Consensus-Based Distributed Linear Filter for Target Tracking With Uncertain Noise Statistics

The conventional distributed Kalman filter (DKF) assumes that the noise sources have known statistical properties. However, the statistical properties of noises are not completely known in practical applications. Therefore, they need to be estimated from some side information. Clearly, an imperfect estimation of the statistical properties with a non-zero error can deteriorate the performance of DKF. In this paper, we develop a novel distributed linear filter, which is based on H
_∞
filter. It is called as distributed H
_∞
filter (DHF). Compared with the DKF, the DHF does not require the statistical properties of noises. It can alleviate the effect of estimation error of statistical properties. So, it is more flexible and robust than the DKF in applications. Then, we extend the DHF, and develop a combination of the DHF and the interacting multiple model algorithm for the maneuvering target tracking. Simulation results illustrate that the DHF outperforms the DKF when the statistical properties of noises are not completely known.