Particle based probability density fusion with differential Shannon entropy criterion

This paper focuses on a decentralised nonlinear estimation problem in a multiple sensor network. The stress is laid on the optimal fusion of probability densities conditioned by different data. The probability density conditioned by the common data is supposed to be unavailable. The optimal fusion is elaborated in the particle filtering and differential Shannon entropy framework. The conversion of weighted particles into a continuous probability density function is performed implicitly by the time update. Further, the issue of sampling density proposal is explored. The proposed approach is illustrated in numerical examples.