Terrain Adaptive Autonomous Vehicles for Uncertain Off-Road Environments

Project Team

Principal Investigator

Tulga Ersal, University of Michigan

Government

Michael Cole, U.S. Army GVSC

Industry

Jin Ge, Toyota Research Institute

Akshar Tandon, Tesla

Tamer Wasfy, AScience

Student

James Dallas, Congkai Shen, Siyuan Yu, University of MIchigan

Project Summary

Project #1.A89 began August 2020 and completed in Q3 2023.

Four considerations about military autonomous ground vehicles motivate this proposal: (1) Military vehicles are often required to operate on off-road deformable terrains where the vehicle’s mobility is highly dependent on the highly nonlinear tire forces generated at the tire-terrain interface. (2) Terramechanics models are often parameterized by many terrain properties that are unknown or varying during vehicle operation. (3) Increasing the mobility of military autonomous ground vehicles is a critical need. (4) State-of-the-art approaches to navigate such vehicles typically rely on model dependent architectures, such as Model Predictive Control (MPC), and require efficient, dynamic, and twice continuously differentiable mathematical models

The project is based on the vision that autonomous vehicles that are aware of and adaptive to changes in the terrain conditions will offer increased safety and higher performance. Accordingly, the overarching goal of this work is to develop an off-road trajectory planning algorithm that is capable of learning terrain conditions online and modifying decisions accordingly. This constitutes basic research, because while the project focuses on the application of off-road autonomous navigation, the techniques developed can impact other applications where real time decision-making in autonomous systems are influenced by an unknown environment. Furthermore, the techniques developed can in general improve the robustness of MPC and be applied to other autonomous vehicles, for example commercial vehicles operating in on-road winter conditions. Towards this end, the following basic research questions will be answered.

Can a dynamic, accurate, efficient, and twice continuously differentiable terramechanics model be achieved to enable trajectory planning in off-road environments through MPC?
Can such terramechanics models be used in estimators to determine terrain properties with sufficient accuracy for mobility prediction purposes?
To what extent can such terrain estimation improve safety and robustness of off-road autonomous vehicles on uncertain terrains?

1.A89

Publications:

Yu, S., Shen, C., Dallas, J., Epureanu, B. I., Jayakumar, P., & Ersal, T. (2024). A Real-Time Terrain-Adaptive Local Trajectory Planner for High-Speed Autonomous Off-Road Navigation on Deformable Terrains. IEEE Transactions on Intelligent Transportation Systems.
Dallas, J., Cole, M. P., Jayakumar, P., & Ersal, T. (2021). Terrain adaptive trajectory planning and tracking on deformable terrains. IEEE Transactions on Vehicular Technology, 70(11), 11255-11268.
Dallas, J., Jain, K., Dong, Z., Sapronov, L., Cole, M. P., Jayakumar, P., & Ersal, T. (2020). Online terrain estimation for autonomous vehicles on deformable terrains. Journal of Terramechanics, 91, 11-22.
Dallas, J., Weng, Y., & Ersal, T. (2020, October). Combined trajectory planning and tracking for autonomous vehicles on deformable terrains. In Dynamic Systems and Control Conference (Vol. 84270, p. V001T15A001). American Society of Mechanical Engineers.