Machine learning-based channel quality prediction in 6G mobile networks

Channel quality is an essential information for management of radio resources in mobile networks. To acquire the channel quality information, pilot (or reference) signals are commonly transmitted, measured, and reported to the network. However, the process of channel quality acquisition is both time and energy consuming. Moreover, the radio resources are competitively shared by the pilot signals and users' data. This motivates an employment of prediction-based approaches determining the channel quality at low cost to avoid over-consumption of resources for pilots. Machine learning is seen as an efficient way to deal with the channel quality prediction, since it allows to reveal usually hidden relations among known and unknown channel quality measurements. In this article, we first overview state-of-the-art works leveraging the time, frequency, and spatial correlations among already known channel qualities and the channel(s), whose quality should be predicted. Furthermore, we outline a framework for a network correlation-based channel prediction enabling to determine the quality of unknown channel between any two communicating nodes by knowing only channels of these two nodes to reference nodes. Then, we demonstrate use-cases and application scenarios for all machine learning-based channel quality predictions. We also assess potential reduction in channel quality measurement-related overhead by all approaches to demonstrate their complementarity and capabilities to support low-overhead and energy-friendly massive deployment of devices in 6G mobile networks.