| Zumrud Isgandarli, Ilkin Sharafkhanov, Rufat Ismayilov, Musa Afandiyev, Jamaladdin Hasanov Adaptive Traffic Light Optimization for Dynamic Urban Intersections |
|---|
| Abstract. Traffic congestion is a major urban challenge, and smart cities increasingly adopt Adaptive Traffic Light Systems (ATLS) to address it. In Azerbaijan, traffic lights follow static schedules, requiring manual adjustments. This paper presents an Adaptive Traffic Flow Optimization System using Reinforcement Learning (RL) and Deep Q-Network (DQN) to introduce dynamic responsiveness. The Simulation Urban Mobility (SUMO) simulator is utilized to model real-world traffic sce- narios from Baku. Initially, a single-agent RL model adjusted signal durations based on vehicle-to-capacity ratios, optimizing decisions through Q-learning. Later, the DQN algorithm was implemented to conduct further experiments. The system modernizes Azerbaijan’s traffic infrastructure, aligning it with advanced global solutions. |
| Keywords: Adaptive Traffic Light System, Rein- forcement Learning, Deep Q-Network, SUMO, Traffic Congestion, Q-learning, Smart Cities |
 Download PDF |
| DOI: |
