A Distributed 3D UAV Placement Algorithm for Integrated Ground-Air Cellular Networks

The problem of deploying unmanned aerial vehicles (UAVs) to form a wireless cellular network has gained much attention. Most of the recent works either need to know the locations of users or to make decision in a centralized way. In this paper, we design a distributed learning-based 3D UAV placement approach, named SDQ-H. Given a set of UAVs, SDQ-H builds an integrated ground-air cellular network to maximize the number of ground users (GUs) covered, without requiring knowledge of geographic locations of GUs. SDQ-H decouples the 3D UAV placement problem into two sub-optimal optimization problems. First, we utilize a Q-learning algorithm to optimize UAV positions in 2D space, i.e., in longitude and latitude. UAVs are modeled as agents, each of which takes action independently according to its position and reward. Then, we finetune the position of UAVs in the third dimension, i.e., height. The above two optimization processes are carried out alternately until convergence. We conduct simulations on several different user distributions. Simulation results demonstrate the superiority of SDQ-H.

I. Introduction

In recent years, the rapid development of wireless technology has provided new solutions for solving some specific problems in communication scenarios. In order to satisfy the growing number of requirements of users, UAVs equipped with base stations have gained attention in the field of wireless communication systems due to the characteristics of easy deployment, mobility, high altitude, and less cost. As air base stations, UAVs can be deployed in relatively high positions, providing ground users with a communication link with a higher probability of line-of-sight (LoS) [1], moreover, they can improve wireless network performance in scenarios such as natural disasters, field surveys, and traffic congestion. When the terrestrial network is damaged by an accident, UAVs can quickly set up a temporary reliable communication network for users on the ground. Unlike fixed ground base stations, which are expensive to deploy and cannot move, UAV can help the ground cellular network to extend the coverage to remote areas, and release the burden of overloaded ground base station traffic in large-scale activity scenarios.