An Efficient Planning Framework for Ground Vehicles Navigating on Uneven Terrain

Autonomous navigation on uneven terrain is significant for ground vehicles to perform complex tasks. This paper proposes an efficient planning framework that achieves safe and fast navigation on 3D uneven terrain. The framework utilizes probabilistic elevation mapping method and SLAM (Simultaneous Localization and Mapping) techniques to construct a global elevation map, while using terrain normal vectors and roughness measures to capture the undulations and roughness of the ground vehicle's path. Subsequently, a terrain-kinematics interaction-based path search method (TK A ∗) is proposed. This method fully leverages the interaction between the vehicle model and the terrain to effectively expand 3D path nodes on the elevation map. Additionally, a terrain-based traversal cost is designed to reflect the vehicle's ability to travel on non-planar surfaces, while an efficient heuristic cost is devised to improve search efficiency.