Active STAR-RIS-Aided RSMA IoT Systems: Performance Analysis, Model-Based and Data-Driven Resource Allocation Frameworks

Introduction to the show: ident: Electrical Engineering. The day in electrical engineering: signals, systems and the hardware that carries them.

Rosa: I'm Rosa.

Dev: And I'm Dev, here with Taro.

Rosa: We're looking at the paper: "Active STAR-RIS-Aided RSMA IoT Systems".

Dev: Here's the gist of the paper. Simultaneously transmitting and reflecting reconfigurable intelligent surfaces (STAR-RISs) have emerged as a promising technology for achieving full-space signal coverage with low hardware complexity and energy consumption The resulting double-fading cascaded channels and the inherent imbalance between near and far users pose reliability and fairness challenges This study presents comprehensive information-theoretic and optimization frameworks for active STAR-RIS-assisted RSMA Internet-of-Things systems System Setup and Challenges The research focuses on a downlink RSMA IoT system supported by a STAR-RIS, where the near user Un lies within the reflection region of the STAR-RIS, and the far user Uf is positioned in its transmission region The system operates according to an ES protocol, which allows each element to flexibly divide the incoming energy into refracted and reflected components Both direct and cascaded links between S and Ul are considered, introducing a binary parameter wl ∈

0, 1: where wl = 1 indicates the existence of a direct link The near/far configuration is the canonical RSMA setting that isolates the near–far fairness problem and admits tractable closed-form OP/EC analysis Information-Theoretic Analysis The paper establishes a complete and unified analytical framework for active STAR-RIS-assisted RSMA IoT systems by deriving closed-form expressions for the OP and EC under both with and without direct link scenarios The outage probability OPl is mathematically expressed as 1 − Pr Rc,l ≥ Rthc, Rl ≥ Rthl = 1 − Pr then Fκl(max (℘c,l, ℘l)), where Rthc and Rthl are the target rates for sc and sl respectively The random variable κl follows an approximate generalized Gamma distribution with first shape parameter ξl, second shape parameter 1/2, and scale parameter omega2l The asymptotic OP corresponding to (7) is calculated as OPgl = 1/Γ(ξl + 1) P∞ k=0 (−1)k k!(k+ξl) / Ωl p max (℘c,l, ℘l) k+ξl The diversity order is obtained as Dl = − lim γ¯→∞ log OPgl / log (¯γ) = ξl/2 Resource Allocation Frameworks The work develops two fairness-oriented RA frameworks: a model-based one and a data-driven one The model-based RA framework formulates joint resource allocation (RA) to optimize rate-, reliability-, energy-, or sensing-related objectives, addressing the non-convex problem via successive convex approximation (SCA) and block coordinate descent (BCD) techniques To enhance user fairness, the outage fairness optimization problem is formulated as min α,τ maxl OPl subject to system constraints The capacity fairness optimization aims to reduce the EC gap between users by solving max α,τ minl ECl subject to QoS constraints Data-Driven Approximation To overcome the real-time and scalability limitations of the model-based RA approaches, a data-driven RA framework is proposed based on machine learning (ML) models This framework approximates the optimal RA policy learned from model-based optimization through supervised regression, training on 500,000 samples The DMNN extends traditional deep neural networks by jointly learning multiple output variables within a single model The CMNN is designed to exploit local feature correlations among the system parameters while maintaining low inference complexity The MXGB model is adopted as a tree-based ensemble learning approach for predicting the RA parameters, achieving the lowest inference complexity Numerical Validation and Conclusion Extensive Monte Carlo simulations validate the analytical results and demonstrate the effectiveness of the proposed frameworks under various system settings The numerical results show that both OP and EC performance improve significantly as STAR-RIS elements or amplification gain increase, confirming the accuracy of the information-theoretic analysis framework The MXGB-based RA framework is identified as the most efficient scheme for real-time fairness-aware RA strategies due to its low inference time Future research directions include a rigorous treatment of residual SI accumulation and nonlinear amplification under high gains, and an iCSI-robust analytical and RA framework The over-the-air, software-defined-radio, or hardware emulation validation is essential to confirm end-toend real-time operation

Paper summary: Rosa: To recap, this paper explores active STAR-RIS systems for RSMA IoT, focusing on how they handle the challenges of double-fading channels and unfairness between near and far users.

Dev: They set up the system where a near user Un is in the reflection region and a far user Uf is in the transmission region, which creates that classic near-far fairness issue.

Taro: The analysis part gives us closed-form expressions for OP and EC when there's no direct link, and when there is a direct link too, which is important for real systems because you don't always have perfect channels.

Rosa: They also calculate the asymptotic OP corresponding to equation number seven as one over Gamma of xi sub l plus one times P infinity k equals zero minus k to the power of minus one times k plus xi sub l over omega sub l p max (pi c sub l, pi L) k plus xi sub l.

Dev: And they find the diversity order, DO, is equal to xi sub l over two. That tells us how well we can improve performance as we get closer to infinity in terms of the average channel gain.

Taro: On the allocation side, they have a model-based framework that uses successive convex approximation and block coordinate descent to optimize objectives like rate or energy, but they specifically formulate an outage fairness optimization problem by minimizing alpha tau max l OP sub l subject to system constraints.

Rosa: Then they also have a capacity fairness optimization aiming to reduce the EC gap between users by solving for max alpha tau min l EC sub l subject to QoS constraints.

Dev: To make this practical for real-time systems, they introduce a data-driven framework based on machine learning models trained on five hundred thousand samples to approximate the optimal resource allocation policies.

Taro: They use a Deep Neural Network called DMNN that jointly learns multiple output variables, and another one called CMNN that exploits local feature correlations among system parameters while keeping inference complexity low.

Rosa: Plus they adopt an MXGB model, which is a tree-based ensemble learning approach for predicting the allocation parameters because it achieves the lowest inference complexity.

Dev: The paper states that this unified integration of analysis, optimization, and learning under a fairness objective is what's new when you compare it to other works in Table I.

Taro: So, for someone listening just to the show, this means they're showing how to move from theoretical analysis right into practical AI-driven decision-making for these complex communication systems.

Conclusion: Rosa: Looking at the title again, "Active STAR-RIS-Aided RSMA IoT Systems: Performance Analysis, Model-Based and Data-Driven Resource Allocation Frameworks," it tells us this isn't just one thing; it covers the whole pipeline from theory to practical allocation.

Dev: The authors are Ngo Hoang Tu, Member IEEE, Vo Nguyen Quoc Bao, Senior Member IEEE, and Tran Thien Thanh, Member IEEE. They're researchers focused on these kinds of integrated communication systems.

Taro: What this implies for the real world is that we now have a comprehensive way to understand how to deploy these surfaces to cover a wide area while making sure the users who are closest don't get completely ignored by the system design.

Rosa: The implication is that for things like smart cities or massive IoT deployments, this framework gives engineers a solid starting point for designing systems that are both efficient and fair.

Dev: It moves beyond just finding the best signal in isolation; it forces you to consider how the near and far users interact through those active surfaces in a way that respects fairness constraints.

Taro: The paper's own limitation is that they focus on active mode, and they mention that future work needs a more rigorous treatment of residual signal accumulation and nonlinear amplification under high gains.

Rosa: That’s the caveat: when you crank up the performance, like with higher gains in the STAR-RIS elements, things get nonlinear and we need better math to handle that noise.

Dev: And they also suggest needing an interference channel state information robust framework to handle situations where your knowledge of the channel isn't perfect during real-time operation.

Taro: So, it’s a solid analysis for now, but pushing the hardware gains will require those deeper treatments of nonlinearities and imperfect information.