A 2D Laser SLAM Graph Optimization Based On A Position And Angle Partition And Cholesky Decomposition

Abstract

The traditional 2D laser slam back-end graph optimization (LSBGO) is not efficient in some special situations, such as fast scene switching, limited computing time, and limited hardware facilities. This paper presents a partitioned pose vector method (PPVM) to optimize a pose vector by dividing it into two parts: a position term and an angle term. Based on this partition, the traditional graph optimization problem has been transformed into two linear equations. The least square solutions of the two equations help to find the pose vector. Furthermore, the paper applies a Cholesky decomposition (CD) to improve the speed of solving the two linear equations. Cholesky decomposition has great advantages in solving linear equations with symmetric positive definite coefficient matrix. The effectiveness of PPVM-CD is numerically verified by MATLAB simulation. Compared with the traditional LSBGO method, PPVM-CD improves the optimization speed by 26% and the optimization accuracy by 11.1%.

Keywords

• laser slam
• graph optimization
• cholesky decomposition
• matlab