A Novel Edge-Enabled Adaptive Traffic Light System Using Nonlinear Optimization

Abstract

Urban traffic congestion remains a critical challenge due to the exponential growth of automobiles, rapid urbanization, and escalating demand for transportation, all of which strain global road infrastructure. Traditional traffic light systems and recent baseline technologies have proven inefficient under dynamic conditions. To address this, we propose an Adaptive Traffic Signal Control (ATSC) system designed to alleviate urban congestion by adjusting signal timing in real-time based on fluctuating demand. Deep Q-learning Network (DQN) recent approaches begin with no previous knowledge and display substantial unpredictability in action selection, which can lead to poor control effects initially. This study proposes an efficient traffic prediction model based on the integration of edge computing near to IoT sensors to ensure accurate real-time information. The proposed Nonlinear Mathematical Adaptive Traffic Light System (NMATCS) algorithm employs an exponential function to dynamically modify traffic signal duration based on road vehicle density, ensuring no time is wasted at any signal phase. The focus of this study is to reduce the Average Waiting Time (AWT) across multiple intersections. The (NMATCS) algorithm was tested in a simulation environment using authentic real traffic data from four intersections in Hangzhou, China, during peak hours. Performance was benchmarked against the state-of-the-art Priority-based Double Deep Q-Learning (Pri-DDQN) method. Our proposed model achieved an (AWT) of 25.2 seconds, outperforming Pri-DDQN (26.0 seconds) at isolated intersections and DQN-Baseline (28.3 seconds), with improvements of 3.08% and 10.95%, respectively. The results demonstrate that edge computing-enhanced IoT architectures ensure a smooth, nonlinear response and effectively address latency, scalability, and real-time decision-making limitations inherent in cloud-dependent systems. The findings highlight the potential of lightweight, edge-enabled adaptive traffic control in improving urban mobility without relying on computationally intensive learning techniques.