Point Mass Filtering Gravity Matching Method Based on Simulated Annealing Resampling
When utilizing point mass filtering algorithms for gravity-assisted navigation computations,after several iterations,the variance of the importance weights of the particle swarm tends to increase over time.This results in the particle swarm being unable to effectively express the posterior probability density of the state variables.This paper optimizes the resampling process of the point mass filtering algorithm using a simulated annealing algorithm.The optimization involves retaining low-weight particles and introducing new particles through a random probability mechanism.This approach ensures diversity in both particle weights and spatial distribution,effectively reducing the variance of particle weights,increasing the dispersion of particle distribution,and endowing the point mass filtering algorithm with the capability to overcome local optima.Experimental validation demonstrates that the improved algorithm can reduce the average matching position error to within 500 meters within a relatively short period.Evidently,the enhanced algorithm strikes a balance between effectiveness and stability,mitigating the rate of particle degeneration and resolving the issue of subsequent estimation inaccuracies caused by particle degradation.This leads to an enhancement in the positioning accuracy and long-term stability of the gravity matching algorithm.
gravity-assisted navigationpoint mass filteringBayesian estimationresamplingsimulated annealing algorithm
NETL
NSTL
万方数据
