A Motion Planning Framework Based on Stackelberg Games for Autonomous Driving in Interactive Scenarios

A motion planning framework based on Stackelberg games is proposed for autonomous driving in interactive scenarios. Firstly, a hybrid path planner is designed to determine the coupling relationship among traffic participants. Secondly, a speed planner based on Stackelberg games is designed for the autonomous vehicle to interact with dynamic environments. Thirdly, a quantitative leader-follower model is built. Finally, the payoff is obtained through the speed planning under the corresponding strategy. The proposed framework can obtain an optimal strategy through heuristic planning, which integrates the decision and planning modules. It improves the traffic efficiency in intersections, and has the ability to cope with different interactive scenarios in simulations.