The Resesrch Of Particle Filter Algorithm

Particle filter algorithm is implemented by a Bayesian recursion process though a Monte Carlo simulation method.As a nonlinear filter algorithm based on Bayesian estimation, particle filter has an original advantage at treating the parameter estimation and stating filtering aspects of nonlinear non-Gaussian time-varying system.So particle filter has been widely used in the fields such as fault diagnosis,navigation position,radio communication,vision tracking and finance etc.At first,this paper makes a detailed analysis for Particle filter algorithm and summarys it’s shortcomings.And then, we present the improved algorithm.The paper’s research and innovative can be summarized as the followings:Through the study of the particle filter algorithm based on Unscented kalman filter(UKF-PF),in order to solve the particles impoverishment problem of UKF-PF. Firstly,we presents a new methods the UKF-PF based on the artificial fish swarm algorithm(AFSA),which uses the AFSA to optimize the resampling process of UKF-PF and makes prior particles move towards the high likelihood region by use of the behaviors of preying,following and swarming in the AFSA,effectively solves the problem of particles impoverishment. Secondly, through the study of the UKF-PF based on AFSA. Aiming at the large amount of calculation leads to the problem of real time, we proposes a new algorithm the UKF-PF based on the modified AFSA(MAFSA).This method has an important improvement in behaviors,parameter and process,effectively solves the real-time problem of the UKF-PF based on the traditional artificial fish swarm algorithm.The experimental results show that the proposed algorithms in this paper not only can solve the problem of UKF-PF, but also the filter accuracy is improved by one order of magnitude.