Enhancing traffic signal systems in Sri Lanka using artificial intelligence

Abstract

This study investigates the performance improvement potential of an artificial intelligence (AI)-based real-time adaptive traffic signal control system in comparison to a conventional fixed-time signal system. The research addresses the inefficiencies of fixed-time traffic lights in adapting to highly variable traffic patterns, particularly in Sri Lankan urban intersections. Nupe Junction in Matara was selected as the case study location. A real-time adaptive signal model was developed conceptually using vehicle detection logic based on the YOLO object detection algorithm. Microsimulation was carried out in PTV VISSIM to model both fixed time and real-time adaptive systems under four distinct traffic scenarios: weekend noon midday, weekday noon peak, weekday evening peak, and Poya day. Key performance indicators such as queue delay, queue length and travel time were analyzed. These findings suggest that AI-based adaptive control can serve as an effective and scalable solution for improving urban traffic efficiency in Sri Lanka. The study concludes with recommendations for future pilot implementations and highlights the potential of AI integration in smart mobility initiatives.