基于代价地图的非结构化道路参考线生成算法

A multi-layer cost map model based on grid-based cost mapping is proposed in this paper to ad⁃ dress the challenges of unstructured roads characterized by undulating terrain， uneven surfaces， and the absence of distinct boundary markers， which integrates slope cost， roughness cost， and boundary cost. A multi-stage algorith⁃ mic framework of search-optimization-sampling-optimization is designed： initial paths are generated and smoothed in grid maps through an enhanced Astar algorithm， followed by the generation of motion primitive sample sets in the Frenet coordinate system. Smooth global reference paths compliant with vehicle kinematics are produced via a two layer optimization mechanism that combines dynamic programming and quadratic programming. The real vehicle ex⁃ periments show that the proposed method can plan smooth global paths in unstructured road environment that satisfy kinematic constraints， exhibiting excellent ride comfort while fully complying with vehicle kinematic requirements.