A Distributed Multi-Agent Intelligence Framework For Autonomous Traffic Coordination In Smart Transportation Systems

Abstract

Rapid urbanization and increasing vehicle density have made efficient traffic management a critical challenge in modern smart cities. Traditional centralized traffic control systems often suffer from poor scalability and delayed responsiveness under dynamic and uncertain traffic conditions. This research proposes Multi-Agent Traffic Signal Coordination Network (MATSC-Net), a distributed multi-agent intelligence framework for autonomous traffic coordination in smart transportation systems. The framework employs a decentralized architecture, where agents representing vehicles and traffic control units make decisions based on local observations while collaboratively optimizing global traffic flow. Smart Transportation Traffic Coordination Dataset of 4,000 is carried out describing traffic conditions, vehicle movement characteristics, road infrastructure attributes, intelligent transportation indicators, and traffic coordination factors. Min–max normalization is applied to scale traffic features into a uniform range and reduce variability across different data sources. A convolutional neural network (CNN)-based extraction method is utilized to effectively capture spatial traffic patterns such as vehicle density, lane occupancy, and road connectivity. The decision-making process of each agent is driven by a Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm, which enables efficient learning in continuous action spaces. Furthermore, an Intelligent Artificial Bee Colony (IABC) optimization technique is integrated to enhance policy optimization, improve convergence speed, and avoid local optima. Experimental results demonstrate that the 98.34% accuracy, 98.12% Precision, and 97.95% recall using python 3.10 and the proposed framework significantly enhances traffic throughput, reduces congestion, and making it a scalable solution for intelligent transportation systems.