Cooperative-competitive Game-based Method for Local Path Planning with Obstacle Avoidance in Racing Scenarios

In this paper, a cooperative-competitive game based planning method (CCGP) is proposed to address the local path planning with obstacle avoidance in racing scenario. During the racing, the robots in the same group cooperate with each other to achieve the goal, while competing with another group of robots. The goal of each group of robots is to reach the same target point ahead of the other group of robots, and there is no collision with other robots and static obstacles. This cooperation competition relationship is described as a cooperation competition game in theory. The objective function of each robot is formulated as an optimization problem, and the collision avoidance in robot-robot and robot-obstacle is regarded as constraints. The rationality of the proposed objective function in the local path planning problem is theoretically analyzed. Then MPC is used iteratively to compute the approximate optimal local trajectory of each robot under Nash equilibrium by considering the best response of other robots, without collision with other robots and obstacles. We compare our method with the MPC method by conducting racing simulations in a scene with static obstacles. The results show that our method has better performance, and the robots in our group can obstruct and overtake robots of the MPC group, while they avoiding collision with teammates and static obstacles.