As the need for spectrum access has dramatically increased over the past decade, the radio spectrum has become a valuable and scarce resource. A great deal of research has been conducted to increase the efficiency of spectrum assignment and utilization. This ranges from methods for resource scheduling among users of the same operator (intra-operator), to approaches for Dynamic Spectrum Access (DSA) among users belonging to different providers (inter-operator). Cognitive Radio (CR) is a DSA-based approach that allows secondary users to sense the spectrum and opportunistically utilize vacant spectrum bands without causing significant interference to licensed users. Furthermore, new radio access technologies (RATs) have been developed by 3GPP for the 5G (fifth generation) mobile networks, such as the 5G New Radio (NR), where new frequency bands are exploited for mobile communications including the sub-6 GHz frequency bands and frequency bands in the mmWave range (24–100GHz).
The use of Machine Learning (ML) and Deep Learning (DL) for optimizing the operation of 5G mobile systems has been investigated in the past few years. ML- and DL-based approaches have been proposed recently for Radio Resource Management (RRM), DSA, and for spectrum sensing. When it comes to the latter, data-driven approaches eliminate the analytical complexity factor and they are model-agnostic. Moreover, deep learning can learn the characteristics of signal features, which can be helpful for spectrum sensing and CR applications.
In this project, we will investigate the use of ML and DL for spectrum sharing applications in 5G systems. This includes different aspects of spectrum sharing such as resource and power allocation, signal feature extraction, spectrum sensing, and DSA.