Distributed Fault-tolerant Fusion Estimation Based On Multiple-model Extended Kalman Filter

As practical nonlinear systems become progressively complex, the extended Kalman filter (EKF) is limited for many applications because of its performance degradation. In this paper, we propose a novel multiple-model extended Kalman filter (MMEKF), which remarkably reduced its large deviation. The expansion points designed in the MMEKF algorithm obey the Gaussian distribution in the process of probabilistic models, which are used to approximately represent the whole state space by using multiple probabilistic weighted method. Compared with other filters such as EKF, UKF and CKF, the MMEKF shows higher estimation accuracy for unreliable measurements especially in multi-sensor systems. This paper also considers fault-tolerant distributed data fusion estimation, whose feasibility and effectiveness through a numerical example.