More recently, the advancement of reinforcement learning brings increasing attention in transport areas. Several studies have utilized them to address the challenge of adaptive traffic signal control (TSC) and have obtained encouraging outcomes. In this paper, a new adaptive traffic signal control algorithm is proposed based on reinforcement learning (RL) to alleviate congestion in urban corridors. To associate with congestion and signal control plans, first, the method employs the negative value of queue lengths for links in the urban corridor as the reward and takes the queue lengths at several time points as the state features. Compared to traditional traffic flow indicator, queue length is able to identify the deteriorating trend of traffic conditions after intersections become saturated, which helps in early detection of traffic congestion trends. Then, adjusting the green split at each intersection is chosen as the action to control congestion levels by adjusting the traffic flow at the upstream and downstream intersections. Through the SUMO simulation platform, the proposed signal control method is comprehensively evaluated at different levels of traffic demand and several RL algorithms have been compared. It has been proved that the proposed signal control approach not only can be trained efficiently on a single low-performance GPU but also can reduce congestion significantly.
Adaptive Traffic Signal Control for Urban Corridor Based on Reinforcement Learning
Mechan. Machine Science
International Conference on Computational & Experimental Engineering and Sciences ; 2024 ; Singapore, Singapore August 03, 2024 - August 06, 2024
2024-12-03
11 pages
Article/Chapter (Book)
Electronic Resource
English
Automated Adaptive Traffic Corridor Control Using Reinforcement Learning
Transportation Research Record | 2006
|Automated Adaptive Traffic Corridor Control Using Reinforcement Learning: Approach and Case Studies
Transportation Research Record | 2006
|Automated Adaptive Traffic Corridor Control Using Reinforcement Learning: Approach and Case Studies
British Library Conference Proceedings | 2006
|Automated Adaptive Traffic Corridor Control Using Reinforcement Learning: Approach and Case Studies
Online Contents | 2006
|