Mobility-Aware Decentralized Federated Learning with Joint Optimization of Local Iteration and Leader Selection for Vehicular Networks

TL;DR

The paper tackles mobility-induced dynamics in vehicular network FL by proposing Mobility-Aware Decentralized Federated Learning (MDFL) with a joint Local Iteration and Leader Selection Optimization (LSOP). LSOP is reformulated as a Dec-POMDP and solved via Multi-Agent Proximal Policy Optimization (MAPPO), enabling two agent types to coordinate local iterations and leader choices using centralized value estimation. Extensive simulations with SUMO mobility data and FashionMNIST/LeNet demonstrate MDFL’s superiority over Random and DFL baselines, highlighting energy-efficiency, convergence behavior, and robustness to non-IID data, with Scaffold-style leader strategies performing best under heterogeneity. The work’s Dec-POMDP/MAPPO combination offers a scalable, mobility-aware approach for practical vehicular FL, with demonstrated generalization to traffic-flow prediction in the PeMS dataset.

Abstract

Federated learning (FL) emerges as a promising approach to empower vehicular networks, composed by intelligent connected vehicles equipped with advanced sensing, computing, and communication capabilities. While previous studies have explored the application of FL in vehicular networks, they have largely overlooked the intricate challenges arising from the mobility of vehicles and resource constraints. In this paper, we propose a framework of mobility-aware decentralized federated learning (MDFL) for vehicular networks. In this framework, nearby vehicles train an FL model collaboratively, yet in a decentralized manner. We formulate a local iteration and leader selection joint optimization problem (LSOP) to improve the training efficiency of MDFL. For problem solving, we first reformulate LSOP as a decentralized partially observable Markov decision process (Dec-POMDP), and then develop an effective optimization algorithm based on multi-agent proximal policy optimization (MAPPO) to solve Dec-POMDP. Finally, we verify the performance of the proposed algorithm by comparing it with other algorithms.

Figures (14)

• Figure 1: MDFL framework.
• Figure 2: Timeline of MDFL.
• Figure 3: Communication model of vehicles.
• Figure 4: The relationship of agents.
• Figure 5: Comparison of $F^{acc}$ with respect to initial vehicle energy level.