Adaptive Traffic Lights Control System Using Deep Q-Network and Salp Swarm Optimization

Effective traffic light control is critical for reducing congestion and improving urban mobility. This paper presents a novel algorithm combining Salp Swarm Optimization (SSO) with Deep Q-Network (DQN) to enhance decision-making in adaptive traffic management systems. The proposed DQN-SSO approach was evaluated against traditional DQN and Double DQN (DDQN) methods in two scenarios: an uncalibrated scenario with high traffic in localized lanes and a fully crowded scenario with high traffic across all lanes. Results indicate that DQN-SSO achieved substantial improvements, reducing total waiting times by up to 10% compared to DDQN and 12% compared to DQN in the uncalibrated scenario and by 10% and 18%, respectively, in the fully crowded scenario. The integration of SSO’s global optimization capabilities with DQN’s reinforcement learning framework enabled the algorithm to adapt efficiently to varying traffic conditions. These findings highlight the potential of DQN-SSO as a scalable and robust solution for adaptive traffic control in urban environments.