People tracking with a mobile robot: a comparison of Kalman and particle filters

Abstract

People tracking is an essential part for modern service robots. In this paper we compare three different Bayesian estimators to perform such task: Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF) and Sampling Importance Resampling (SIR) Particle Filter. We give a brief explanation of each technique and describe the system implemented to perform people tracking with a mobile robot using sensor fusion. Finally, we report several experiments where the three filters are compared in terms of accuracy and robustness. In particular we show that, for this kind of applications, the UKF can perform as well as a particle filter but at a much lower computational cost.