Research On GNSS/INS/OD Fusion Position Algorithm Base On Graph Optimization

Multi source fusion positioning technology is currently the main research direction in the field of positioning and navigation.Graph optimization based fusion positioning technology is receiving increasing attention due to the variable sensor components and the ability to achieve asynchronous heterogeneous positioning source fusion.Due to the fact that graph optimization algorithms are essentially optimization problems,there are inevitably problems such as computational complexity and the accumulation of computational complexity with increasing positioning time.Moreover,with the proliferation of tall buildings and increasingly harsh positioning environments,the robustness of graph optimization based fusion positioning algorithms has also been challenged.In response to the above issues,this article focuses on the GNSS/INS/OD fusion localization algorithm based on graph optimization to improve the efficiency and robustness of fusion localization.The main research content includes:(1)A modified IMU/OD pre-integration algorithm is proposed to address the issues of cumbersome and high computational complexity in graph optimization calculation.It constructs pose constraints between adjacent nodes,and can directly use first-order approximation to correct the pre integration results when there is a change in IMU bias or OD scale factor,without the need for recalculation,improving computational efficiency;A hybrid sliding window optimizer was designed,which divides the window into mature and growing regions,and uses nodes within the mature region to update prior factors.This reduces computational complexity while reducing information loss during the marginalization process.Through experiments,it was found that the positioning accuracy of the running trajectory is improved compared to traditional algorithms,provided that the solution time is much shorter than the sampling time.(2)In terms of improving the robustness of fusion localization algorithms,this thesis proposes an adaptive EM method for solving Gaussian mixture error models.Design a nested EM algorithm to simultaneously estimate the state and error model,where the inner EM algorithm estimates the Gaussian mixture error model and the outer EM algorithm estimates the state.Compared with Gauss,c DCE,DCS,SM,and MM algorithms,the experimental results show that the adaptive EM algorithm proposed in this thesis has a much faster solution speed than the sensor information transmission speed in complex environments and has good convergence.Especially,the adaptive EM algorithm-MM has high positioning accuracy in most scenarios.The thesis consists of 31 figures,4 tables,and 89 references.