Adaptive pose and location estimation for indoor mobile robot

In this paper, a new method for mobile robot positioning is proposed. The method is a combination of particle filter (PF) and extended Kalman filter (EKF). Under normal driving situation, EKF is sufficient to estimate the vehicle's pose and location. Subject to external disturbances, EKF does not converge from time to time. PF is then introduced and the switching criteria are governed by the estimation confidence. We cluster the particles of PF into groups at the end of each iteration. The number of clusters is used as one of the parameters to determine whether the PF has converged. In the paper, the formulations and algorithms are illuminated and experimental results are also given and analyzed.