The SLAM backend graph optimization based on traditional PDL algorithm has the disadvantages of instability and poor adaptability. The step update strategy based on Cauchy point algorithm and the trust region update strategy based on adaptive scaling factor are proposed in this paper, and they are combined to obtain the improved PDL algorithm. The improved PDL algorithm not only solves the problem that the step iteration cannot continue when it encounters the singularity of the approximate Hessian matrix, but also improves the adaptive ability of the algorithm by making full use of the information of gain ratio and current iteration step. The trajectory error analysis and optimization time analysis are performed on the TUM dataset, and the real scene tests are carried out with a mobile robot, which verified the improved algorithm has better stability and adaptive ability.
SLAM Backend Graph Optimization Based on Improved PDL
2021-10-20
1276994 byte
Conference paper
Electronic Resource
English
Landmark based radar SLAM using graph optimization
IEEE | 2016
|ExplORB-SLAM: Active Visual SLAM Exploiting the Pose-graph Topology
Springer Verlag | 2022
|