A Parallel Hybrid Evolutionary Particle Filter for Nonlinear State Estimation

Particle filters (PF) are widely used for state estimation in non-linear and non-Gaussian environments. However, conventional particle filters possess some drawbacks such as sample impoverishment and sample size dependency. In this paper, a novel parallel hybrid evolutionary particle filter is proposed to solve those problems from the perspective of evolutionary computation. In the proposed algorithm, an effort has been made to fuse a genetic algorithm (GA) and particle swarm optimization (PSO) together to improve the standard particle filter. Genetic operators such as crossover and mutation are utilized to maintain the particle diversity and PSO is used to optimize the particle distribution. A parallel scheme is employed to reduce the computation time so it is more suitable to implement by multithreaded programming for real-time system. The simulation results demonstrate the effectiveness of the proposed algorithm.