Reinforcement Learning for a Cellular Internet of UAVs: Protocol Design, Trajectory Control, and Resource Management

Abstract

Unmanned aerial vehicles (UAVs) can be powerful Internet-of-Things components to execute sensing tasks over the next-generation cellular networks, which are generally referred to as the cellular Internet of UAVs. However, due to the high mobility of UAVs and the shadowing in the air-to-ground channels, UAVs operate in an environment with dynamics and uncertainties. Therefore, UAVs need to improve the quality of service (QoS) of sensing and communication without complete information, which makes reinforcement learning suitable to be employed in the cellular Internet of UAVs. In this article, we propose a distributed sense-and-send protocol to coordinate the UAVs for sensing and transmission. Then, we apply reinforcement learning in the cellular Internet of UAVs to solve key problems such as trajectory control and resource management. Finally, we point out several potential future research directions.