When graphs meet game theory: a scalable approach for robotic car racing

Ahmet Tikna, Marco Roveri, Daniele Fontanelli, Luigi Palopoli

Abstract: Autonomous vehicle racing is facing a growing interest both in industrial and academic settings spanning multiple disciplines. In this paper, we will explore how to create a robust, efficient, and reliable decision-making mechanism to decide, at every point in time, the trajectories that a vehicle should take to overtake its opponents and win the race. The proposed framework combines a graph-based path planner with a game-theoretic model to generate powerful racing strategies. We implemented the framework, and we carried out an experimental evaluation to show its effectiveness and evaluate the impact of the different parameters.