Mobility-Aware Multi-Task Decentralized Federated Learning for Vehicular Networks: Modeling, Analysis, and Optimization
Dongyu Chen, Tao Deng, He Huang, Juncheng Jia, Mianxiong Dong, Di Yuan, Keqin Li
TL;DR
This work tackles multi-task federated learning in highly dynamic vehicular networks by introducing a mobility-aware decentralized framework (MMFL) that jointly optimizes task scheduling, subcarrier allocation, and leader selection (TSLP). It establishes theoretical results, including a single-task convergence bound and the existence of a Nash equilibrium for multi-task resource allocation, and recasts the problem as a DEC-POMDP solved via a HAPPO-based algorithm. The proposed approach demonstrates superior accuracy and stability over strong baselines across realistic mobility and communication settings, validating its potential for privacy-preserving, multi-task ITS applications. Overall, the paper provides a rigorous modeling, analysis, and algorithmic pipeline that links mobility, communication constraints, and multi-task learning to deliver efficient, scalable FL in vehicular networks.
Abstract
Federated learning (FL) is a promising paradigm that can enable collaborative model training between vehicles while protecting data privacy, thereby significantly improving the performance of intelligent transportation systems (ITSs). In vehicular networks, due to mobility, resource constraints, and the concurrent execution of multiple training tasks, how to allocate limited resources effectively to achieve optimal model training of multiple tasks is an extremely challenging issue. In this paper, we propose a mobility-aware multi-task decentralized federated learning (MMFL) framework for vehicular networks. By this framework, we address task scheduling, subcarrier allocation, and leader selection, as a joint optimization problem, termed as TSLP. For the case with a single FL task, we derive the convergence bound of model training. For general cases, we first model TSLP as a resource allocation game, and prove the existence of a Nash equilibrium (NE). Then, based on this proof, we reformulate the game as a decentralized partially observable Markov decision process (DEC-POMDP), and develop an algorithm based on heterogeneous-agent proximal policy optimization (HAPPO) to solve DEC-POMDP. Finally, numerical results are used to demonstrate the effectiveness of the proposed algorithm.