Development of off-road autonomy for unmanned ground vehicles

Unmanned ground vehicles (UGVs) are increasingly recognized as assets that can benefit military operations, offering the potential to reduce risk to personnel. Having autonomous capabilities on board UGVs can further enhance their utility by reducing operator workload and enabling more mission capabilities.

This report describes the development of a modular autonomy stack for UGVs and its evaluation through experiments in off-road terrain. The autonomy stack integrates three core capabilities – perception, motion planning, and trajectory tracking – implemented on our Milrem Themis UGV platform, Tor. This autonomy system relies on onboard sensors and local mapping to navigate without predefined paths or precise satellite positioning, aiming to address the fundamental challenge of autonomous terrain traversability.

To measure performance, we defined a set of benchmark routes representing diverse terrain types, from open fields and dirt roads to narrow trails and forests. These routes serve as structured test cases for assessing autonomy requirements and identifying limitations of our system. We have conducted experiments to assess our autonomy system at FFI’s LandX demonstrations and in long-range trials. The experiments revealed that Tor with our autonomy stack can successfully navigate simple environments such as open terrain and wide vehicle trails, but struggles in complex scenarios involving dense vegetation and narrow passages. Key limitations include the lack of semantic terrain classification and robust decision making in challenging scenarios. Predictability and reliability remain major challenges, with decisions influenced by randomness rather than consistent reasoning.

While our current autonomy stack is not yet ready for operational use, it provides a solid foundation for continued research and experimentation. The benchmark routes and evaluation methodology offer a repeatable framework for measuring progress, ensuring that future iterations move closer to the ultimate goal: UGVs capable of operating effectively and autonomously in demanding environments.