Improved Comprehensive Learning Particle Swarm Optimization Algorithm And Neighborhood Field Optimization Based On Resampling Step Of Particle Filter

Since the searching accuracy of the standard PSO algorithm is not high and its convergence rate gets slow in the later searching stage,the resampling step of particle filter is introduced into the later search process of PSO algorithm.To prevent the identity of particles,the resampling step is followed by the existing method of particle variation.The improved PSO algorithms is compared with its standard algorithms and other six improved PSO algorithms through extensive simulations under seven different benchmark functions.The experimental results show that the convergence rate and search accuracy of the improved PSO both upgrade,and the improved algorithm is able to solve multi-modal problems globally.In order to verify this new technique which is the resampling step of particle filter combining with mutation is somewhat general,and it can improve other particle-based optimization algorithms and the improved algorithms have better performance.Choose two typical kinds of optimization algorithm,one is the comprehensive learning particle swarm optimization algorithm(CLPSO),it is a typical improved particle swarm optimization algorithm and a global optimization algorithm with high reference;Another is neighborhood optimization algorithm,it utilizes the local information with the principle of “learning from neighbors”.Both also have a shortcoming that their convergence rate get slow in the later searching stage.In the same way the new technique is introduced into the later periods of the algorithm introduced,at the same time using five and seven benchmark functions to test respectively.The simulation results show that the improved comprehensive learning particle swarm optimization algorithm and the improved neighborhood optimization algorithm achieve significantly better performance respectively,in terms of the accuracy of the solution,the global search ability and the stability and robustness.Furthermore,based on the simulation results,it can be thought that the proposed technique can be some-what general on the performance of improved intelligent algorithm based on particle.