Moving Aerial Base Station Networks: A Stochastic Geometry Analysis and Design Perspective

• Published in: IEEE transactions on wireless communications 2019-06, Vol.18 (6), p.2977-2988
• Main Authors: Enayati, Saeede, Saeedi, Hamid, Pishro-Nik, Hossein, Yanikomeroglu, Halim
• Format: Article
• Language: English
• Subjects: Aerial base station; Base stations; binary point process; Energy consumption; Geometry; Low speed; Power consumption; Radio equipment; stochastic geometry; Stochastic processes; Throughput; Trajectory; trajectory processes; Wireless communication
• ISSN: 1536-1276; 1558-2248
• DOI: 10.1109/TWC.2019.2907849

Recently, the utilization of aerial base stations (ABSs) has attracted a lot of attention. For the static implementation of ABSs, it has been shown that if the ABSs are statistically distributed in a given height over a cell, according to a binomial point process (BPP), a fairly uniform coverage across the cell is achievable. However, such a static deployment exhibits poor performance in terms of average fade duration (AFD) for the static or low speed moving users and power consumption. Therefore, considering a network of moving ABSs is of practical importance. On the other hand, once such a moving ABS network is considered, the coverage probability may not necessarily remain at an acceptable level. This paper is concerned with the design of stochastic trajectory processes such that if according to which the ABSs move, in addition to improving the AFD, an acceptable coverage profile can be obtained. We propose two families of such processes, namely, spiral and oval processes, and analytically demonstrate that the same coverage as the static case is achievable. We then focus on two special cases of such processes, namely, radial and ring processes, and show that the AFD is reduced about two orders of magnitude with respect to the static case. To obtain a more practical scenario, we also consider deterministic counterparts of the proposed radial and ring processes and show that similar coverage and AFD as the stochastic case can be obtained.