An efficient particle filter-based potential game method for distributed sensor network management

This paper addresses information-based sensor selection that determines a set of measurement points maximizing the mutual information between the measurements and the target states. The problem is formulated as a potential game in which each player computes a local utility function defined by the conditional mutual information. A new approximation method is proposed for computing the conditional mutual information when the target states are represented using a particle filter to handle a non-linear system with non-Gaussian noise. This method approximates the conditional entropy of an agent conditioned on other agents sensing decision by sampling the other agents measurements from a particle filter. This computational method makes it possible to apply the potential game approach to non-linear/non-Gaussian problems with a large number of the measurements. We performed simulations for localization and tracking of a target with mobile/deployed sensor networks.