Energy-efficient device selection and resource allocation for edge-driven hierarchical federated learning
Hierarchical federated learning has emerged as a promising approach to enhance the scalability and efficiency of federated learning in large-scale Internet of Things (IoT) environments. However, existing frameworks often struggle with energy efficiency, adaptability to dynamic conditions, and the co...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | Results in Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S259012302502064X |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Hierarchical federated learning has emerged as a promising approach to enhance the scalability and efficiency of federated learning in large-scale Internet of Things (IoT) environments. However, existing frameworks often struggle with energy efficiency, adaptability to dynamic conditions, and the complex interplay between device selection and resource allocation. In this paper, we propose an energy-efficient hierarchical federated learning framework that jointly optimizes IoT device selection, local resource allocation, and base station resource allocation across a three-layer architecture. We decompose the mixed-integer non-linear programming problem into two subproblems and propose an iterative algorithm that alternates between solving these subproblems until convergence. For the IoT device selection and local resource allocation subproblem, we introduce a multi-agent actor-critic approach enabling autonomous decision-making at the device level. The base station resource allocation subproblem is addressed using a cooperative game-theoretic approach, facilitating distributed coordination among base stations. Our simulation results demonstrate significant improvements in energy efficiency, learning performance, and adaptability compared to baseline methods. |
|---|---|
| ISSN: | 2590-1230 |