Application of Adaptive Federated Filter Based on Innovation Covariance in Underwater Integrated Navigation System

For the underwater integrated navigation system composed of multiple navigation sensors, the uncertainty of measurement noise has a direct impact on the performance of standard Kalman filtering algorithm for each local filter, which results in the performance degradation of entire federated filter or even abnormal use. Based on the hypothesis of standard Kalman filter, an adaptive federated filtering method based on innovation covariance is proposed to improve the adaptive ability of the whole system in this paper. First, the popular real-time estimation of innovation covariance is derived in according to maximum likelihood estimation (MLE) criterion. Then, a scaling factor is introduced in each local filter to modify the filter gain directly under uncertain measurement noise. The simulation and analysis of the proposed algorithm mapplied in SINS/DVL/TAN/MCP underwater integrated navigation system, verify its validity and robustness in the presence of measurement noise uncertainty. A comparison to traditional federated Kalman filtering method demonstrates that our method provides a considerably improved accuracy and performance.