Whitepaper: Artificial Intelligence Empowered Multiple Access for Beyond 5G Networks
As the wireless world moves towards the B5G era, the need of additional bandwidth to satisfy the data rate requirement of killer applications, turned the attention of both industry and academia towards higher frequency bands, such as the terahertz (THz), creating the vision of THz networking. This shift to THz wireless networks comes with several challenges; especially, in the physical and medium access control (MAC) layers design, as a result of the directional nature, line-of-sight requirement, and ultra-dense deployment of THz networks. From several studies, it became clear that user association and resource allocation tactics need to be rethought to incorporate artificial intelligence (AI), which can offer “real-time” answers in challenging contexts that change often. Novel mobility management technologies are also necessary to meet the requirements for ultra-reliability and low latency of some B5G applications.
This white paper presents a comprehensive MAC layer strategy that enables intelligent user association, resource allocation, flexible mobility management, and blockage minimization, while maximizing system reliability. Specifically, a novel metaheuristic-machine learning (ML) framework is proposed, which enables quick and centralized joint user association, radio resource allocation, and blockage avoidance. This framework improves the performance of THz networks, while reducing association latency by roughly three orders of magnitude. A deep reinforcement learning (DRL) strategy for beam-selection is explored to assist mobility management and blockage avoidance inside the access point (AP) coverage region. Finally, a proactive hand-over system based on quick channel prediction with AI assistance is provided to allow user movement across the coverage zones of neighboring APs.
View full paper here
The publication was created by ARIADNE project (www.ict-ariadne.eu), which has received funding from the European Horizon 2020 Programme under grant agreement number 871464.
