Abstract

Mobile robots require digital twins to test and learn algorithms while minimising the difficulty, expense and risk of physical trials. Most mobile robots use wheels, which are notoriously difficult to simulate accurately due to friction. Physics engines approximate complex tribology using simplified models which can result in unrealistic behaviors such as inability to turn or sliding sideways down small slopes. Methods exist to characterise friction properties of skid steer vehicles \cite{khaleghian2017technical} but use has been limited because they require expensive measurement equipment or physics models not available in common simulators. We present a new simple protocol to obtain dynamic friction parameters from physical four-wheeled skid-steer robots for use in the Gazebo robot simulator using ODE (Open Dynamics Engine), assuming only that calibrated IMU (Inertial Measurement Unit) and odometry, and vehicle and wheel weights and geometry are available.