Multiple model based unscented particle filter algorithm for a SINS/CNS integrated navigation system
Fangfang Zhao, Shuzhi Sam Ge, Wei He · 2017
The single model filter has poor adaptability under the uncertain or unknown system parameters, multiple model filters can be used to resolve this problem. This paper investigates the multiple model adaptive estimation (MMAE) algorithm and proposes a multiple model unscented particle filter (MMUPF) algorithm. Forward the traditional MMAE method which uses extend Kalman filter (EKF) or unscented Kalman filter (UKF) as each filter to execute the state estimation process. In this method, we combine MMAE with unscented particle filter (UPF), and replace UKF with UPF which shows higher accuracy than UKF for the nonlinear system. At the same time, this algorithm is applied to the integrated navigation system which combines strap-down inertial navigation with celestial navigation by a ballistic missile. Simulations are provided to demonstrate that MMUPF algorithm has the highest navigation precision, but at the cost of heavier computational burden.