et un algorithme génétique

Abstract

In recent decades, wireless sensor networks have received a new impetus thanks to the miniaturization of electronic components. These components have made it possible to manufacture small systems equipped with memory devices, microprocessors, communication and sensor devices. These systems are called "sensor nodes". When deployed in an area of interest, they interact and communicate with each other autonomously to form a network of sensors capable of transmitting information to the base station. Sensor networks are used in several application areas such as health, transportation, military and agriculture. However, these applications are particularly sensitive on the quality of the coverage of the field and the network connectivity. In addition, several researchers are interested in the problems raised by these applications, such as energy, coverage, connectivity and redundancy, in other words, the problem of node deployment. In this thesis project, we address the problem of the deployment of sensor nodes in a specific application which is the prevention of forest fires. We propose a deployment approach comprising two phases : a phase of area representation based on an unsupervised learning algorithm which is the K-means clustering, and a phase of sensor node deployment based on the genetic algorithm. The objective of the first phase is to determine the points of interest to be covered taking into account the structure of the geographical area. For the second phase, its objective is to determine the optimal location of the sensor nodes taking into account various constraints such as coverage, connectivity, cost of deployment and overlap between nodes. The simulations carried out have demonstrated the performance of our solution in optimizing the deployment of sensor nodes in large and small-scale geographical areas while ensuring maximum coverage, full network connectivity, minimum number of nodes and minimum redundancy. In addition, experimental results have proven the effectiveness of our approach to provide optimal solutions.

Date	26 Apr 2021
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Abdelouahed Gherbi (Supervisor)