Multi-Path-Planning in Complex Intersection Based on Improved Hybrid A* and Minimum Jerk Optimization

Pre-definition of paths in intersections is an essential step in intersection management for autonomous vehicles, and it is difficult to ensure safety and smoothness of paths of all entry-exit pair of a complex intersection while obeying traffic rules. However, only a few researches cover this part and existing path/trajectory planning methods can hardly be implemented directly in the scenario, especially in terms of complex intersections. This paper is aimed to solve multi-path planning in complex intersection scenarios based on the improved hybrid A* - minimum jerk pipeline, which could adaptively generate collision-free, smooth and traffic-rule-complying paths for each entryexit pair in various complex intersections, ensuring safety and trackability. After single-path and multipath simulations in different intersections with comparison with two common path/trajectory generation methods, we tested the performance of our method and verified its feasibility and advantages.