Performance analysis of an aerial assisted cellular network

Abstract

An Aerial Assisted Cellular Network (AACN) using Unmanned Aerial Vehicles (UAVs) as Aerial Base Stations (ABSs) is envisioned as a cost-effective way to provide network connectivity especially in rural and remote areas. Performance analysis of AACNs often considers dense deployment of ABSs, ignores low power ground BSs and does not quantify the effect of ABSs on ground BSs. Additionally, interference mitigation for AACNs deserves attention because of the strong LineOf-Sight (LoS) links from the ABSs. This thesis, therefore, formulated a downlink analytical model for the coverage probability, average rate and energy efficiency of an AACN. It also compared the performance of three Joint Transmission Coordinated Multipoint (JT CoMP) schemes and Dynamic Point Selection (DPS) for interference mitigation. Since ABSs are size, weight and power limited, it was imperative to investigate low complexity Coordinated Scheduling/Coordinated Beamforming (CS/CB) techniques for a multi-antenna system under imperfect channel state conditions. An analytical model based on Poisson Point Processes (PPPs) and Monte Carlo (MC) simulations were used to establish the performance of an AACN with low density of ABSs in a rural environment. The analytical results were closely matched by the simulation results hence validating the analytical model. Both results show that the coverage probability initially increases with ABS altitude due to increasing probability of LoS, reaches an optimum and then drops due to the high path loss. At a Signal-to-Interference Ratio (SIR) threshold of -15 dB and UAV density of 2 / km2 , the optimum ABS altitude is 100 m which gives a coverage probability of 0.97. However, at 100 m, the optimum density is 5 / km2 with coverage probability of 0.98. The simulation results show that interference aware JT CoMP gives the best performance compared with fixed number and fixed region. It inherently chooses the strongest interferers and includes them in the cluster. Three low complexity CS/CB techniques with proportional fair scheduling are compared. Block diagonalisation using Householder transformation (BD HH) gives the lowest overall complexity and strong resilience to CSI error. Therefore, it is the most pragmatic beamforming technique for an AACN. Through performance analysis for an AACN, performance bounds are established and inherent trade offs revealed. The results will contribute to the establishment of UAV operational parameters developed by UAV manufacturers, network operators, network regulators and airspace regulators to ensure their safe and efficient integration into cellular networks. Keywords: Aerial Base Stations, Block Diagonalisation, Coordinated Multipoint, Coverage Probability, Unmanned Aerial Vehicles