Comparison of a Distributed Kalman Filter Versus a Centralized Kalman Filter with Fault Detection Considerations.

摘要

This project examined the results obtained by simulating an aircraft navigation system with a partial complement of a typical avionics sensor array. Two different techniques of estimation processes were utilized and compared: the conventional Kalman and the federated filter architectures. Areas of interest include error state estimation accuracy and overall performance, residual behavior under normal and induced sensor failure conditions, and potential for failure detection and isolation. Several simulations were accomplished for each filter design and the results were compared in order to verify the validity of the newly developed federated filter architecture. Comparison of the error state estimation accuracies of the two filter designs revealed excellent overall performances for both. The identification of failures showed a definite advantage in the federated filter design. Having sensor-dedicated local filters allowed for easy sensor failure identification for the federated filter. while the centralized filter design suffered from navigation solution corruption under the same circumstances. Once established as a valuable estimation technique, the federated filter will add significantly to the viable alternatives when choosing a particular filter architecture for current avionics modifications or future avionics implementations. Kalman filter, Distributed kalman filter, Federated filter, Residual monitoring, Fault detection and isolation.