High-Speed, All-Terrain Autonomy: Ensuring Safety at the Limits of Mobility
A novel energy-based constraint and GPU acceleration allow vehicles to handle extreme terrain without rolling over.
A team of researchers from the University of Michigan and the U.S. Army DEVCOM Ground Vehicle Systems Center has developed a breakthrough local trajectory planner for autonomous off-road vehicles. Published on arXiv, the paper "High-Speed, All-Terrain Autonomy: Ensuring Safety at the Limits of Mobility" introduces a novel Model Predictive Control (MPC) formulation designed to solve a critical problem: current autonomous systems either fail to predict and prevent rollovers on rough terrain or are too computationally slow for real-time, high-speed operation.
The core innovation is a new dynamics model specifically derived for vehicles on rough, non-planar terrain, paired with a novel energy-based constraint. This constraint is key—it safely allows for extreme mobility maneuvers, including temporary tire liftoff, without leading to a catastrophic rollover. The researchers analytically prove their formulation mitigates rollover types that many state-of-the-art methods ignore. To achieve the necessary speed for real-time control, the planner leverages parallelized General-Purpose computing on Graphics Processing Units (GPGPU).
The system's performance was validated through extensive simulated trials and full-scale physical experiments. The results demonstrated a clear advantage over a leading baseline method, with the new planner achieving fewer rollovers and a higher rate of successful navigation across multiple challenging scenarios that pushed test vehicles to their absolute mobility limits. This work directly addresses the gap between theoretical off-road autonomy and practical, safe, high-speed operation in unstructured environments.
- Uses a novel energy-based constraint to safely enable extreme maneuvers like tire liftoff, preventing rollovers ignored by other methods.
- Achieves real-time feasibility through parallelized GPGPU computation, making high-speed control on complex terrain possible.
- Demonstrated superior performance in physical tests, with fewer rollovers and more successes than a state-of-the-art baseline.
Why It Matters
Enables practical, high-speed autonomous vehicles for defense, disaster response, and logistics in the world's most challenging terrains.