Dynamic Event-Triggered Feedback Fusion Estimation for Nonlinear Multi-Sensor Systems With Auto/Cross-Correlated Noises

This paper aims to solve the distributed fusion estimation problem for a nonlinear system with auto/cross-correlated noises. An equivalent nonlinear system with uncorrelated noises is obtained by means of a de-correlation method. Due to the nonlinear characteristics, the order of de-correlation affects whether the noises are completely uncorrelated or not. In order to improve accuracy of fusion estimation while avoiding the increase of communication burden, fusion predictions are fed back to local filters according to a dynamic event-triggered scheduling (DETS). The feedback frequency is reduced by introducing real-time adjusted offset variables into the DETS, which makes the event-triggered scheduling more strict. Subsequently, a local filter in the form of unscented Kalman filter (UKF) is designed using the measurement and received feedback information. Based on the Kalman-like fusion strategy, a distributed fusion estimation algorithm subject to auto/cross-correlated noises is developed, and boundedness of the fusion error covariance as well as complexity of the fusion algorithm are analyzed. Finally, performance of the proposed fusion estimation algorithm is verified by a numerical simulation.