Federated filtering for fault tolerant estimation and sensor redundancy management in coupled dynamics distributed systems

This paper discusses the application of the idea of federated filtering to the estimation of intrinsically nonlinear distributed systems by examining its impacts on filtering performance by using the Extended Kalman Filter (EKF) as state estimator. Specifically, the performance of the traditional centralized solution is compared with the filtering structure obtained using the federating idea, and their conceptions, and their ability to balance between fault tolerance and estimation accuracy is examined. Our research demonstrates how successfully the EKF for solving nonlinear estimation problems in federated structures can be used, noting that the idea of federation have only been demonstrated for linear problems previously. In addressing the demands of both fault tolerance and estimation accuracy, it is shown that increased filtering accuracy is relied on the proper choice of sharing of the error covariance information between local filters, while sensor fault tolerance is provided by the utilization of an appropriate resetting policy of the filter.