Yin, Jianhua (Wuhan University of Technology), Zhan, Xianyuan (Tsinghua University), Ji, Tianying (Tsinghua University), Xu, Bingrong (Wuhan University of Technology), HE, Yi (Wuhan University of Technology), Zhang, Daqing (Sunward Intelligent Equipment Co., Ltd.), Li, Lingxi (Indiana University-Purdue University Indianapolis)

Motion Planning Integrated with Vehicle-Terrain Interactions for Off-Road Autonomous Ground Vehicles

Scheduled for presentation during the Regular Session "Motion planning" (ThBT9), Thursday, September 26, 2024, 14:30−14:50, Salon 17

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

Off-road autonomous ground vehicles (AGVs) work in unstructured environments with varied elevation and deformable ground, where vehicle-terrain interactions are more complex resulting in unpredictable dynamics behavior. Current sensor-based hazards/obstacle identification methods mainly focus on geometric hazards/obstacles with explicit geometry, which might be adequate for vehicles in on-road conditions. However, in the off-road environment, vehicle-terrain interactions could introduce non-geometric hazards that cannot be detected by current geometric hazards-oriented sensing techniques, e.g. vehicle stuck in mud. To enable AGVs to traverse the target terrain and meet the goal while avoiding non-geometric hazards, this work proposes a motion planning method integrated with coupled vehicle-terrain dynamics response. Specifically, terramechanics-based modeling and simulation (M&S) is utilized to characterize vehicle-terrain interactions. Based on the dynamics response, a reinforcement learning (RL) algorithm with a specialized reward function is used to navigate the AGV to find the goal with high mobility and safety. We demonstrate the proposed method with two case studies, 1) high mobility is the foremost concern with less safety requirement, and 2) both high mobility and safety are required. The results show that the proposed motion planning method is effective for navigating the AGV in off-road terrains.