Spectrum-Sharing UAV-Assisted Mission-Critical Communication: Learning-Aided Real-Time Optimisation

Abstract

We propose an unmanned aerial vehicle (UAV) communications scheme with spectrum-sharing mechanism to provide mission-critical services such as disaster recovery and public safety. Specifically, the UAVs can serve as flying base stations to provide extended network coverage for the affected area under spectrum-sharing cognitive radio networks (CRNs). To cope with the effects of network destruction in a disaster, we propose a real-time optimisation framework for resource allocation (e.g., power and number of UAVs) for CRNs assisted by UAV relays. The proposed optimisation scheme aims at optimising the network throughput of primary and secondary networks under the stringent constraint of maximum tolerable interference impinged on the primary users. We also propose a deep neural network (DNN) model to significantly reduce the execution time under real-time solution of mixed-integer UAV deployment problems. For both primary and secondary networks, our real-time optimisation algorithms impose low computational complexity, hence, have a low execution time in solving throughput optimisation problems, which demonstrates the benefit of our approached proposed for spectrum-sharing UAV-assisted mission-critical services.

Keywords

• Real-time optimisation
• unmanned aerial vehicle (UAV)
• spectrum sharing
• machine learning
• mission-critical communications