Monte Carlo WLS Fuser for Nonlinear/Non-Gaussian State Estimation

Over the last few decades there has been an explosion of work in the area of nonlinear/non-Gaussian state estimation. Probably the best known methods to solve this problem is extended Kalman filter, unscented Kalman filter, quadrature Kalman filter and particle filter, etc. The majority of algorithms are based on the framework of the Bayesian filter and limited to the additive Gaussian noise. Motivated by these reasons, this paper proposes a novel approach which fuses two kinds of independent state information from the prediction and the observation, respectively. The approach, named the Monte Carlo weighted least squares (WLS) fuser, is not depended on the prediction-correction processes and can be applied to the nonadditive non-Gaussian scenario. The simulation demonstrates the feasibility of the Monte Carlo WLS Fuser and compares the performance of the proposed approach with some cutting-edge methods.