Deep Reinforcement Learning for User Access Control in UAV Networks

Unmanned Aerial Vehicles (UAV) are recently proposed as the flying base stations (BS), namely, UAV-BS, to boost the capacity as well as extend the coverage of the current wireless network. Nevertheless, the mobility of UAV-BS leads to a highly dynamic network environment, which poses new challenges on the access control of the users on the ground. Two key challenges are the acquisition of the global network information at each user and frequent handovers. Motivated by recent works, which show preferable performance of the deep reinforcement learning (DRL) in the dynamic network environment, we propose a DRL framework to help the user make access decisions. With the proposed DRL framework, the user is able to intelligently make access decisions and maximize the long-term throughput with partial network information, meanwhile avoid frequent handovers. Simulation results have validated the effectiveness of the proposed algorithm and shown the advantage of the proposed DRL framework over the state of arts.