2025/07/11 by Shamsabadi, Afsoon Alidadi, Yadav, Animesh, Yanikomeroglu, Halim
#FOS: Electrical engineering #Systems and Control (eess.SY) #electronic engineering #information engineering
paper · doi:10.48550/arxiv.2507.08299
Next-generation wireless networks (xG) must provide ubiquitous connectivity while enhancing user experience in both densely populated urban areas and rural regions. To achieve this, a disruptive network architecture is essential, and high altitude platform stations (HAPS) offer a promising solution. By integrating HAPS with terrestrial networks, we can create HAPS-empowered vertical heterogeneous networks (vHetNets), which significantly improve coverage and capacity, as well as support emerging use cases. In HAPS-empowered vHetNets, different tiers can share the same spectrum, forming harmonized spectrum vHetNets that enhance spectral efficiency (SE). However, we face two major challenges: i) co-channel interference in harmonized spectrum vHetNets, and ii) the large-scale nature of the network. To address the first challenge, we adopt a cell-free approach as the underlying network architecture for the HAPS-empowered vHetNet. In this approach, base stations use beamforming to direct high-gain, narrow beams toward users, which helps mitigate interference. However, this creates a nonconvex and high-dimensional optimization problem, which highlights the second challenge of dealing with a large-scale network. Consequently, centralized solutions become impractical due to the computational and communication overhead involved. The standard two-block alternating direction method of multipliers (ADMM) is one option, but nonconvex constraints can hinder its convergence. As an alternative, we have developed a two-level distributed proportional fairness beamforming weight design (PFBWD) algorithm. This algorithm uses a combination of the augmented Lagrangian method (ALM) and a three-block ADMM framework. The proposed method effectively tackles nonconvexity, reduces complexity, and enables scalable, distributed optimization with guaranteed convergence.