A Hierarchical Planning for Autonomous Racing in Multi-Vehicle Competitive Scenarios

This paper presents a two-layer planner for the autonomous racing trajectory planning problem in multi-vehicle competitive scenarios, which consists of two parts: global planning and local planning. We perform offline global planning using an optimization-based approach to find the global trajectory with the minimum lap time. Online local trajectory planning is performed using an approach based on game theory to avoid vehicle collisions while accomplishing blocking, overtaking, and following maneuvers. A parameter estimation method is proposed to estimate the driving style of the opponent vehicle. Statistical results under simulation conditions show that our method has obvious advantages over ordinary game theoretic planner. Finally, we carried out a hardware experiment using the wheeled mobile robot in an indoor environment to prove the effectiveness of the algorithm.