A comparison of several nonlinear filters for mobile robot pose estimation

Abstract

Pose estimation for mobile robots is one of subjects attracting a lot of attention in recent years. In order to remove process and measurement noise of the non-linear/non-Gaussian system, a number of filtering approaches are available: the extended Kalman filter (EKF), the unscented Kalman filter (UKF) and several variants of the particle filter (PF). In this paper, we compare the accuracy and computational load of the EKF, UKF and particle filter (bootstrap algorithm). A mobile robot is simulated. The simulation results indicate that the bootstrap particle filter has the best state estimation accuracy and the most computational cost. The UKF performs almost equivalently with EKF and they both have much less computational cost than the PF.