Game-Theory-Guided Reinforcement Learning for Autonomous Vehicles in Highway Scenarios

Autonomous vehicles have become a popular topic with the development of technology. In the early stages of an automated driving system, the key to correctly guiding the vehicle’s behavior lies in the effective collection and application of expert driving experience. This study introduces a game-theoretic guided reinforcement learning (GTG-RL) framework that merges game-theoretic expert models with reinforcement learning to enhance decision-making efficiency for autonomous vehicles in complex motorway scenarios. Validation confirms the framework’s effectiveness in improving performance and adaptability under dynamic traffic conditions. The results underscore the potential of integrating game-theoretic expertise with reinforcement learning to tackle autonomous driving challenges.