Latency Optimization for Wireless Federated Learning in Multihop Networks

2025-06-08

IEEE Transactions on Vehicular Technology

AI Summary

To address significant data heterogeneity, high end-to-end latency, and poor node coordination in multi-hop wireless federated learning (FL), this paper proposes the Personalized Adaptive Federated Learning (PAFL) framework. PAFL is the first to jointly optimize leaf-node scheduling, relay-node resource allocation, and dynamic routing selection in multi-hop FL, while incorporating relay-side energy harvesting modeling to enhance system sustainability. A non-convex joint optimization problem is solved via block coordinate descent and successive convex approximation (SCA). Furthermore, personalized model aggregation and adaptive weight updates are integrated to mitigate data heterogeneity. Experimental results demonstrate that, compared to single-tier node optimization, greedy algorithms, and routing-agnostic baselines, PAFL reduces system end-to-end training latency by up to 69.37%, substantially improving both efficiency and practical deployability of multi-hop FL.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Machine Learning: Distributed Machine Learning & Federated Learning
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

In this paper, we study a novel latency minimization problem in wireless federated learning (FL) across multi-hop networks. The system comprises multiple routes, each integrating leaf and relay nodes for FL model training. We explore a personalized learning and adaptive aggregation-aware FL (PAFL) framework that effectively addresses data heterogeneity across participating nodes by harmonizing individual and collective learning objectives. We formulate an optimization problem aimed at minimizing system latency through the joint optimization of leaf and relay nodes, as well as relay routing indicator. We also incorporate an additional energy harvesting scheme for the relay nodes to help with their relay tasks. This formulation presents a computationally demanding challenge, and thus we develop a simple yet efficient algorithm based on block coordinate descent and successive convex approximation (SCA) techniques. Simulation results illustrate the efficacy of our proposed joint optimization approach for leaf and relay nodes with relay routing indicator. We observe significant latency savings in the wireless multi-hop PAFL system, with reductions of up to 69.37% compared to schemes optimizing only one node type, traditional greedy algorithm, and scheme without relay routing indicator.

Problem

• Minimize latency in wireless federated learning across multi-hop networks
• Address data heterogeneity via personalized and adaptive FL framework
• Optimize leaf and relay nodes with energy harvesting for efficiency

Innovation

• PAFL framework harmonizes individual and collective learning
• Joint optimization of leaf and relay nodes
• Energy harvesting scheme for relay nodes