UKF Sensor Data Fusion for Localisation of a Mobile Robot

Matthias Baumann, Daniel Eck, László Lemmer, Klaus Schilling

Abstract: The localisation of outdoor mobile robots is one of the most important challenges for implementing applications such as search and rescue, reconnaissance, surveillance and monitoring. The Global Positioning System (GPS) is a common used sensor system for localisation but the drawbacks of its limited accuracy are well known. These effects can cause mission failure especially for small sized mobile robots. To compensate these drawbacks, a sensor data fusion is introduced based on an Unscented Kalman Filter (UKF) that fuses GPS, inertial and incremental sensor data in an adaptive way. In case of GPS outages the typical INS drift can be avoided by a new alternative position update, which is calculated based on the last pose and the kinematic model that uses incremental encoder and yaw rate sensor data. The whole system is implemented on a low power micro controller.