Integrating UAV-Enabled Base Stations in 3D Networks: QoS-Aware Joint Fronthaul and Backhaul Design

TL;DR

The paper tackles the challenge of jointly placing drone base stations (DBSs) and designing a backhaul mesh in 3D networks to meet QoS constraints. It introduces an offline two-subproblem framework: an agglomerative hierarchical clustering (HC) method to determine the number and locations of DBSs while satisfying a minimum neighbor degree and a maximum fronthaul radius, and a genetic algorithm (GA) to optimize backhaul interconnections (the Drone Network Problem, DNP) with the aim of maximizing surplus throughput. The authors prove NP-hardness of DNP and demonstrate the effectiveness of HC and GA against baselines, highlighting the operational dynamics and resilience of 3D backhaul meshes under load variability. The approach offers a scalable, modular solution for deploying UAV-based infrastructure in complex 3D wireless networks, with potential applicability to future B5G/6G deployments and disaster scenarios.

Abstract

The emerging concept of 3D networks, integrating terrestrial, aerial, and space layers, introduces a novel and complex structure characterized by stations relaying backhaul loads through point-to-point wireless links, forming a wireless 3D backhaul mesh. A key challenge is the strategic placement of aerial platform such as drone base stations (DBSs), considering the locations and service demands of ground nodes and the connectivity to backhaul gateway nodes for core network access. This paper addresses these complexities with a two-fold approach: a novel Agglomerative Hierarchical Clustering (HC) algorithm that optimizes DBS locations to satisfy minimum backhaul adjacency and maximum fronthaul coverage radius requirements; and a Genetic Algorithm (GA) that designs backhaul connections to satisfy the cumulative load across the network and maximize the throughput margin which translates to network resilience to increasing demands. Our results showcase the effectiveness of these algorithms against benchline schemes, offering insights into the operational dynamics of these novel 3D networks.

Figures (9)

• Figure 1: Illustration of diverse UAV-based 3D network configurations for B5G and 6G wireless communication. Low-altitude DBSs facilitate dense network deployments in urban settings to enhance spectrum reuse and LOS connectivity. High-altitude platforms, such as balloons, extend coverage across vast areas. Backhaul solutions include inter-UAV links and connections to terrestrial base stations or satellites, utilizing FSO, cmWave, and mmWave technologies for robust wireless communication.
• Figure 2: Visualization of the backhaul optimization, with pre-determined DBS locations and GNs loads. Potential backhaul links are shown with throughput values. Each DBS's total GNs load is labeled. Two sample connectivity paths are highlighted: green lines signify sufficient backhaul capacity, and red indicates congestion with respective accumulated loads.
• Figure 3: Average minimum required number of DBSs for different backhaul distances $d_\text{max}$ given the specific $N_\text{B}$ and $R_\text{A}$ constraints.
• Figure 4: Left: Histogram of DBS backhaul neighbors across $N_\text{B}$ settings. Right: Histogram of GN-DBS distances for various $R_\text{A}$ settings. Results from the proposed HC method are compared with K-means, keeping the cluster count ($M$) constant.
• Figure 5: Probability of success vs. DBSs count $M$ for different GA settings and the benchline scheme. The NVP strategy outperforms the rest for large values of $M$.

Theorems & Definitions (4)

• Theorem 1
• proof
• Theorem 2
• proof