Preliminary Design of a Test Bench for Terrain Classification and Soil Parameters Identification in Deformable Terrains

Accurate terrain identification represents a crucial challenge to improve safety and performance of vehicles in unstructured environments. This work presents the preliminary design of a reconfigurable test bench for terrain recognition and soil parameter estimation through a physics-informed ML approach. The system is developed based on existing semi-empirical models and features a well-structured set of proprioceptive and exteroceptive sensors to measure all relevant variables describing the wheel-ground contact phenomenon. Rather than being limited to a qualitative and quantitative analysis, the system lays the foundation for the development for scalable traction and stability control in autonomous off-road navigation.