Decoupled Lateral and Longitudinal Local Path Planning Method Based on LiDAR

The use of autonomous driving technology to improve mining efficiency and safety is currently a focal point of development. This paper addresses the issue of local path planning for unmanned trucks tasked with transportation between loading and unloading points in mining areas. It proposes a method for collision-free, smooth, and efficient path planning based on the decoupling of longitudinal and lateral movements. Initially, the vehicle is transformed from the Cartesian coordinate system to the Frenet coordinate system, and the problem is decomposed into two independent subproblems: longitudinal and lateral. In the longitudinal aspect, the global path from the cloud control platform is captured as the planning result. For lateral planning, a quintic polynomial model parametrized by longitudinal displacement is established. Obstacle information is acquired through the lateral system, and planning results are derived using numerical optimization. Subsequently, the system converts back to the Cartesian coordinate system, where B-spline curve smoothing is applied, and a collision detection function based on the occupied grid lookup table is established to determine the optimal local path.