Reinforcement learning based adaptive control method for traffic lights in intelligent transportation

Huang, Zhongyi (2024) Reinforcement learning based adaptive control method for traffic lights in intelligent transportation. Alexandria Engineering Journal, 106. 381 - 391. ISSN 1110-0168

Text (Huang_reinforcement-learning-based-adaptive-control--published) - Published Version Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (4MB)

• Scopus publication

Abstract

Addressing the requirements and challenges of traffic light control, a reinforcement learning based adaptive optimal control model for traffic lights in intelligent transportation systems is proposed. In the model design, we combined Markov decision process, Q-learning algorithm, and Deep Q-Learning Network (DQN) control theory to establish a comprehensive signal light Adaptive Optimal Control of Signal Lights in Intelligent Transportation Systems (AOCITL) control model. Through simulation experiments on the model and the application of actual road scene data, we have verified the superiority of the model in improving traffic system efficiency and reducing traffic pressure. The experimental results show that compared with traditional fixed cycle signal light control, the adaptive optimal control model based on reinforcement learning can significantly improve the traffic efficiency of roads, reduce the incidence of traffic accidents, and enhance the overall operational effectiveness of urban transportation systems. The proposed method is possible to further optimize the model algorithm, expand its application scope, and promote the development and practical application of intelligent transportation systems.

Item Type:	Article
Official URL:	https://www.sciencedirect.com/journal/alexandria-e...
Additional Information:	© 2024 The Authors
Divisions:	Geography & Environment
Subjects:	H Social Sciences > HE Transportation and Communications
Date Deposited:	25 Jul 2024 14:06
Last Modified:	15 Nov 2024 03:51
URI:	http://eprints.lse.ac.uk/id/eprint/124348

Actions (login required)

View Item