An Efficient Surrogate Modeling Method for Reliability-Based Global Path Planning of Off-Road Autonomous Ground Vehicles

Abstract

Abstract The off-road autonomous ground vehicle operation is challenging because of the complex and highly uncertain working environment they operate in. This brings risks to their mission success. Physics-based global path planning combining physics-informed surrogate modeling with path planning is recently employed to address this issue by ensuring a reliability to complete the mission. However, the current two-stage methods require a relatively high computational cost to obtain the reliability map undermining the practicability for real-life applications. To decrease the computational cost, this study proposes an efficient surrogate modeling framework for physics-based path planning under uncertainty. The proposed method couples adaptive surrogate modeling with path planning sequentially through several iterations of path planning and path verification. Instead of obtaining an accurate reliability map for the whole target area in the two-stage approach, the proposed method only refines the map in the vicinity of the path identified, thereby reducing the computational cost. The final path is determined if the path verification is successful; i.e. the specified mission reliability for the path is met. The presented case study demonstrates that the proposed method can significantly reduce the number of surrogate refinement iterations and reduce the computational cost for physics-based path planning.