Estimating the Odometry Error of a Mobile Robot during Navigation

This paper addresses the problem of the odometry error estimation during the robot navigation. The robot is equipped with an external sensor (like laser range finder). Concerning the systematic error an augmented Kalman Filter is introduced. This filter estimates a vector state containing the robot configuration and the parameters characterizing the systematic component of the odometry error. It uses encoder readings as inputs and the readings from the external sensor as observations. The estimation of the non-systematic component is carried out through another Kalman Filter where the observations are obtained by two subsequent robot configurations provided by the previous augmented Kalman Filter. Both synchronous and differential drive systems are considered.