Slip parameter estimation for tele-operated ground vehicles in slippery terrain

Operating a ground vehicle remotely is a cooperative activity between the vehicle and the operator. Existing tele-operated/driven vehicles suffer from poor traversability in slippery terrain during task execution. Improving vehicle traversability will facilitate their efficiency when manoeuvring in difficult environments. Slip is important for determining traction forces and handling characteristics when vehicles traverse slippery terrain. This paper presents a slip-estimation approach using an optical-flow technique and a non-linear observer/filter. An onboard downward-looking camera acquires image sequences of the terrain surface during vehicle motion. By tracking salient natural features on terrain surfaces, the optical-flow algorithm provides estimates of vehicle pose and velocity. Using estimates from the optical-flow algorithm and wheel angular velocity obtained from wheel encoders, the observer estimates longitudinal slip values and the slip angle. The proposed approach has been evaluated on test rigs and a skid-steering mobile robot. A comprehensive comparison has been made between a sliding mode observer (SMO) and an extended Kalman filter (EKF) in terms of estimation accuracy, convergence speed, robustness, and computational costs. The SMO is shown to have good accuracy and fast convergence speed comparable to that of the EKF; moreover, the SMO does not require knowledge of noise statistics, and computes faster than the EKF.