Traffic signal timing in saturated mode using reinforcement learning

Articles in Press

Document Type : Original Article

Authors

1 Professor-School of Civil Engineering-Iran University of Science and Technology

2 Department Transportation, Faculty of Civil Engineering, Iran University of Science and Technology, Tehran, Iran

3 Iran University of Science and Technology

4 Iran University of Science and technology

10.22034/tri.2023.396364.3152

Abstract

Today, with the expansion of urbanization, the need for a dynamic transportation system is felt more than ever. For this reason, to achieve a stable and orderly system, the control of transportation networks is considered essential. Although the modeling of networks today has become a complex and difficult issue and it faces problems in modeling to be closer to the environmental conditions, in the meantime, the framework of reinforcement learning as a model-independent method can play a better role in controlling and provide us with traffic simulation. In this study, we tried to use different reinforcement learning algorithms, such as DQN and DDPG algorithms, to simulate the considered traffic network in a faster and more regular way, and to be able to determine the influencing factors such as queue length. formed in the streets and traffic lights by using algorithms and proper planning, in a new way to reduce the amount of traffic and to optimize it, and according to the results obtained from the two mentioned algorithms, an algorithm that We propose that it had a better performance as the superior algorithm from the subset of reinforcement learning algorithm and finally our network by reducing the queue length and also reducing the amount of time spent behind traffic lights in urban networks in saturated state, which as a result improves passing and Review and smooth the flow of traffic.

Keywords

Main Subjects

Journal of Transportation Research

Articles in Press, Accepted Manuscript
Available Online from 01 August 2023

Files

Share

How to cite

Statistics