Smart traffic light system with multi-head attention deep Q-learning for AI-driven vehicle prioritization

Kavina Shree, Ganason (2026) Smart traffic light system with multi-head attention deep Q-learning for AI-driven vehicle prioritization. Final Year Project, UTAR.

PDF Download (8Mb)

Abstract

Heavy traffic in populated areas leads to significant breakdowns in the performance of traditional traffic control systems. AI and adaptive traffic control systems are used to minimize these challenges. This project investigates the development of an advanced Smart Traffic system with Multi-head Attention Deep Q-Learning for AI-driven vehicle prioritization. The goal of this project is to offer a simulation-based traffic control system that adjusts real-time traffic flow and prioritizes emergency vehicles. The system uses real-time data to make effective decisions. Emergency vehicle prioritization is a key component to ensure that the vehicles are given instant clearance to proceed through intersections. In this era, many individuals face difficulties reaching their destinations on time due to traffic delays. This results in irregular traffic distribution, where one lane is packed with vehicles while another remains empty. Moreover, current traffic management systems are often too slow at adapting the signal conditions. In this project, traffic conditions will be monitored by detecting the number of vehicles in each lane. Priority is given to lanes with the highest vehicle count. This system will continuously make real-time adjustments based on traffic simulations. In addition, the vehicle priority management system can detect emergency vehicles and override regular traffic control rules to allow them to pass first. To address the traffic congestion problems featured in this project, we are developing an advanced Smart Traffic system with an AI-driven solution through this final year project to improve traffic signal control and reduce delays caused by irregular traffic flow. By merging emergency vehicle prioritization and real-time vehicle detection, this system aims to enhance traffic signal timing and improve overall road efficiency in high traffic urban zones.

Actions (login required)