Estimation of 3D form of the Path for Autonomous Driving in Terrain

Abstract

3D path estimation, i.e., determining the height profiles of the path, is crucial in motion control for autonomous driving in terrains. It is essential to prevent rollover from abrupt peaks, dips, and briefly wet ground. In this study, wheel displacement measurements and height estimation of an off-road vehicle are presented. A calibration process is suggested to measure the instantaneous vertical displacement of each wheel. GNSS-pose-based methods are used to estimate the vehicle's geometry parameters yielding the centimeter-level accuracy of the 3D-path estimation method. The accuracy of the vehicle's instrumentation is examined on a test track to create a 3D terrain model of the path. The outcome of the proposed scheme is compared to a reference elevation profile created with structure from motion on the basis of machine vision and GNSS using data collected by an unmanned aerial vehicle. The comparison of the results demonstrates that the 3D path can be estimated with sufficient accuracy in open terrain using ground vehicles.