Sensor Fusion and Model Verification for a Mobile Robot

Abstract

This paper presents the results of modeling, sensor fusion and model verication for a four-wheel driven, four-wheel steered mobile robot moving in outdoor terrain. The model derived for the robot describes the actuator and wheel dy-namics and the vehicle kinematics, and includes friction terms as well as slip. An Unscented Kalman Filter (UKF) based on the dynamic model is used for sensor fusion, feed-ing sensor measurements back to the robot controller in an intelligent manner. Through practical experiments with the robot, the UKF is demonstrated to improve the reliability of the sensor signals signicantly, and the model is seen to show surprisingly good agreement with the practical exper-iments.