The core idea rests on two technologies that have each attracted enormous attention in the race toward sixth-generation wireless. Non-orthogonal multiple access, or NOMA, boosts spectral efficiency by serving multiple users on the same time and frequency resource, separating their signals in the power domain: users with weaker channels receive more transmit power, and stronger users peel off the weaker users’ messages first through successive interference cancellation before decoding their own. Fluid antenna systems, or FAS, take a different tack entirely. First proposed by Koon-Wong Wong and collaborators in a series of influential papers between 2020 and 2021, FAS lets a single radiating element switch among many pre-defined ports distributed over a compact space. Because wireless fading varies rapidly over distance, selecting the port with the strongest instantaneous channel can rescue a link that would otherwise be trapped in a deep fade, all without the cost and complexity of multiple radio frequency chains.

Previous efforts to marry the two technologies, however, mostly considered only two users and frequently assumed ideal conditions, such as perfect port selection and flawless interference cancellation. Those assumptions fall apart in realistic short-packet scenarios, where IoT and URLLC traffic arrives in bursts so brief that the channel may remain essentially constant for the entire transmission. In this finite-blocklength regime, the Shannon capacity formula no longer applies cleanly; codes cannot average over fading the way long packets can, and the achievable rate carries a penalty tied to the channel’s dispersion. The Nantong team set out to close this gap with a generalized three-user downlink framework that scales to practical multi-tier deployments, such as smart factories and emergency rescue networks where different user groups have sharply different reliability requirements.

The system model is deliberately concrete. A base station with a conventional fixed antenna transmits a superimposed NOMA signal to three users arranged by channel quality: a central user with the strongest link, an intermediate cell-edge user and a distant cell-edge user with the weakest link. Each user carries a one-dimensional fluid antenna whose ports are evenly spaced across a linear space measured in wavelengths, with switching delays assumed negligible, a reasonable approximation for pixel-based switchable antenna hardware. Crucially, the fading coefficients across ports are not independent. The researchers adopted Jakes’ correlation model, in which the correlation between any two ports follows the zero-order Bessel function of the first kind, and they exploited the symmetry of the resulting correlation matrix through eigenvalue decomposition to characterize the channel at each port analytically.

Decoding follows the classic NOMA hierarchy with a twist that favors the weak. The central user first decodes and cancels the messages of both cell-edge users, then retrieves its own. The intermediate user cancels only the distant user’s signal before decoding its own message, treating the stronger central user’s transmission as interference. The weakest user, which suffers the worst path loss, performs no cancellation at all and decodes its own signal directly while treating everything else as noise, but it is compensated by receiving the largest share of transmit power. This ordering ensures that users in the poorest channel conditions face the least decoding burden, a deliberate engineering choice that the analysis shows pays off in block error rate performance.

On the analytical side, the team derived closed-form approximations for the average block error rate in short-packet transmissions, capturing the interplay among fluid antenna port selection, power allocation and imperfect successive interference cancellation. Using the finite-blocklength normal approximation pioneered by Polyanskiy, Poor and Verdu, they expressed the error probabilities through the cumulative distribution functions of the relevant signal-to-interference-plus-noise ratios, which in turn incorporate the spatial correlation of the fluid antenna ports. Perhaps the most striking theoretical result emerges in the high signal-to-noise regime: every user in the system achieves a diversity order equal to the number of fluid antenna ports, even in the presence of multiuser interference. In other words, adding ports steepens the error-rate curve in a predictable, quantifiable way, confirming that the spatial diversity harvested by FAS survives the rigors of short-packet NOMA decoding.

Reliability is only half the story, because latency-critical applications also demand throughput. The researchers formulated a sum-rate maximization problem under power normalization and SIC ordering constraints, and solved it with a projected gradient descent algorithm. A subtle but important simplification underpins the optimization: although finite-blocklength rate expressions are used for performance evaluation, the power allocation problem is posed with the infinite-blocklength sum rate, because the channel dispersion term does not materially shift the optimal allocation, a fact the authors verified numerically and note is common in the literature. The per-iteration cost of the algorithm involves only elementary operations on precomputed channel statistics, and the method’s Lipschitz-continuous gradient guarantees convergence to a stationary point with a backtracking line search ensuring the optimized sum rate never drops below the fixed baseline. That efficiency matters, because power control in real-time URLLC systems cannot afford heavy computation between packets.

The Monte Carlo simulations bring the mathematics to life. With three fluid antenna ports, a blocklength of 200 symbols, packet sizes on the order of tens to hundreds of bits and a fluid antenna size of one wavelength, increasing the number of ports from three to four to five consistently improved block error rates for all three users, while a conventional single-antenna scheme lagged far behind, especially for the cell-edge devices. Three-dimensional surface plots and contour maps of the average block error rate against the power allocation coefficients revealed how tuning the power split between the two edge users enables a flexible reliability trade-off: the closer edge user needs its coefficient above roughly 0.3 for low error rates, while the more distant user performs best near 0.2, letting network operators tailor settings to scenario-specific requirements. On the throughput side, the finite-blocklength penalty was clearly visible, with the sum rate at a reference signal-to-noise ratio falling from approximately 3.96 bits per second per hertz under the Shannon limit to about 2.70 bits per second per hertz under a 10 percent target block error rate, a gap that underscores why short-packet analysis cannot simply lean on classical capacity formulas.

The optimization results were equally encouraging. The projected gradient algorithm, initialized from a fixed power allocation and refined with an Armijo-conditioned line search, consistently outperformed or matched the fixed baseline across the entire signal-to-noise range, with the most substantial gains appearing in the low-to-moderate signal-to-noise region where interference and rate trade-offs are hardest to balance. At higher signal-to-noise ratios the gap narrows as the fixed allocation approaches optimality, but the adaptive scheme never loses ground. For network designers, this means the framework is most valuable precisely where edge devices struggle: in the unfavorable channel conditions that define the cell edge, where every decibel of effective gain translates directly into packets delivered on time.

Taken together, the study sketches a flexible and effective blueprint for short-packet edge access in the networks of the coming decade. By letting each device’s antenna roam across a small patch of space and dynamically select the port where the channel happens to cooperate, and by layering NOMA power-domain multiplexing on top with carefully tuned and optimizable power splits, the proposed framework converts the cell edge from a reliability dead zone into a manageable engineering problem. The authors’ simulations confirm that the combination achieves a diversity order equal to the number of fluid antenna ports while preserving throughput favorable for latency-critical traffic. As URLLC and massive IoT deployments push connectivity into factories, hospitals, rescue operations and smart infrastructure, frameworks like this one, which trade a modest antenna switching capability for quantifiable gains in reliability and throughput, may well become a standard tool in the sixth-generation toolkit, proving that sometimes the smartest way to fight a bad channel is simply to move the antenna to where the signal is stronger.

Subject of Research: A fluid antenna system-assisted NOMA framework for reliable multiuser short-packet communications at the cell edge

Article Title: A multiuser FAS-NOMA framework for short-packet communications

Article References: Ju, J., Yao, H., Sun, Q., & Wu, Y. (2026). A multiuser FAS-NOMA framework for short-packet communications. Frontiers in Communications and Networks, 7, Article 1842339. https://doi.org/10.3389/frcmn.2026.1842339

Image Credits: AI Generated

DOI: 10.3389/frcmn.2026.1842339

Keywords: fluid antenna system, NOMA, short-packet communications, URLLC, Internet of Things, block error rate, successive interference cancellation, diversity order, sum-rate maximization, cell-edge users, 6G wireless, power allocation