Enhancing Urban Traffic Flow: An AI-Driven Traffic Light Management System

As urban traffic congestion continues to rise, intelligent control systems are becoming essential for efficient transportation management. This paper explores an AI-driven approach to traffic light control that balances the needs of both vehicles and pedestrians. The proposed design, implementation, and evaluation of the proposed system aim to optimise urban traffic flow and minimise waiting times for vehicles and pedestrians. In this paper, the authors propose three pipelines: an epsilon-greedy AI controller, an Upper Confidence Bound (UCB) exploration controller, and a Boltzmann exploration controller to manage traffic flow at an intersection with the help of a dataset containing real-time data. A deterministic controller was used as a baseline for comparison. The results demonstrated that while the Boltzmann controller minimized vehicle stops most effectively, it increased pedestrian waiting times substantially. The epsilon-greedy AI controller achieved a balanced performance, making it the most suitable for real-world scenarios. Further areas for improvement include integrating real-world vehicle and pedestrian movement in the environment to increase the model's generalizability.