A non-degenerate particle filter under the mutation of nonlinear system state

To address particle degeneracy and sample impoverishment in traditional particle filters, this paper proposes an improved algorithm incorporating Particle Swarm Optimization into the sampling stage. The method selects the maximum-weight particle state as a reference before resampling, then relocates particles toward high-likelihood regions using a PSO-based strategy to enhance diversity and mitigate degeneracy. Furthermore, we introduce and theoretically validate a novel computable proposal distribution. Simulation results demonstrate superior performance over the Unscented Particle Filter (UPF) in both accuracy and computational efficiency, along with robust adaptability to nonlinear system state mutations.