A novel odometry model for wheeled mobile robots incorporating linear acceleration

Abstract

One of the fundamental problems in mobile robotics is simultaneous localisation and mapping (SLAM), where one or more robots build a map and localise within the map at the same time. Many mobile robots have differential drive wheels with pneumatic tyres. These robots are dependent on odometry pose estimation when performing map building or localisation tasks. Currently a three parameter odometry estimation model is widely used [1-2] consisting of the left and right wheel radii and wheel separation distance. This paper introduces a new fourth parameter that accounts for wheel slip proportional to linear acceleration of the robot. To the authors' knowledge, this is the first robot odometry model that accounts for linear acceleration in a differential wheeled robot. The model requires little extra computation and accelerations can be obtained from existing robot motion commands. We experimentally validate our new odometry model on a hard lino floor. Experimental results show an error range of 9% in the estimation of the wheel radii when a Pioneer2 DX H-8 indoor mobile robot accelerates and decelerates on straight paths at an acceleration value of ±0.45 m / s <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> . We use another independent experiment to show a 0.3% improvement of the quality of the occupancy grid (OG) map [3] built using the three parameters model. The maps were generated by attaching laser scans to robot poses based only on odometry estimation. For fairness in the number of parameters, we also compare our model with another four parameter model. The experimental results also demonstrate two methods for calibration of the new model.