Research on Autonomous Ground Vehicle Path Planning in Unstructured Environments

Abstract

As the demand for autonomous ground vehicles (AGVs) in complex unstructured environments continues to grow, their autonomous navigation capabilities are also facing increasingly higher requirements. Unstructured scenarios typically feature irregular terrain elevations, which poses core challenges to traditional path planning algorithms, including insufficient terrain traversability assessment, rough modeling of dynamic constraints, and poor coupling in multi-objective optimization. These challenges can lead to safety risks and execution failures in planned paths. To address these issues, this paper proposes a path planning method that integrates platform physical constraints and multi-dimensional optimization. We construct a multi-dimensional constraint model that includes geometric traversability, motion stability, and power availability, and transform it into hard boundary conditions and dynamically weighted cost terms for path search. Furthermore, we propose an improved A* algorithm-based path planning architecture that, through a dynamic constraint feedback mechanism, achieves closed-loop coupling between theoretical constraints and planning decisions, systematically addressing the challenges of autonomous navigation in unstructured environments. Experimental results show that the comprehensive performance of path planning has improved by more than 8%.