Up in the Air, Split on the Ground: STAR-RIS vs. RIS in 3D Networks

The upgrade is not the answer; the angle is

Hardware upgrades are comforting because they sound linear. A faster chip is faster. A bigger antenna array has more elements. A surface that can both transmit and reflect sounds strictly better than one that only reflects.

Wireless geometry is less polite.

The paper behind this article compares aerial RIS and aerial STAR-RIS in three-dimensional wireless environments.¹ The obvious story would be simple: conventional RIS reflects signals on one side; STAR-RIS can simultaneously transmit and reflect; therefore STAR-RIS should win. The paper’s actual story is more useful. STAR-RIS often wins, especially at lower altitudes and farther deployment positions. But conventional RIS can outperform it near the base station at higher altitude. STAR-RIS is also highly sensitive to orientation.

That means the practical lesson is not “buy STAR-RIS.” It is “first ask where the surface is, how high it flies, and which way it faces.” Revolutionary, I know: the laws of geometry still refuse to read vendor brochures.

What RIS and STAR-RIS are trying to fix

A reconfigurable intelligent surface, or RIS, is a passive surface whose elements can adjust phase shifts so that wireless propagation becomes less accidental. Instead of treating buildings, walls, and blockages as fixed annoyances, a RIS tries to reshape the radio environment.

A conventional RIS is reflection-only. It is useful when the transmitter, the surface, and the served users sit on the right side of the geometry. But that one-sidedness is also the architectural constraint.

A STAR-RIS, short for simultaneously transmitting and reflecting RIS, relaxes that constraint. Its elements can split energy between a transmitted component and a reflected component, allowing users on both sides of the surface to be served. In the paper’s model, STAR-RIS operates in energy-splitting mode, meaning each element can concurrently transmit and reflect the incident signal.

Mount either technology on a UAV and the problem becomes more interesting. Aerial deployment adds altitude, horizontal position, and orientation to the planning problem. It also creates better line-of-sight opportunities. But the gain is not automatic because an aerial surface does not merely exist in space. It points somewhere.

That pointing direction is the hinge of the paper.

The mechanism: channel gain is distance plus direction

The authors do not compare RIS and STAR-RIS using a flat, orientation-free channel abstraction. They model the BS-to-surface and surface-to-user links as Rician fading channels, combining line-of-sight and non-line-of-sight components. More importantly, their large-scale path loss includes directional radiation patterns.

In plain language, the channel gain is affected by both distance and angular alignment. A simplified way to read the model is this:

The directional terms include cosine-based radiation-pattern factors. The more misaligned the incoming or outgoing signal is relative to the surface’s effective direction, the more the channel suffers. This is why “higher” is not always better, and “more flexible hardware” is not always superior.

The distinction between the two architectures enters through orientation:

For the horizontal RIS, the relevant angular relationship depends heavily on elevation. At low altitude, users and the base station can create large incident and reflection angles, reducing effective channel gain. As altitude increases near the base station, alignment improves.

For STAR-RIS, the vertical surface gives full-space coverage. That is valuable when users occupy both transmission and reflection half-spaces. But the STAR-RIS also has an orientation angle $\eta$. Rotate it badly, and the same full-space capability becomes less impressive. A door that opens both ways is still less useful if installed sideways.

The optimization is there to make the comparison fair

The paper does not merely place two surfaces in space and compare raw links. It formulates sum-rate maximization problems for both architectures.

For STAR-RIS, the objective is to maximize the total transmission rate across users, subject to three important constraints:

The resulting problem is non-convex. The authors transform the sum-rate maximization into a weighted minimum mean square error problem, then solve it using block coordinate descent. For the phase-coupling difficulty in STAR-RIS, they introduce penalty dual decomposition and auxiliary variables.

This part of the paper is mostly implementation infrastructure. It matters because the comparison would be weak if one architecture were optimized carefully and the other treated casually. The paper’s algorithmic pipeline is not the business headline, but it is what makes the headline credible.

The experimental role of each figure is worth separating:

This distinction matters. Figure 2 is not a second business thesis; it checks whether the optimization behaves sensibly and reveals orientation sensitivity. Figures 3 and 4 carry the deployment argument.

The first finding: STAR-RIS wins low because full-space coverage matters

At lower altitude, the paper finds that aerial STAR-RIS significantly outperforms aerial RIS. The reason is not mysterious once the geometry is visible.

A low-altitude horizontal RIS faces unfavorable incident and reflection angles. The surface may be close to users, but closeness alone does not rescue poor angular alignment. In the paper’s channel model, angular penalties directly reduce effective channel gain.

STAR-RIS has a different advantage. Because it can transmit and reflect, it can serve both sides of the surface. That full-space coverage is especially useful when the aerial surface is operating closer to the user plane and must handle a wider spread of user directions.

This is the correction to the naive misconception. STAR-RIS is not better because it has a more impressive acronym. It is better in this regime because the geometry makes one-sided reflection costly.

For business interpretation, this points toward low-altitude aerial coverage use cases: temporary events, emergency connectivity, industrial yards, construction zones, ports, and dense urban pockets where direct BS-user links are blocked or weak. In such settings, the surface is not simply extending a clean beam. It is managing messy spatial coverage.

That is where STAR-RIS looks like a real upgrade.

The second finding: RIS recovers when altitude improves alignment

As altitude increases, the performance picture changes. The paper reports that conventional RIS gradually exhibits better performance in regions closer to the base station. The mechanism is angular alignment.

A horizontal RIS near the BS at higher altitude can align more favorably with the base-station side of the channel. In that regime, reflection-only behavior is less damaging, and the conventional RIS can become more efficient than the more flexible STAR-RIS.

This is the uncomfortable result for anyone hoping to turn the paper into a simple procurement slogan. More capability does not erase geometry. STAR-RIS has to split energy between transmission and reflection and obey phase-coupling constraints. If the deployment setting already favors a simpler reflection geometry, conventional RIS can look better.

The business implication is not that RIS is “old but still good,” which is the kind of sentence that should be retired along with most innovation panels. The sharper implication is this:

RIS may remain economically rational in high-altitude, BS-proximate aerial deployments where angular alignment is strong and full-space coverage is less valuable.

That matters because network infrastructure decisions are not made per paper. They are made per site, per budget, per maintenance regime, and per operational envelope.

The third finding: STAR-RIS is powerful but orientation-sensitive

Figure 4 adds the most operationally important complication: STAR-RIS performance can deteriorate sharply when orientation is wrong. The authors test different orientation angles, including $\eta=0$, $\eta=\pi/4$, and $\eta=\pi/2$. In their convergence test, $\eta=\pi/4$ gives the best STAR-RIS performance under the tested deployment. In one altitude-position case, $\eta=\pi/2$ performs poorly.

This is not a minor tuning detail. STAR-RIS is vertically deployed, so its normal vector changes with the horizontal rotation angle. That angle changes how the surface sees both the base station and the users. In the model, this flows directly into the cosine-based directional terms.

A STAR-RIS deployment therefore needs orientation control, not just placement control. In UAV-assisted networks, that means the communications design touches navigation, stabilization, and control systems. A surface with better theoretical coverage may require more precise actuation and monitoring.

The hardware value chain then changes:

So yes, STAR-RIS expands the design space. It also expands the number of ways to be wrong.

What this directly shows, and what Cognaptus infers

The paper directly shows three things under its simulation assumptions. First, aerial STAR-RIS can outperform aerial RIS at low altitude because full-space coverage offsets RIS angular penalties. Second, aerial RIS can outperform STAR-RIS near the base station at higher altitude because alignment improves. Third, STAR-RIS orientation is a first-order performance variable.

Cognaptus infers a broader deployment principle from this: aerial intelligent surfaces should be selected through a geometry-aware planning workflow, not through architecture labels. The choice between RIS and STAR-RIS should come after simulating altitude, horizontal position, base-station placement, likely user distribution, and orientation control capacity.

A practical planning framework would look like this:

This is not a recommendation to standardize on one technology. It is a recommendation to stop pretending the comparison is one-dimensional.

The ROI is in avoiding wrong deployments

The most immediate business value of the paper is not a new algorithm for operators to deploy tomorrow. It is cheaper diagnosis.

If an operator tests STAR-RIS in a weak geometry and sees poor performance, the wrong conclusion would be that STAR-RIS is overhyped. If the operator tests RIS at low altitude in a wide angular spread and sees poor performance, the wrong conclusion would be that RIS is obsolete. Both conclusions are too simple.

The paper suggests that poor performance may come from architecture-geometry mismatch. That is an expensive mistake if discovered after hardware procurement, UAV integration, or field trials.

For a telecom operator, infrastructure vendor, or smart-city contractor, the ROI path is therefore diagnostic:

1. build a site-specific 3D geometry model;
2. simulate RIS and STAR-RIS under realistic altitude and orientation ranges;
3. identify crossover regions where the preferred architecture changes;
4. evaluate whether orientation control costs erase STAR-RIS gains;
5. select hardware after placement regimes are known, not before.

This is not glamorous. It is also the difference between engineering and shopping.

The boundary: this is a clean simulation, not a field manual

The paper’s results are useful, but they should not be inflated beyond the testbed.

The simulations assume a base station at $(0,0,0)$, four single-antenna users randomly distributed in a $100 \times 100$ m² area, blocked direct BS-user channels, Rician fading, fixed transmit power, fixed directivity parameters, and specified noise and path-loss assumptions. Figure 2 uses $N=20$ surface elements for its convergence comparison. The analysis focuses on sum-rate, deployment altitude, horizontal position, and orientation.

Those choices are reasonable for isolating the architectural comparison. They also leave out several operational factors:

The boundary is not a weakness; it is the point of reading the paper correctly. The authors isolate the role of 3D geometry and orientation. They do not claim to solve the full deployment economics of UAV-mounted intelligent surfaces.

The useful conclusion is a map, not a winner

The easiest article to write from this paper would say: STAR-RIS beats RIS at low altitude, RIS can beat STAR-RIS near the base station at higher altitude, and orientation matters. Accurate, but slightly dead on arrival.

The more useful reading is that aerial intelligent surfaces are not plug-and-play upgrades. They are geometry-dependent infrastructure components. The channel does not care whether a surface is newer, more flexible, or better funded. It cares where the surface is, where it points, and how the signal arrives.

STAR-RIS expands coverage possibilities by serving both transmission and reflection regions. RIS remains competitive when the geometry rewards simple reflection. The crossover between them is not a footnote; it is the deployment strategy.

For 6G planning, the message is clean: do not choose between RIS and STAR-RIS as technologies in the abstract. Choose between deployment regimes. The surface in the sky only works if the angles on the ground cooperate.

Cognaptus: Automate the Present, Incubate the Future.