Speaker: 苏荣教授，新加坡南洋理工大学

Inviter: 李怡蓓

Title: Real-Time Traffic Light Control for Large Traffic Networks

Language: Chinese

Time & Venue: 2025.07.08 10:15-11:15 南楼205

Abstract: Traffic congestion in urban areas significantly increases the commuting time for passengers and introduces unnecessary fuel burns and carbon emissions to the fragile urban ecosystem. Simply increasing the road capacity has been proven futile in solving congestion problems. Optimally balancing demands and road capacities dynamically is believed to be one promising solution among a bundle of other solutions, which, together, holds the key to solving the congestion problems. Traffic light control is one major control means to achieve such balance. In this talk, I will briefly introduce some recent works done in my group on real-time traffic light control, which, by leveraging a proper network volume dynamic model, can effectively predict how vehicle volumes vary in the network. Upon such predictions, a computationally efficient optimization strategy is proposed to update the signal plan for each traffic light in a large network in real time. Relying on learning techniques, this method can automatically update model parameters. By dynamically allocating computational power to regions that need optimization most, scalability can be effectively attained. Simulations show that this proactive traffic light control method can significantly improve the traffic network performance such as reducing the total travel delays and increasing average speed, while leveraging an existing network infrastructure.