User Coverage Maximization for a UAV-mounted Base Station Using Reinforcement Learning and Greedy Methods

This paper proposes two methods to maximize the coverage of distinct ground users by an unmanned aerial vehicle (UAV) -mounted mobile base station: a deep reinforcement learning (DRL) approach, and a reward-based greedy approach. The performance of the proposed methods are evaluated based on two aspects: the number of distinct ground users covered, and the delay experienced by a user until it first receives coverage. The distribution of ground users is modelled as a Gaussian Mixture Model (GMM) with fixed and time-varying means with the latter scenario mimicking the mobility of users. Simulation results show that both proposed methods lead to efficient coverage and latency performance with the DRL based approach significantly outperforming the rewards-based greedy algorithm, especially when ground users are allowed to be mobile.