Intelligent Traffic Management System Using Ant Colony and Deep Learning Algorithms for Real-Time Traffic Flow Optimization

Abstract

Urban traffic congestion presents a formidable global challenge that necessitates innovative and adaptive solutions, surpassing the capabilities of traditional traffic management systems. This research introduces an Intelligent Traffic Management System (ITMS) that synergistically integrates Ant Colony Optimization (ACO) and Deep Learning (DL) methodologies, effectively optimizing real-time traffic flow. To dynamically adapt to complex urban environments, the ITMS leverages ACO for agile routing and DL for precise traffic prediction, enabled by a novel Long-Short-Combination (LSC) framework designed to accommodate both congested and uncongested traffic attributes. Real-time data acquisition is achieved using a computer vision model, which detects and classifies vehicles into four categories (cars, bikes, buses, and trucks) with updates every 15 minutes. Data preprocessing addresses inconsistencies to ensure integrity. The ITMS employs ACO to optimize vehicle routing dynamically by simulating artificial "ants" that evaluate routes based on pheromone levels representing congestion and distance, thus adapting to real-time fluctuations. Reinforcement learning dynamically adjusts traffic signal timings, minimizing congestion and optimizing overall traffic flow. Six Machine Learning models were tested, finding a weighted average precision, recall, and f1-score of 0.95. More specifically, for traffic situation classification, a detailed model performance analysis was conducted, revealing that Class 0 achieved a precision of 0.99, recall of 0.98, and F1-score of 0.99. Class 1 achieved a precision of 0.90, recall of 0.87, and F1-score of 0.88. Class 2 achieved a precision of 0.93, recall of 0.96, and F1-score of 0.95, and Class 3 had a precision of 0.96, recall of 0.96, and F1-score of 0.96. These results highlight the transformative potential of AI-driven traffic optimization.