Variable-Step-Length Hybrid A* Based on Dichotomy Optimization for Path Planning of Autonomous Mining Trucks*

Global path planning for autonomous mining trucks in shovel-loading areas requires path optimality and high computational efficiency. However, generating qualified paths for the loading process is more difficult for conventional methods under accurate final pose constraints. Furthermore, due to large-scale maps with variable obstacle distribution, considerable computation time needs to be allocated for path planning when using conventional methods. To address this problem, a novel Variable-Step-Length Hybrid A* based on Dichotomy Optimization (DO-VSLHA*) algorithm is proposed to generate obstacle-free paths considering mountain morphology and vehicle constraints while reducing computation time. To avoid U-shape obstacles and unnecessary node search, the clustering method is applied to the grid map to generate convex polygon obstacles. Subsequently, with joint sampling of step length and steering angle, we put forward dichotomy optimization based on the cost function in order to generate near-optimal nodes in each loop of the node expansion process, thus reducing the overall computing time of path planning. Field experiments are carried out on autonomous mining trucks at an open-pit mine, validating the improvement in effectiveness and computational efficiency of our method compared to conventional methods.