An experimental implementation or sensor fusion for mobile robots based on kalman filtering

In this paper, an experimental investigation of the Kalman filtering technique aimed at implementing sensor fusion for mobile robots is presented. A mobile robot is usually equipped with encoders mounted on the wheels axes in order to control the motion and to estimate its position via dead-reckoning algorithms. It is well known, however, that this suffers from drift and errors due to an imperfect calibration. In case of vision-based navigation, the vehicle might be equipped with a video-camera as well. However, image acquisition generally works at a lower frequency than odometry, the camera might be then used to obtain a second measurements of the displacement and fused with the vehicle´s odometry. Experimental results with a Magellan PRO mobile robot equipped with a low cost vision system are finally given