Game Theory-Based Overtaking-Preventing Controller for Competitive Racing Scenarios

This paper introduces a novel overtaking-preventing controller for competitive racing scenarios, employing a game-theoretical approach. The proposed methodology generates fifth-order trajectories for each agent and utilizes level-k game theory to select the most effective trajectories. Online estimation is conducted to gauge the level of an opponent's mobility, subsequently, the ego mobility chooses the trajectory that can obstruct the opponent's trajectory. The mobilities follow the trajectories through a model-predictive-control approach based on a nonlinear differential drive model. Simulations conducted in a Matlab environment demonstrate the computational efficiency of our approach, with computation times in milliseconds. Our overtaking-preventing controller was tested against various opponent models and proved its effectiveness in terms of blocking success rate in human-involved scenarios.