EIGENVALUE DECOMPOSITION BASED ADAPTIVE INFORMATION FUSION FILTER ALGORITHM FOR INTEGRATED NAVIGATION SYSTEMS

An efficient, adaptive federated Kalman filter is developed which can be applied to integrated navigation systems. Researchers have indicated that the information sharing scheme can influence the performance of the decentralized navigation filter. The estimation precision of local filters is directly reflected from the corresponding covariance matrix. So the aim is to calculate the information sharing coefficients on line through the variation of the covariance matrix. Based on a new adaptive algorithm, the information sharing coefficients of the local filters are decided in a federated structure, according to the change of the estimation quality in the local filters. The global information thus can be feasibly allocated and reset the estimation in local filters after information fusion stage. A practical federated filter example is given. Simulation results demonstrate that the adaptive INS/GPS/Doppler system can realize accurate real time navigation task and have great advantages in fault tolerance and isolation. Thus, the algorithm can be deemed as practical scheme for integrating different kinds of navigation sensors.