Terrain and Vehicle-Terrain Sensing and Estimation in Real-Time for Use in Autonomous Vehicles

Abstract

Autonomous vehicles are being used more and more every day. While there are many applications for autonomy in off-road scenarios terrain, such as in military and agriculture, most autonomous models focus on on-road applications and do not account for the effect of deformable terrains. This project focuses on obtaining information about vehicle-terrain interaction in real-time, so it can be used in an autonomous model in the future. This is done mainly in two ways: direct sensing and parameter estimation. In the area of direct sensing the tire rut depth left behind by a tire is measured using a system of two Intel RealSense D405 stereo cameras. The rut depth can be assumed to be the same as the tire sinkage in this application and is then used to get the tire entry angle. The tire entry angle can then be used to further obtain the drawbar pull and the tractive effort. The cameras were experimentally tested and validated using the Terramechanics testing rig at the TMVS laboratory at Virgina Tech. Both single pass and multi-pass scenarios were tested and the results analyzed. The terrain tested was GRC-1 lunar simulant, sandy loam, and clay. In the area of parameter estimation, the estimation model of interest is the generalized polynomial chaos extended Kalman filter (gPC-EKF). This filter is used to estimate the vehicle and tire slip angles, as well as the yaw rate using a regression model. Project Chrono was used to collect data from a FED Alpha for the filter.