Unscented Particle Filtering with Particle Swarm Optimization for Estimating Nonlinear System

A new Unscented Particle Filter incorporating Particle Swarm Optimization for estimating nonlinear systems state is proposed. The proposed method employs an intelligence optimization approach to mitigate sample degeneracy and impoverishment, and the computation complexity is also reduced. Studies are conducted: through comparing particles' present fitness value with the optimum fitness value of objective function, PSO makes particles with insignificant weights of UPF move towards to the higher likelihood region, and then finds the optimal position where particles with larger weights. Results are promising, especially indicate that the state estimation precision of PSO-UPF is superior to the traditional UPF algorithm and offers an improvement performance compared with PF.