Consensus-based distributed information filter for a class of jump Markov systems

This study investigates the problem of distributed fusion for a class of jump Markov systems in a not fully-connected sensor network. A distributed information filter is proposed from the point of view of the consensus theory. To this end, the best-fitting Gaussian (BFG) approximation approach is applied to overcome the difficulty of lacking a global model for multiple model estimation fusion, and a recursive formula is presented for calculating the mean and covariance of this Gaussian distribution. Based on the approximated linear Gaussian system, local information filter is derived for each sensor and the filtering estimates are fused with its neighbouring sensor nodes using the dynamic average-consensus strategy. Performance comparison of the proposed filter with the optimal centralised fusion filter is demonstrated through a multi-static manoeuvring target-tracking simulation study.