A DRL-Based Control Strategy for Efficient Highway Entry in Congested Scenarios

Efficient traffic management in high-density urban areas remains a critical challenge, particularly at highway entry ramps and merging points where congestion, sudden braking, and collision risks frequently occur. This chapter addresses the problem of coordinating vehicles entering ramps and intersections by formulating it as an optimization task aimed at minimizing delays, enhancing safety, and improving overall flow. To tackle this challenge, we propose a deep reinforcement learning approach based on the actor–critic framework, where intelligent traffic light agents are trained to dynamically manage vehicle entries. The design of the reward function plays a central role, as it balances safety and efficiency by penalizing excessive braking events while encouraging smooth traffic throughput. Simulations were conducted in a representative high-density traffic scenario located in Bologna, Italy. The results demonstrate that the proposed approach can significantly reduce congestion, lower waiting times, and increase safety margins compared with conventional management strategies. These findings highlight the potential of reinforcement learning techniques to support intelligent traffic control systems and provide a promising direction for future deployment in real-world urban and highway infrastructures.