Fairness-Aware Power Allocation Keeps Every User Connected in Hybrid Satellite Networks

…

…

The study arrives at a moment when the architecture of non-terrestrial networks is shifting from single-layer designs toward genuinely hybrid systems. In the space–air–ground configuration examined by the researchers, low Earth orbit satellites supply the wide-area backbone, high-altitude platform stations operating in the lower stratosphere act as intermediate relays that shorten the effective communication distance, and terrestrial users connect through whichever layer offers the most favorable link conditions. This layering matters because signal quality degrades dramatically with distance: a link routed through a platform at roughly 17 to 25 kilometers experiences far less path loss than one traversing thousands of kilometers from orbit. The practical consequence is that the same user can present very different channel qualities depending on which node serves it, and the network must allocate scarce power and spectrum across all of these heterogeneous links simultaneously.

The fairness problem the paper addresses is rooted in a well-known tension in multi-user wireless systems. When a network operator maximizes the sum of all users’ data rates, the mathematically optimal solution tends to concentrate resources on the users with the strongest channels, because each unit of power spent on a good channel yields more throughput than the same unit spent on a poor one. This is efficient in aggregate but can leave users on weak or impaired links with so little signal-to-noise ratio that they effectively disconnect. In terrestrial cellular networks this trade-off is annoying; in non-terrestrial networks serving remote regions, disaster zones, oceans, and deserts, a disconnected user may have no alternative connectivity path at all. The authors therefore frame fairness not as an abstract proportional allocation but as a concrete, verifiable requirement: every user must remain above a minimum signal-to-noise ratio threshold, which guarantees a baseline connection quality regardless of channel conditions.

The choice of multiple access techniques is central to how this requirement can be met without sacrificing too much capacity. Orthogonal schemes such as FDMA, TDMA, CDMA, and OFDMA divide resources cleanly among users, which simplifies receivers but caps the number of simultaneous users at the number of available orthogonal slots. Power-domain NOMA relaxes this by letting users share the same time-frequency resources and separating them at the receiver through successive interference cancellation, but that cancellation becomes fragile in large-scale systems where residual interference accumulates and imperfect cancellation compounds. Sparse Code Multiple Access takes a code-domain approach, allowing the number of users to exceed the available orthogonal resources through deliberately sparse codeword structures that keep receiver complexity manageable. Rate-Splitting Multiple Access goes further by treating interference in a more nuanced way: part of each message is decoded at all receivers as common information, while the remainder is decoded privately, so interference is neither fully suppressed nor fully treated as noise but partially decoded. This flexibility is what gives RSMA its robustness when channel state information is imperfect, which is the normal condition in satellite links where propagation delays and mobility make accurate channel estimates difficult to maintain.

The simulation comparison between these two techniques produced a consistent advantage for RSMA across both link types. On radio frequency links the gain reached up to 19 percent in energy efficiency and sum data rate, while on free-space optical links the transmission capacity advantage was approximately 15 percent. The authors attribute this to RSMA’s more efficient interference management, and the mechanism is worth unpacking. In SCMA, users are separated by codebook design, and when channel conditions diverge sharply among users sharing a resource, the code-domain separation provides limited ability to adapt the split between useful signal and interference. RSMA, by contrast, can tune the power split between common and private portions of each transmission to match the actual channel state, effectively converting part of the interference into decodable signal. In a hybrid network where some users see strong satellite links and others depend on weaker relayed paths, this adaptability translates directly into the measured gains.

The channel models used in the evaluation deserve attention because they shape how meaningful the results are for real deployments. For the radio frequency access links, the researchers adopted the Nakagami-m fading model, which represents fading severity through a single parameter m and remains a workhorse of satellite communication analysis precisely because it is tractable while spanning weak, moderate, and strong fading regimes. The paper situates this choice within a broader literature of more elaborate composite distributions: the λ–κ–μ model captures multipath clustering, line-of-sight components, and correlation between in-phase and quadrature signal components; its shadowed extension adds shadowing effects; the double λ–κ–μ variant handles cascaded relaying scenarios; and the α–η–μ/IGA distribution addresses nonlinear multipath propagation with inverse-Gamma shadowing. These generalized models offer richer physical fidelity but at the cost of analytical and computational complexity, and the authors’ decision to retain Nakagami-m reflects a deliberate trade-off between realism and tractability for a system-level optimization study.

The free-space optical links face an entirely different impairment landscape. Atmospheric turbulence induces random fluctuations in the refractive index along the propagation path, producing scintillation that the Málaga turbulence model captures with a mathematically flexible mixture-of-distributions formulation. On top of turbulence, the model incorporates atmospheric attenuation, which scales with distance and weather, and pointing errors, which arise when the transmitter’s beam and the receiver’s aperture fail to align perfectly — a persistent concern for platforms that move relative to one another. Optical links are attractive in non-terrestrial networks for their enormous unlicensed bandwidth and directional energy efficiency, but their performance is far more sensitive to these environmental and geometric effects than radio frequency links. Evaluating the same resource allocation framework across both channel types therefore tests whether the proposed approach generalizes across fundamentally different propagation regimes, and the results suggest it does.

The Fairness-Aware Geometric Water-Filling algorithm is the paper’s principal methodological contribution, and its logic follows from the classical water-filling solution to power allocation. Conventional geometric water-filling distributes power across users or channels by filling from the strongest upward, which maximizes sum capacity but naturally starves the weakest links — exactly the behavior the fairness requirement forbids. The proposed extension introduces an SNR-based minimum power constraint: before the water-filling optimization proceeds, each user is guaranteed the power needed to reach the specified signal-to-noise ratio floor. Users whose channels are strong enough to exceed the threshold with modest power are served by the ordinary water-filling logic, while users on severely faded links receive the guaranteed minimum first, and the remaining power is then distributed to maximize total capacity. The result is a two-tier allocation that preserves the near-optimality of classical water-filling for the system as a whole while converting the worst-case user experience from potential outage to guaranteed connectivity.

The simulation findings on this front are notable for how small the capacity penalty of fairness turns out to be. The FAGWF algorithm achieves performance close to unconstrained geometric water-filling while ensuring that no user falls below the SNR threshold, and it mitigates user outage under severe fading with only marginal loss in aggregate throughput. This is an encouraging empirical result because it suggests the capacity–connectivity trade-off, while real, is not steep in the regimes studied: the power that must be diverted to keep weak users connected comes disproportionately from the surplus capacity that strong-channel users would otherwise absorb, rather than from the capacity needed to keep the network’s overall throughput competitive. In other words, the marginal utility of power spent beyond a user’s connectivity needs diminishes, so redirecting some of that surplus costs little in sum-rate terms while preventing disconnection entirely.

Several implications follow for the design of future non-terrestrial networks. First, the demonstrated superiority of RSMA over SCMA in both RF and FSO settings strengthens the case for rate-splitting as a foundational multiple access layer in next-generation standards, particularly where channel state information is imperfect and user density is high. Second, the finding that fairness constraints can be imposed with modest capacity cost challenges the assumption that connectivity guarantees and throughput maximization are inherently opposed goals; the right power allocation structure can deliver both. Third, the evaluation across heterogeneous channel models — fading on the RF side, turbulence with pointing errors on the optical side — provides a template for how hybrid space–air–ground systems should be benchmarked, since single-channel analyses can mask failures that only appear when the network must serve users across very different link qualities simultaneously.

The work also leaves open questions that define the research frontier. The fairness criterion studied is a hard SNR floor, which guarantees connectivity but does not by itself equalize throughput across users; richer fairness objectives, such as proportional fairness or max-min rate allocation, could be integrated with the same geometric water-filling machinery. The multi-user scalability of the optimization, the interaction between beam-level resource management in multibeam satellite systems and the per-user power constraints studied here, and the behavior of the framework under mobility and time-varying channel estimates all remain areas for further investigation. As Internet of Things deployments continue to expand into regions where terrestrial infrastructure is absent, the ability to serve massive numbers of devices fairly — without letting the weakest links silently drop out — will become an increasingly important measure of whether non-terrestrial networks fulfill their promise of truly ubiquitous coverage.

Subject of Research: …

Article Title: Fairness-aware resource allocation for FSO/RF non-terrestrial networks

Article References: Ghanem, M., Nordin, R., Mohd Ramly, A., Behjati, M., & Bani-bakr, A. (2026). Fairness-aware resource allocation for FSO/RF non-terrestrial networks. Frontiers in Communications and Networks, 7, Article 1880413. https://doi.org/10.3389/frcmn.2026.1880413

Image Credits: AI Generated

DOI: 10.3389/frcmn.2026.1880413

Keywords: Fairness-aware, resource, allocation, non-terrestrial, networks, scientific research