AI-Powered Traffic Signal Control for Lower Emissions in Smart Cities

This study presents a privacy-sensitive traffic signal control system based on Deep Q-Networks (DQN) aimed at reducing carbon emissions in urban dense scenarios by minimizing vehicle waiting time at road intersections. The system utilizes data from the city infrastructure (non-sensitive data) while addressing privacy concerns. We validate the model's effectiveness using a testing framework that includes various reward function models, training scenarios, and traffic conditions. Preliminary results indicate that during peak hours, the system can reduce vehicle waiting times at intersections by up to 50%. This work serves as a reference for developing intelligent and sustainable transportation systems.