Addressing the issues of insufficient adaptability and limited energy efficiency optimization capabilities in traditional tunnel lighting control methods under complex traffic conditions, this paper proposes a dynamic dimming strategy for tunnel lighting based on the Proximal Policy Optimization (PPO) algorithm.First, the tunnel lighting system is modeled as a reinforcement learning environment. A state space integrating multi-dimensional information—including traffic flow, vehicle speed, external brightness, and tunnel section location—is constructed, and a continuous action space is designed to enable precise dimming control for each functional section. Based on this, a multi-objective reward function is established that integrates brightness tracking error, energy consumption optimization, control stability, and environmental adaptability to guide the agent in learning the optimal dimming strategy.Subsequently, model training and experimental validation were conducted using actual tunnel operation data.Experimental results indicate that, compared to traditional L20 control strategies, the proposed method achieves smoother brightness regulation and higher zone control accuracy while ensuring driving safety and visual comfort, and demonstrates significant energy-saving advantages during periods of high lighting demand. In summary, the dynamic dimming strategy based on the PPO algorithm shows promising application prospects and engineering value in intelligent tunnel lighting systems.