Event-triggered distributed cooperative extended Kalman filter based on formation estimation

The existing distributed filters focus on the consensus estimation of multi-sensor information, but ignore the cooperative characteristics of formation targets, resulting in information loss and limited estimation accuracy. To address the problem of formation target tracking, a novel event-triggered distributed cooperative extended Kalman filter based on formation estimation is proposed, which improves the estimation performance without additional information by mining formation characteristics. Different from previous works, the formation estimation based on the best-matched model is developed to establish the connection between different targets, by which the unknown formation evolution can be modeled in the sense of minimum innovation. Considering the communication topology and the formation detection, an event-triggered mechanism is constructed to detect formation status reliably and reduce the communication burden. Afterwards, a novel distributed filter is designed by employing the multi-agent consensus theory. The cooperative term can be considered as a pseudo measurement, which forces local estimations to reach a consensus. By virtue of the variance-constrained approach, the filter gain is derived by minimizing the upper bound of the estimation error covariance. The proposed filter can also realize optimal estimation in the absence of cooperation and the multi-sensor fusion function for a single target. Moreover, the exponential stochastic boundness of the proposed filter is analyzed based on the stochastic stability theory. Finally, numerical experiments are carried out in three different simulation scenarios, and simulation results show that estimation errors of the proposed filter are significantly decreased compared to existing filters. The advantages of the proposed filter become more apparent as the number of formation members increases or the information quality of formation members improves.