Optimal Coverage Enhancement for Multiple UAVs Using Multi-agent Learning Technique

Abstract

Robustness of terrestrial cellular wireless networks becomes challenging in times of disasters such as earthquakes. This paper studies the deployment of multiple Unmanned Aerial Vehicles (UAVs) above the earth's surface to provide ubiquitous connectivity to under-laid users on earth. We provide an analytical framework using tools from stochastic geometry to model the UAV-user equipment (UE) network. We specifically model the UAVs in a finite three dimensional (3-D) space with their associated UEs as the marks on a two dimensional (2-D) earth surface. Tractable expressions for the UE's received signal strength and signal-to-interference plus noise ratio (SINR) are derived in Nakagami fading environments. A new paradigm in the study of UAV cellular communication is also developed in this work with a multi-agent learning technique. With this technique, the UAVs learn from each other by communicating, as well as interacting with their environment to provide optimal coverage. Our numerical results show that our method drastically reduces the interference from adjacent UAVs leading to improved coverage in terms of SINR values. Also, the results show that, UAV deployed wireless network provides better coverage compared to conventional terrestrial base station (BS) deployment.