Multi-Vehicle Coordinated Motion Planning Based on Interactive Primitive Tree

Abstract

Guiding multiple vehicles to leave the conflict zone under consideration of interaction is the core of multi-vehicle coordinated planning. The solution efficiency can be improved by decomposing complex planning tasks into primitives. Therefore, this paper provides a novel coordinated algorithm for general scenes based on an interactive primitive tree. First, the optimization algorithm is applied to generate a motion primitive (MP) library that considers driving behavior and tracked skid-steering vehicle dynamics. Subsequently, multiple vehicles simultaneously extend MPs under environmental constraints and generate the interactive primitive tree in the conflict zone. Finally, the mixed-integer linear programming algorithm is utilized to optimally select the MP sequence to be executed. The experimental results show that our proposed MPs achieve a more reasonable scene-adapted trajectory extension than MPs defined in Hybrid A* and improve the algorithm's efficiency. Besides, in the two typical conflict scenes of road narrowing and intersection, the proposed algorithm demonstrates good coordination ability and significantly reduces the average travel time in the conflict zone. By selecting the MP sequences to be executed from the interactive primitive tree, the coordinated motion-planning algorithm proposed in this paper improves the algorithm's adaptability to different scenes and obtains the planning results related to time and space.