Estimating vehicle-terrain interaction parameters from tracked-robot sensor data

Abstract

With a wider range of robotic vehicle being deployed with either semi- or full autonomous control, the need to provide useful information about the operational terrain remains essential for reliable operation. In particular, small-scale tracked robotic vehicles are especially found to have widely varying behavior when operating on uncertain and highly variable soil properties. It can be helpful to have information about specific vehicle-terrain parameters that influence traction and resistance to mobility. This paper describes an approach for estimating such terrain parameters (i.e., soil cohesion, shearing resistance, and shear modulus) online, particularly for deformable terrain. By combining an Extended Kalman Filter (EKF) and Newton-Raphson techniques, soil parameters can be estimated using onboard sensor data. Preliminary results from field testing on sandy and clay-like soil terrains show the ability to distinguish between these terrains. These results show promise for implementing online and real-time methods that can inform and guide planning and traction control algorithms.