Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Resource Efficient Asynchronous Federated Learning for Digital Twin Empowered IoT Network

Shunfeng Chu, Jun Li, Jianxin Wang, Yiyang Ni, Kang Wei, Wen Chen, Shi Jin

TL;DR

The study tackles privacy-preserving DT-based modeling in industrial IoT by integrating asynchronous FL with a dynamic resource scheduler. It presents a Lyapunov-based decomposition to convert long-term energy-latency goals into per-round optimization and uses Lambert W-based closed-form power control alongside a CU-UCB-based device selection strategy. Theoretical analysis provides convergence and regret guarantees for CU-UCB, and simulations show faster training speeds and lower energy-latency trade-offs on CIFAR-10 and Fashion-MNIST compared to baselines. The work offers a practical framework for scalable, privacy-aware DT orchestration in dynamic IoT networks with strong implications for real-time resource management and edge intelligence.

Abstract

As an emerging technology, digital twin (DT) can provide real-time status and dynamic topology mapping for Internet of Things (IoT) devices. However, DT and its implementation within industrial IoT networks necessitates substantial, distributed data support, which often leads to ``data silos'' and raises privacy concerns. To address these issues, we develop a dynamic resource scheduling algorithm tailored for the asynchronous federated learning (FL)-based lightweight DT empowered IoT network. Specifically, our approach aims to minimize a multi-objective function that encompasses both energy consumption and latency by optimizing IoT device selection and transmit power control, subject to FL model performance constraints. We utilize the Lyapunov method to decouple the formulated problem into a series of one-slot optimization problems and develop a two-stage optimization algorithm to achieve the optimal transmission power control and IoT device scheduling strategies. In the first stage, we derive closed-form solutions for optimal transmit power on the IoT device side. In the second stage, since partial state information is unknown, e.g., the transmitting power and computational frequency of IoT device, the edge server employs a multi-armed bandit (MAB) framework to model the IoT device selection problem and utilizes an efficient online algorithm, namely the client utility-based upper confidence bound (CU-UCB), to address it. Numerical results validate our algorithm's superiority over benchmark schemes, and simulations demonstrate that our algorithm achieves faster training speeds on the Fashion-MNIST and CIFAR-10 datasets within the same training duration.

Resource Efficient Asynchronous Federated Learning for Digital Twin Empowered IoT Network

TL;DR

The study tackles privacy-preserving DT-based modeling in industrial IoT by integrating asynchronous FL with a dynamic resource scheduler. It presents a Lyapunov-based decomposition to convert long-term energy-latency goals into per-round optimization and uses Lambert W-based closed-form power control alongside a CU-UCB-based device selection strategy. Theoretical analysis provides convergence and regret guarantees for CU-UCB, and simulations show faster training speeds and lower energy-latency trade-offs on CIFAR-10 and Fashion-MNIST compared to baselines. The work offers a practical framework for scalable, privacy-aware DT orchestration in dynamic IoT networks with strong implications for real-time resource management and edge intelligence.

Abstract

As an emerging technology, digital twin (DT) can provide real-time status and dynamic topology mapping for Internet of Things (IoT) devices. However, DT and its implementation within industrial IoT networks necessitates substantial, distributed data support, which often leads to ``data silos'' and raises privacy concerns. To address these issues, we develop a dynamic resource scheduling algorithm tailored for the asynchronous federated learning (FL)-based lightweight DT empowered IoT network. Specifically, our approach aims to minimize a multi-objective function that encompasses both energy consumption and latency by optimizing IoT device selection and transmit power control, subject to FL model performance constraints. We utilize the Lyapunov method to decouple the formulated problem into a series of one-slot optimization problems and develop a two-stage optimization algorithm to achieve the optimal transmission power control and IoT device scheduling strategies. In the first stage, we derive closed-form solutions for optimal transmit power on the IoT device side. In the second stage, since partial state information is unknown, e.g., the transmitting power and computational frequency of IoT device, the edge server employs a multi-armed bandit (MAB) framework to model the IoT device selection problem and utilizes an efficient online algorithm, namely the client utility-based upper confidence bound (CU-UCB), to address it. Numerical results validate our algorithm's superiority over benchmark schemes, and simulations demonstrate that our algorithm achieves faster training speeds on the Fashion-MNIST and CIFAR-10 datasets within the same training duration.
Paper Structure (19 sections, 4 theorems, 50 equations, 8 figures, 1 algorithm)

This paper contains 19 sections, 4 theorems, 50 equations, 8 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Federated Learning for Digital Twin
  4. Resources-efficient Asynchronous Federated Learning
  5. System Model
  6. Digital Twin for Internet of Things Networks
  7. Asynchronous Federated Learning Based DT Modelling
  8. Computation and Communication Models
  9. Problem Formulation
  10. Single-round Optimization Problem
  11. Solution for Single-round Optimization Problem ($\mathcal{P}1$)
  12. Decisions for IoT Devices Selection
  13. Algorithm Analysis
  14. Simulations
  15. Conclusion
  16. ...and 4 more sections

Key Result

Theorem 1

We assume that the loss function $l$ satisfies Assumption smooth function and weak convexity, and each IoT device trains at most $\hat{D}_{\max}$ local samples before uploading models to the edge server. We also assume that the staleness is bounded by $\Delta$ for all IoT devices in any communicati

Figures (8)

  • Figure 1: The architecture of digital twin empowered IoT networks.
  • Figure 2: The average cumulative energy consumption and latency versus the minimum average training sample quality $D_{\min}$ of the proposed CU-UCB and other baselines.
  • Figure 3: The average cumulative energy consumption and latency versus the energy efficient coefficient $\lambda_{e}$ under the proposed CU-UCB and other baselines.
  • Figure 4: Impact of $\tilde{V}$ on the total queue length of all IoT devices.
  • Figure 5: Test accuracy versus training time for different schemes on the Fashion-MNIST dataset.
  • ...and 3 more figures

Theorems & Definitions (5)

  • Theorem 1
  • Remark 1
  • Theorem 2
  • Theorem 3
  • Theorem 4