A Deep Learning Method for Predictive Channel Assignment in Beyond 5G Networks

Abstract

In Beyond Fifth Generation (B5G) networks, Internet of Things (IoT) and massive Machine Type Communication (mMTC) traffic are anticipated to be offloaded by multi-hop, Device-to-Device (D2D)-enabled relay networks. The relays offer an energy and spectral-efficient solution to the rising problem of spectrum scarcity and overloading of cellular base stations. Moving beyond the conventional paradigm of the relay nodes employing channels on a specific band at a time, in this article, we aim to investigate how to simultaneously leverage multiple bands at a relay node to improve spectral efficiency. We address the challenge associated with dynamic channel conditions in the multi-band relay networks, and envision a deep learning-based predictive channel selection method to solve the problem. A 1-D (one-dimensional) Convolutional Neural Network (CNN) model is employed to predict the suitable channels across multiple bands with the best Signal-to-Interference-plus-Noise Ratio (SINR). The packets received from the source or previous relay node are scheduled to be transmitted to subsequent relay node/destination based on the best modulation and coding rates to transmit over the predicted band. Our envisioned approach, based on shallow and deep-CNN models, proposes two proactive channel assignment strategies, namely controlled and smart prediction. Our proposal is evaluated with several, comparable machine/deep learning methods. Experimental results, based on datasets, demonstrate encouraging performance of our proposed lightweight deep learning-based proactive channel selection in multi-band relay systems.