| The paper analyzes GNSS and INS navigation systems and builds a mathematical model of the GNSS/INS integrated navigation system.The commonly used filtering algorithms such as UKF,PF,and UPF are further studied,based on the advantages and disadvantages of UKF and PF.Two improved methods based on different importance resampling optimization strategies are proposed,namely AFSA-UPF algorithm based on steepest descent method and RBUPF algorithm based on diffusion sequence importance resampling.The AFSA-UPF algorithm based on the steepest descent method achieves rapid optimization and maintains the diversity of particles through the combination of cluster intelligence algorithms,optimization methods,and other methods,and effectively solves the particle depletion problem of UPF.The RBUPF algorithm based on the diffusion sequence-based importance resampling strategy aims at solving the large computational defect of the UPF algorithm.It adopts dimension reduction to decompose the subspace.Each subspace selects the optimal algorithm to solve the problem,which greatly reduces the computational complexity and supplements the diffusion.Sequence importance resampling strategies to prevent particle degradation.Both algorithms are fused by a variety of methods across multiple domains.According to the actual needs of nonlinear complex positioning problems,they are designed to use the advantages of multiple algorithms to improve the deficiencies of UPF algorithms,and to use the complementary advantages of various algorithms to improve Fusion.The classical nonlinear,non-Gaussian system target tracking model verifies that the two improved fusion filtering algorithms show good results in alleviating particle degradation,reducing computational complexity and improving positioning accuracy.In order to further demonstrate the conclusions of this paper,the simulation of real train operation data in GNSS/INS integrated navigation system was designed and simulated by matlab.Through the analysis of simulation graphs,algorithm running time,error and other comparative data,it is verified that the two improved algorithms have higher performance in practical applications. |
