Abstract: To solve the particle impoverishment caused by resampling in particle fileter (PF), the Chicken Swarm Optimization (CSO) was integrated into the sampling phase of generic particle filter and an intelligent optimized particle filter of CSO was proposed. According to the weight of focused particles as the fitness, the type of each particle in the population and interrelation between each one was determined. Various designed mechanisms about individual movement were introduced to update the location. Moreover, the dynamical structure of particle population was utilized to overcome weakness of local optimum and improve the optimization. On the basis, particles moved towards to the high likelihood region of posterior probability density. As a result, the diversity of samples was kept and quality of particles was ameliorated. The result of simulation experiment shows that this algorithm has higher estimation accuracy and keeps the diversity of particles, and reduces the quantity of particles required by the state estimation.

Key words: chicken swarm optimization, particle filter, particle impoverishment, state estimation