Game theory control solution for sensor coverage problem in unknown environment

This paper studies the coverage problem in an unknown environment by a Mobile Sensor Network (MSN). Each agent in the MSN has sensing, communication, computation and moving capabilities to complete sensing tasks. Here the agents need to relocate themselves, from their initial random locations, to their optimal configuration. The proposed algorithm is based on game theory control where a collection of distributed agents use their local information to make decisions. A state-based potential game is defined in which each agent´s utility function is designed to consider the trade off between the worth of the covered area and the energy consumption. The agents employ binary log-linear learning to update their actions in each iteration in order to converge to the Nash equilibrium. As the agents do not have the knowledge of the sensing area, a Maximum Likelihood estimation scheme is used to estimate the unknown parameters of a Gaussian Mixture Model (GMM). Then in order to feed the estimation algorithm with more informative data, a mutual information term is added to the agents´ utility functions. The mutual information is utilized to determine which observation can improve the agent´s knowledge of the unobserved area more. Simulation results are provided to verify the performance of the proposed algorithm.