An Improved Particle Filter Based on Robustness Factor and Weight Optimization

Abstract

Aiming at the problems of poor robustness and limited precision of UPF (Unscented Particle Filter) in processing nonlinear and non-Gaussian systems, this paper proposes an improved unscented particle filter algorithm based on robustness factor and weight optimization (IUPF). Combining the latest observation information, IUPF uses the more computationally efficient edge unscented Kalman filter to generate the recommended distribution, and increases the robustness factor when gross errors occur. The relatively unscented particle filter effectively avoids the problem of excessive particle weight variance.; At the same time, IUPF introduces a re-sampling method with optimized weights in the re-sampling process, which effectively solves the problem of particle depletion and improves the diversity of particles. Through theoretical derivation and simulation analysis, it can be known that the estimation accuracy of the IUPF algorithm is improved compared with the UPF algorithm, and the robustness is enhanced.