Enhanced Multi-sensor Data Fusion Methodology based on Multiple Model Estimation for Integrated Navigation System

A novel multi-sensor data fusion methodology is presented in this paper with respect to noise with unknown or randomly varying statistics properties and outliers in the SINS/GPS/Odometer integrated navigation system. The proposed methodology combines an adaptive interacting multiple model filtering (AIMM) and federated Kalman algorithm. The former implements dynamic interaction and dynamic change of multiple modes based on the Markov chain process of system models. To achieve the adaptive outlier detection and processing in the measurement signal, modified Kalman filter based on orthogonality of innovation serves as the parallel model filters in the AIMM approach. The advantage of decentralized filter architecture of the latter federated algorithm is flexibility and modularity. It has received considerable attention because of its outstanding fault detection and isolation capability. Experiment results show that the proposed multi-sensor data fusion methodology significantly improves the navigation estimation accuracy and reliability as compared to the federated extend Kalman filter and federated IMM filter approaches.