Boosting wireless systems with UAVs: how to learn from a quality and energy perspective

The adoption of Unmanned Aerial Vehicles (UAVs) as mobile Base Stations is a promising solution to boost the capacity in hotspot areas and to provide energy aware services in different smart environments.

The adoption of UAV-BSs involves the planning of their missions over time with the aim to provide specific services to ground users, which includes the scheduling of recharging actions of each UAV-BS at ground sites. This is of particular interest in two rising future technologies: Heterogeneous Cellular Networks using 5G and beyond techniques, and Internet of Things. The former leverages the utilization of mmWave technology on UAVs that recently gained attention due to high available bandwidth. The bandwidth budget can be provided to support demanding services, such as uplink live video streaming, or for edge computing purposes, such as those required seamless user interaction with real and virtual objects throughout extended reality multimedia services. As for the IoT, transmitting the huge amount of sensing data through UAVs may on a side alleviate the cellular network from a massive data collection and on the other side entail less power at the IoT end devices.

In the literature,  Artificial Intelligence and Machine Learning based UAV flight planning algorithms aimed at improving energy efficiency as well as service quality related metrics are provided. Specifically, several algorithms leverage Q-learning and Reinforcement learning algorithms by introducing rewards related to key Quality of Service (QoS) and users’ Quality of Experience (QoE) metrics.

We will explore the potential for using UAVs BS in 5G/6G HetNet as well in cyber-physical systems. Emphasis will be placed on recent results using the Q-learning approaches, considering also several exciting emergent research directions paving the way towards novel applications such as eXtended Realty Multimedia Communications, Smart Health, Internet of Everything.