Hierarchical Federated Learning for 6G Edge Computing

Purpose and goal:

As 6G networks introduce massive IoT deployments and advanced AI-driven applications, optimizing resource utilization becomes critical. In this project, we develop a hierarchical federated learning scheme that enables seamless integration and effective participation across a wide range of devices with various resource constraints. This will allow devices—from small-scale IoT sensors to powerful mobile units—to contribute meaningfully to the learning process.

Expected results and effects:

Our federated learning scheme effectively accommodates a wide range of devices with varying resources, enhancing the scalability and efficiency of 6G edge computing. By the end of this project, we will deliver demonstrations and papers on the proposed scheme, including scalability and efficiency analyses. Our project will contribute to leveraging previously untapped computational resources, ultimately enabling the development of more diverse and intelligent AI solutions within 6G.

Approach and implementation:

Our project first proposes two approaches, “Device-centric clustering” and the “Heterogeneous cluster model,” to handle a more diverse and massive array of devices with various resource constraints. We then introduce a “Dual-mode aggregation” approach that considers both 6G and federated learning characteristics to optimize the efficiency of machine learning model training.

These approaches will be developed in parallel and eventually integrated into one system to maximize effectiveness.