A Novel Simulation-Based Framework for Optimizing Traffic Signal Control Using Reinforcement Learning Techniques

Abstract

Traffic congestion remains a critical issue in metropolitan areas, leading to economic inefficiencies and increased travel times. This study introduces a novel simulation-based framework that leverages reinforcement learning (RL) for optimizing traffic signal control. The empirical methodology involves the use of a custom-developed simulation environment in Python (version 3.9) with TensorFlow (version 2.3) and SimPy (version 3.0.11) libraries for modeling traffic flows. Our experiments were conducted across multiple scenarios depicting various traffic patterns and control strategies. The findings reveal a significant reduction in average vehicle delay by 25%, with a corresponding increase in intersection throughput by 18% compared to traditional traffic signal control methods. This research contributes to the ongoing discourse on intelligent transportation systems (ITS) by providing a robust tool for practitioners aiming to enhance urban mobility. Furthermore, the implementation of RL-based methods demonstrates considerable potential for real-time application, paving the way for further exploration in adaptive traffic management systems.