FedHC: A Hierarchical Clustered Federated Learning Framework for Satellite Networks

TL;DR

FedHC addresses the challenge of efficient federated learning in highly dynamic, resource-constrained satellite networks by introducing a two-stage hierarchical clustering approach. It employs a satellite-clustered PS selection algorithm to enable parallel, in-cluster model aggregation and a ground-station stage for global updates, augmented by a meta-learning-driven re-clustering mechanism to rapidly adapt to changing cluster memberships. Experimental results on satellite testbeds and standard datasets show FedHC achieves significant reductions in processing time (up to 3x) and energy consumption (up to 2x) while maintaining accuracy, outperforming centralized and baseline clustered FL methods. This work advances practical, energy-efficient FL for global satellite networks and opens avenues for privacy-preserving extensions such as differential privacy.

Abstract

With the proliferation of data-driven services, the volume of data that needs to be processed by satellite networks has significantly increased. Federated learning (FL) is well-suited for big data processing in distributed, resource-constrained satellite environments. However, ensuring its convergence performance while minimizing processing time and energy consumption remains a challenge. To this end, we propose a hierarchical clustered federated learning framework, FedHC. This framework employs a satellite-clustered parameter server (PS) selection algorithm at the cluster aggregation stage, grouping nearby satellites into distinct clusters and designating a cluster center as the PS to accelerate model aggregation. Several communicable cluster PS satellites are then selected through ground stations to aggregate global parameters, facilitating the FL process. Moreover, a meta-learning-driven satellite re-clustering algorithm is introduced to enhance adaptability to dynamic satellite cluster changes. The extensive experiments on satellite networks testbed demonstrate that FedHC can significantly reduce processing time (up to 3x) and energy consumption (up to 2x) compared to other comparative methods while maintaining model accuracy.

Figures (4)

• Figure 1: The network model with federated learning.
• Figure 2: Overview of the proposed framework FedHC.
• Figure 3: MNIST
• Figure 4: CIFAR-10