The Research On Online Adjustable PF Algorithm Based On Optimized Probability Density

Filtering technology is widely used as a system state estimation method in nonlinear and non-Gaussian systems.A key parameter of particle filter is the number of particles.The more particles the algorithm uses,the closer the distribution of the filter is to the true distribution.The cost of calculations increases dramatically as the number of particles increases.Therefore,it is necessary to reasonably select a certain number of particles to improve the efficiency of filtering.In addition,the current particle filter resampling method is an important technical method to particle degradation.The resampling method can alleviate the particle degradation to a certain extent,but it also causes the particle diversity to decrease.In view of how to improve the efficiency of particle filter and improve the reduction of particle diversity,this paper proposes the following improvements based on the existing particle filter algorithm:Firstly,in order to improve the filtering efficiency,the online method adjustment can reduce the calculation cost and reduce the executive time without affecting the filter precision.This paper uses the convergence assessment to adjust the number of particles.Then the ratio of adjusting the number is the step change rate of the number of particles.In this paper,after performing simulation analysis on different parameter thresholds of unsynchronized long change rate,the optimal step change rate and the optimal parameter range are determined.Secondly,in order to improve the reduction of particle diversity,this paper considers the importance probability density of particle filter.In particle filter,in order to improve the accuracy of particle distribution probability density,an improved particle distribution probability density method based on auxiliary state variables and an improved particle distribution probability density method based on maintaining neighborhood relationship are proposed.The simulation results verify the effectiveness of the proposed method.Finally,the above method is verified by simulation experiments.The MATLAB software is used to compare the algorithm based on auxiliary state variables and neighborhood-based relationship with Bootstrap particle filter algorithm,and the effectiveness of the proposed method is proved.In the optimization of algorithm parameters,the algorithm is evaluated from three aspects:error mean MSE,number of adjusted particles,and execution time.By selecting different step change rates in different parameter ranges for analysis,the optimal parameter range of the filter and the best step rate of change are determined.The particle filter research method in this paper has certain reference value for shortening the running time of the algorithm and improving the estimation accuracy of the algorithm.