Collaborative Machine Learning for Energy-Efficient Edge Networks in 6G

Abstract

To fulfill the diversified requirements of the emerging Internet of Everything (IoE) applications, the future sixth generation (6G) mobile network is envisioned as a heterogeneous, ultra-dense, and highly dynamic intelligent network. Edge intelligence is a vital solution to enable various intelligent services to improve the quality of experience of resource-constrained end users. However, it is very challenging to coordinate the independent but interrelated edge nodes in a decentralized learning manner to improve their strategies. In this article, we propose a decentralized and collaborative machine learning architecture for intelligent edge networks to achieve ubiquitous intelligence in 6G. Considering energy efficiency to be an essential factor in building sustainable edge networks, we design a multi-agent deep reinforcement learning (DRL)-empowered computation offloading and resource allocation scheme to minimize the overall energy consumption while ensuring the latency requirement. Further, to decrease the computing complexity and signaling overhead of the training process, we design a federated DRL scheme. Numerical results demonstrate the effectiveness of the proposed schemes.