A Fusion Method with Information Gains for Distributed Multi-sensor Network

In this paper, the problem of distributed fusion in multi-sensor networks under known correlation is studied. In the network, each node calculates local estimates based on its measurements before estimates are transmitted to be fused at the fusion center. However, there are correlated errors caused by the common prior information involved in estimates and may be double-counted in fusion. To eliminate the correlated errors an information gains method is presented in this work. Firstly, the information form of estimates is utilized to analyze the supplementary information among local estimates. Simultaneously, the difference of information matrices is utilized to formulate the supplementary information. Subsequently, from a geometric point of view, the supplementary information is treated as information gains in calculating fusion covariance. Finally, a linear weighted fusion estimator is designed by considering the fusion covariance and information gains. The method is unbiased and its effectiveness is illustrated by simulation examples of target position.