Integrating FedDRL for Efficient Vehicular Communication in Smart Cities
Mancini, Lorenzo Labbi, Safwan Abed Meraim, Karim Boukhalfa, Fouzi Durmus, Alain Mangold, Paul Moulines, Eric
Mancini, Lorenzo
Labbi, Safwan
Abed Meraim, Karim
Boukhalfa, Fouzi
Durmus, Alain
Mangold, Paul
Moulines, Eric
Supervisor
Department
Machine Learning
Embargo End Date
Type
Book chapter
Date
2025
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
Vehicle-to-everything (V2X) communication technology is changing the way we move. It allows vehicles, devices, and infrastructures to interact, overcoming traditional limitations, and enabling smart mobility. V2X technologies aim to enhance road safety, transportation efficiency, energy savings, and driver assistance systems, thus being an important milestone in the development of smart cities. To improve the reliability and efficiency of these technologies, researchers and practitioners are increasingly turning to Deep Reinforcement Learning (DRL). This chapter offers an introduction to DRL in V2X, and its synergy with Federated Learning (FL). It starts by explaining the principles of DRL, where vehicles learn themselves which behavior to follow. A strong focus is put on deep policy gradient and actor-critic methods. These methods are crucial in reinforcement learning and rely on using deep neural networks to find good policies and evaluate them. FL, a collaborative machine learning paradigm that promotes collective learning, is also introduced. The fusion of FL and DRL leads to Federated Deep Reinforcement Learning (FedDRL), offering scalable solutions to modern V2X challenges. Federated Deep Reinforcement Learning (FedDRL) is then applied to the use-case of access point selection for communication in Vehicle-to-Everything (V2X) technologies. These experiments demonstrate the potential of combining Deep Reinforcement Learning (DRL) and Federated Learning (FL) to advance V2X technology. This offers intelligent, adaptable, and collaborative mobility solutions for the future.
Citation
L. Mancini et al., “Integrating FedDRL for Efficient Vehicular Communication in Smart Cities,” Lecture Notes in Intelligent Transportation and Infrastructure, vol. Part F99, pp. 431–450, 2025, doi: 10.1007/978-3-031-72959-1_19.
Source
Lecture Notes in Intelligent Transportation and Infrastructure
Conference
Keywords
Federated learning, Intelligent transportation systems, Internet of vehicles, Policy gradient methods, Reinforcement learning, Smart cities, V2X, Vehicular networks
Subjects
Source
Publisher
Springer Nature