A UAV-based coverage gap detection and resolution in cellular networks: A machine-learning approach

The use of unmanned aerial vehicles (UAVs) to extend the coverage of terrestrial base stations (BSs) in cellular communication systems has been gaining increasing interest in recent years. This is due to the ease of deploying UAV-mounted BSs within relatively short times with low associated costs. In this work, we formulate the problem of deploying UAV-mounted BSs to mitigate the coverage gaps of the terrestrial BSs of cellular networks in some geographic regions. We then devise a technique that provides the optimal bound of the solution to the coverage gap detection and mitigation problem. We also propose a machine-learning (ML) based technique to provide a real-time solution for deploying UAVs to determine and mitigate the coverage gaps. In both solutions, namely, the optimal and ML-based solutions, the deployment of the UAVs is done in such a way that addresses the coverage gaps at the minimum possible cost. Simulation results show that our ML-based technique performs quite closely to the performance of the optimal solution, at a significantly lower complexity, and hence fulfills the real-time requirements of such deployments. It also provides significantly better performance results than a common benchmark solution from the literature.