Comparison of EKF based SLAM and optimization based SLAM algorithms

Abstract

This paper compares the recent developed state-of-the-art extended Kalman filter (EKF) based simultaneous localization and mapping (SLAM) algorithm, namely, invariant EKF SLAM, with the nonlinear least squares optimization based SLAM algorithms. Simulations in 1D, 2D, and 3D are used to evaluate the invariant EKF SLAM algorithm. It is demonstrated that in most 2D/3D scenarios with practical noise levels, the accuracy of invariant EKF is very close to that of nonlinear least squares optimization based SLAM. In the simple 1D case, the Kalman filter results and the linear least squares results are exactly the same (for any noise levels) due to the linear motion model and linear observation model involved.