Fourth Edition of the International Conference on Research in Applied Mathematics and Computer Science ICRAMCS 2022
March 24-25-26, 2022
Online and Face-to-Face Conference

ICRAMCS, 4 (2022) | Proceedings ISSN: 2605-7700

Research Communication | Open Access
Volume 2022 | Communication ID 409
A hybrid reinforcement learning and cellular automata model
Salma El Bakkal, Abdallah Lakhouili, Essouffi El Hassan
Academic Editor: Youssef EL FOUTAYENI
Received
Accepted
Published
January 31, 2022
March 03, 2345
April 15, 2345

Abstract: Traffic light systems in cities are a very important topic to study due to many challenges such as traffic congestion, traffic accidents and gridlock. In this contribution, we propose a new model for traffic flow in city networks based on cellular automata and deep reinforcement learning, modifying the basic ideas of the most known Biham, Middleton and Levine (BML) model [1]. We show that the traditional BML model can be formulated as a Markov Decision Process. And using the deep Q-learning [2] our approach learns policies that can minimize the traffic jam at a group of intersections. An ...