Adaptive Gaussian particle filter for nonlinear state estimation

The Gaussian particle filter has emerged as a useful tool for nonlinear state estimation problems. The sample size used in the estimation is one of the key factors to the efficiency and accuracy of the filter. However, the fixed sample size which is usually determined empirically may be highly inappropriate since it ignores the varying errors of the processes. This paper presents a sample size adaptive Gaussian particle filter that uses statistical methods and unscented transform technique to estimate the needed sample size in the time update step and the observation update step respectively at each iteration. Simulation results show that the proposed method performs much better than the standard GPF in the nonlinear problems with great accuracy.