A state-of-the-art on federated learning for vehicular communications

Abstract

With the increasing number of connected vehicles on the road, vehicular communications have become an important research area. Federated learning (FL), a distributed machine learning technique that enables learning from distributed data sources while preserving data privacy and efficient learning, has the advantages of privacy preservation, low latency, and energy efficiency, making it a suitable technique for vehicular communications.In vehicular communications, where dynamic data privacy, real-time decision-making, and energy conservation are paramount, FL's inherent benefits find unique relevance. The preservation of data privacy is of utmost concern in vehicular networks, where sensitive information is exchanged in real time. FL addresses this concern by allowing vehicles to collaboratively train machine learning models without sharing raw data, thus ensuring the privacy and security of vehicular users. Furthermore, FL's decentralized nature enables low-latency model updates, crucial for Intelligent Transportation Systems (ITS) real-time applications like collision avoidance and traffic management, where split-second decisions can be life-saving. Additionally, FL can contribute to energy efficiency by reducing the need for centralized data aggregation, which is particularly beneficial in resource-constrained vehicular environments.The paper provides a comprehensive survey of the state-of-the-art in FL for vehicular communications, including existing frameworks and architectures, challenges, and applications in various aspects of vehicular communications. The survey analyzes various FL techniques, their strengths and weaknesses, and the performance metrics of FL in vehicular networks. The paper concludes by presenting the open research issues and future directions for FL in vehicular communications. Our survey provides a timely and comprehensive overview of the recent advances in this emerging research field and serves as a valuable reference for researchers and practitioners in both FL and vehicular communications.