Mobile Robot Localization via Fusion of Ultrasonic and Inertial Sensor Data

We describe two possible structures for a localization system which should exploit ultrasonic sensor measures as well as inertial and odometric data to maintain a correct estimate of the location of a mobile robot. The objective is to reduce the position and orientation error in the presence of slippage, and, at the same time, to identify the bias of the gyroscope. The two algorithms have the classical predictor-corrector structure of the Extended Kalman Filter, but they differ in the use of the angular velocity measure coming from the gyroscope. Experimental trials, performed on a robotic wheelchair equipped with five wide-beam sonars, allow a comparative assessment of the algorithms.