Demo: A Practical Testbed for Decentralized Federated Learning on Physical Edge Devices

2025-05-12

AI Summary

To address feasibility, performance, and energy efficiency of decentralized federated learning (DFL) on resource-constrained edge devices, this work establishes the first real-hardware DFL testbed using Raspberry Pi and Jetson Nano—eliminating reliance on centralized servers while enabling privacy-preserving collaborative training. Our approach innovatively integrates dynamic power monitoring (via INA219), a lightweight P2P communication protocol, and the NEBULA framework to support distributed training across multiple datasets (CIFAR-10 and FEMNIST) on embedded Linux platforms. Experimental results reveal that communication topology density critically impacts both convergence behavior and energy efficiency: dense topologies improve model accuracy by 3.2% and reduce per-round energy consumption by 18%. The system demonstrates stable operation with up to 16 nodes and achieves end-to-end latency under 850 ms, validating its practicality for edge-deployable DFL.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Search and Optimization: Distributed Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Federated Learning (FL) enables collaborative model training without sharing raw data, preserving participant privacy. Decentralized FL (DFL) eliminates reliance on a central server, mitigating the single point of failure inherent in the traditional FL paradigm, while introducing deployment challenges on resource-constrained devices. To evaluate real-world applicability, this work designs and deploys a physical testbed using edge devices such as Raspberry Pi and Jetson Nano. The testbed is built upon a DFL training platform, NEBULA, and extends it with a power monitoring module to measure energy consumption during training. Experiments across multiple datasets show that model performance is influenced by the communication topology, with denser topologies leading to better outcomes in DFL settings.

Problem

• Evaluating decentralized federated learning on physical edge devices
• Addressing deployment challenges in resource-constrained environments
• Analyzing impact of communication topology on model performance

Innovation

• Decentralized FL testbed using edge devices
• NEBULA platform with power monitoring
• Evaluates topology impact on model performance