Design And Implementation Of Particle Filter Based On Evolutionary Programming And Particle Swarm Optimization

Target tracking technique has been developing rapidly in various fields, and filtering method is one of the most important researches. The state of the system can be estimated by filtering algorithm when the system state transition model and observation model are known. However, the system is usually nonlinear and non-Gaussian in reality, which requires nonlinear filtering algorithm.Particle filter algorithm is an effective nonlinear filtering method. Two improved methods were designed for the target tracking problem. The two proposed improved methods were used to estimate information of the train, including position, velocity and acceleration, and so on. The main work of this paper is summarized as follows:Firstly, the basic mathematical knowledge, the rationale and calculation process of particle filter were researched, and the deficiencies of the algorithm and the existing improved algorithms were summarized.Secondly, the evolutionary programming based on particle filter algorithm was designed to solve the problem of insufficient search accuracy. The ideas of variation and competition selection from the evolutionary programming were introduced to the particle filter, which improved samples’ particle diversity and settle the intrinsic degenerate phenomenon of the particle.Thirdly, in order to solve the problem falling into local convergence, the particle swarm optimization (PSO) based particle filter algorithm was proposed. The idea of particle’s optimization from the PSO was introduced to the particle filter, which made particles approximate the optimal solution and thus improves the estimation accuracy.Finally, the two proposed improved methods used to estimate the state of the train operation were verified in the case of train running status of target tracking problem.