Resource Efficient Asynchronous Federated Learning for Digital Twin Empowered IoT Network

As an emerging technology, digital twin (DT) can provide real-time status and dynamic topology mapping for Internet of Things (IoT) devices. However, the implementation of DT within IoT networks necessitates substantial, distributed data support, which 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 considers both energy consumption and latency by optimizing IoT device selection and transmit power control, while adhering to long-term performance constraints derived from the convergence upper bound for asynchronous FL. We utilize the Lyapunov method to decompose the formulated problem into a series of one-slot optimization problems and design a two-stage optimization algorithm to achieve optimal transmit power control and efficient IoT device scheduling. In the first stage, we derive closed-form solutions for optimal transmit power on the IoT device side. In the second stage, to address the uncertainty in partial state information, the edge server leverages a multi-armed bandit (MAB) framework to model the IoT device selection problem and employs an efficient online algorithm, the client utility-based upper confidence bound (CU-UCB), to solve it. Moreover, we analyze the algorithm’s optimality gap and time-averaged regret, further demonstrating its effectiveness in resource allocation. Numerical results confirm the superior performance of our algorithm compared to benchmark methods, with the CU-UCB algorithm achieving an 18.9% reduction in average cost and a 6% improvement in accuracy on real-world datasets.