Aerial-STAR: UAV-Mounted STAR-RIS Systems

• Aerial-STAR is a UAV-mounted STAR-RIS system offering 360° coverage via simultaneous transmission and reflection for robust air-ground communications.
• It jointly optimizes UAV trajectory, passive beamforming, and resource allocation to balance throughput, energy efficiency, and mobility constraints.
• Performance studies show gains up to 28–97% over conventional RIS, highlighting its advantages in multi-user, MEC, and covert networking scenarios.

Aerial-STAR denotes a UAV-mounted Simultaneous Transmit And Reflect Reconfigurable Intelligent Surface (STAR-RIS) system for wireless communications. In this line of work, the system integrates the 360° coverage and reconfigurability of STAR-RIS with the mobility and LoS adaptability of unmanned aerial vehicles, so that transmission, reflection, trajectory design, beamforming, and resource allocation are optimized jointly in air-ground networks (Rizvi et al., 30 Aug 2025). Closely related formulations appear under the names STAR-RIS-UAV, UAV-mounted STAR-RIS, airborne STAR-RIS, and aerial STAR-RIS-assisted systems, covering downlink multi-user communications, coordinated multipoint cellular systems, MEC, emergency networking, IoT NOMA, covert satellite-terrestrial links, and comparative 3D deployment studies (Shi et al., 2023).

1. Terminology, scope, and canonical system models

The literature uses several closely related names for the same architectural idea. One paper studies a STAR-RIS-UAV aided coordinated multipoint cellular system for multi-user networks, in which a UAV is equipped with a STAR-RIS and hovers between two base stations to serve a cell-edge user in the transmission sector and cell-center users in the reflective sector (Shi et al., 2023). Another studies STAR-RIS assisted UAV communication systems, where one UAV at constant altitude is aided by a fixed STAR-RIS to serve multiple ground users via NOMA (Zhang et al., 2022). A later work explicitly names the UAV-mounted STAR-RIS system Aerial-STAR and frames it as a downlink multi-user communication system with a coupled TRC phase shift model (Rizvi et al., 30 Aug 2025).

Formulation in the literature	Core setting	Source
Aerial-STAR	UAV-mounted STAR-RIS for downlink multi-user communications	(Rizvi et al., 30 Aug 2025)
STAR-RIS-UAV	UAV equipped with STAR-RIS for CoMP multi-user networks	(Shi et al., 2023)
STAR-RIS assisted UAV communication systems	UAV, fixed STAR-RIS, multiple ground users via NOMA	(Zhang et al., 2022)

Across these formulations, STAR-RIS is introduced as distinct from conventional RIS because it can simultaneously transmit and reflect incoming signals, thereby providing full-space coverage and additional degrees of freedom (Zhang et al., 2022). In the UAV-mounted setting, this is combined with aerial mobility, dynamic placement, and improved line-of-sight conditions. This suggests that “Aerial-STAR” is best understood not as a single protocol, but as a family of airborne STAR-RIS architectures whose common feature is the co-design of intelligent-surface control and UAV motion.

2. Electromagnetic model, transmission/reflection control, and energy-aware deployment

A central modeling distinction is between independent and coupled transmission and reflection coefficients (TRCs). Much of the earlier literature assumes that STAR-RIS transmission and reflection coefficients can be tuned independently, but later work emphasizes that this is unrealistic for passive surfaces because transmission and reflection are mutually coupled and constrained by energy conservation and electromagnetic properties (Rizvi et al., 30 Aug 2025). In the coupled model used for Aerial-STAR, the amplitude and phase constraints are written as

$$\beta_n^\mathcal{R} = \sqrt{1 - (\beta_n^\mathcal{T})^2}, \qquad |\theta_n^\mathcal{R} - \theta_n^\mathcal{T}| = \frac{\pi}{2},$$

with the TRC matrix

$$\boldsymbol{\Phi}_c = \mathrm{diag}(\beta_1^c e^{j\theta_1^c}, \ldots, \beta_N^c e^{j\theta_N^c}), \; c \in \{\mathcal{R}, \mathcal{T}\}$$

(Rizvi et al., 30 Aug 2025). A related 3D comparison study uses the constraints

$$\beta_{t,n}^2 + \beta_{r,n}^2 = 1,\qquad \cos(\phi_{t,n} - \phi_{r,n}) = 0$$

for aerial STAR-RIS elements (Yang et al., 9 Dec 2025).

Control protocols in the broader STAR-RIS-UAV literature typically include energy splitting (ES) and mode switching (MS). Under ES, each element splits its received signal energy into transmitted and reflected fractions, with $\beta_n^t + \beta_n^r = 1$, while under MS each element is set to either transmission or reflection mode (Shi et al., 2023). Earlier UAV-assisted STAR-RIS studies report that the energy split between reflection and transmission modes is adaptive and highly dependent on the real-time UAV trajectory (Zhang et al., 2022).

In airborne deployment, propulsion cannot be treated independently of surface size. Aerial-STAR work introduces an explicit RIS-drag term in the UAV energy model: $$P_{\text{total}}(t) = P_{UAV} + P_{\text{RIS-drag}}, \qquad P_{\text{RIS-drag}} = \frac{1}{2} \rho A_{RIS} C_d V_\perp^3,$$ with

$A_{RIS} = (N_x-1)^2 \frac{\lambda^2}{\mu^2}$

and communication efficiency

$$\eta_{EE}(t) = \frac{\sum_{j=1}^{J}R_j(t)}{P_{\text{total}}(t)}$$

(Rizvi et al., 30 Aug 2025). This makes Aerial-STAR fundamentally different from fixed STAR-RIS deployment: increasing the number of elements improves beamforming flexibility but also increases drag and energy demand.

3. Optimization objectives and mathematical formulations

The dominant design pattern is joint optimization. In passive Aerial-STAR downlink systems, the objective is to optimize UAV trajectory, passive STAR-RIS TRCs, and active beamforming vectors at the base station subject to QoS, power, speed, and coupling constraints (Rizvi et al., 30 Aug 2025). In one canonical formulation, the objective is

$$\max_{\mathbf{T}, \mathbf{\Phi}, \mathbf{W}_B} \sum_{t=1}^{T} \eta_{EE}(t)$$

with minimum user-rate, UAV-speed, phase-range, and coupled-phase constraints (Rizvi et al., 30 Aug 2025).

Other formulations target sum rate rather than communication efficiency. In STAR-RIS assisted UAV communication systems, the goal is to maximize the sum rate of all users by jointly optimizing the STAR-RIS passive beamforming vectors, the UAV trajectory, and power allocation (Zhang et al., 2022). In the coordinated multipoint cellular setting, the objective is to maximize the system sum rate by jointly optimizing BS beamforming vectors and STAR-RIS transmission and reflection coefficient matrices under BS power and minimum QoS constraints (Shi et al., 2023). In active STAR-RIS IoT NOMA networks, the objective is the system sum rate maximization problem for the joint optimization of active STAR-RIS beamforming, UAV trajectory design, and power allocation, where the beamforming variables include amplification, power-splitting, and phase shift (Zhao et al., 5 Jan 2025).

A second major objective class is energy minimization. In aerial STAR-RIS empowered MEC, the formulated problem minimizes the total energy consumption of both IoT devices and the aerial STAR-RIS-UAV while optimizing task offloading, aerial STAR-RIS trajectory, amplitude and phase shift coefficients, and transmit power under latency, MEC resource, STAR-RIS energy split, and UAV mobility constraints (Aung et al., 2023). A third class adds covertness. In aerial active STAR-RIS-assisted satellite-terrestrial covert communications, the objective is to maximize the sum of the fair channel capacity for all ground users while satisfying a covert constraint derived from the minimal detection error probability of a Warden (Zhang et al., 22 Apr 2025).

This diversity of objectives implies that Aerial-STAR is not tied to a single network utility. Depending on the application, the same airborne STAR-RIS substrate can be tuned for sum rate, energy efficiency, offloading efficiency, fairness, or covert performance.

4. Solution methods: alternating optimization, convexification, and deep reinforcement learning

Model-based optimization is prominent in the earlier literature. For STAR-RIS assisted UAV communication systems, the non-convex problem is decomposed into three subproblems—STAR-RIS beamforming optimization, UAV trajectory optimization, and power allocation optimization—which are solved alternately using convex reformulations, iterative optimization, semidefinite programming relaxations, and successive convex approximation (Zhang et al., 2022). In STAR-RIS-UAV coordinated multipoint systems, a joint penalty-based iterative algorithm is proposed instead of alternating optimization, so that all variables are optimized in every iteration for both ES and MS protocols (Shi et al., 2023). For active STAR-RIS IoT NOMA, an alternating optimization (AO) algorithm decouples the original problem into three subproblems, with a penalty-based method for the rank-one constraint and successive convex optimization for UAV trajectory and power allocation (Zhao et al., 5 Jan 2025). In the 3D performance comparison between aerial RIS and STAR-RIS, the sum-rate problem is addressed via weighted minimum mean square error and block coordinate descent, with Penalty Dual Decomposition for amplitude-phase coupling (Yang et al., 9 Dec 2025).

Dynamic environments and high-dimensional action spaces have motivated reinforcement learning. In aerial STAR-RIS empowered MEC, the joint problem is modeled as an MDP and solved using proximal policy optimization (PPO) because of its sample efficiency and stability (Aung et al., 2023). In STAR-RIS assisted UAV NOMA emergency communications, the long-term constrained optimization is cast as a CMDP and solved by Lagrange based reward constrained proximal policy optimization (LRCPPO), with an inner penalized-reward PPO layer and an outer Lagrange-multiplier update layer (Lei et al., 2023). In aerial active STAR-RIS-assisted covert communications, a generative deterministic policy gradient (GDPG) algorithm uses a generative diffusion model (GDM) as the policy representation and an action gradient mechanism for policy improvement (Zhang et al., 22 Apr 2025). For coupled-phase Aerial-STAR, the hybrid action space motivates a Dual Actor Deep Deterministic Policy Gradient (DA-DDPG) algorithm with one actor for continuous actions and one actor for discrete actions (Rizvi et al., 30 Aug 2025).

These solution choices correspond closely to modeling assumptions. Deterministic convexification is favored when the channel model and constraints are explicit and tractable, whereas DRL is favored when the state-action space is high-dimensional, the objective is long-horizon, or the coupled trajectory–beamforming control is difficult to solve directly within polynomial time.

5. Performance findings and design trade-offs

Aerial-STAR studies repeatedly report gains over conventional RIS baselines, but the gains are qualified by geometry, altitude, hardware constraints, and the chosen objective. In STAR-RIS assisted UAV communication systems, simulations show that the STAR-RIS achieves a higher sum rate than traditional RIS, that the UAV's trajectory is closer to STAR-RIS than that of RIS, and that the energy splitting for reflection and transmission highly depends on the real-time trajectory of UAV (Zhang et al., 2022). In coordinated multipoint multi-user networks, the STAR-RIS-UAV aided wireless communication system has a much higher sum rate than the system with conventional RIS or without RIS, and the proposed structure is described as more flexible than a fixed STAR-RIS (Shi et al., 2023).

For the practically constrained coupled-phase Aerial-STAR model, one study reports that DA-DDPG outperforms conventional DDPG and DQN-based solutions by 24% and 97%, respectively, in accumulated reward; three-dimensional UAV trajectory optimization achieves 28% higher communication efficiency compared to two-dimensional and altitude optimization; and the HFI based reward function provides 41% lower QoS denial rates compared to other benchmarks (Rizvi et al., 30 Aug 2025). The same work states that the mobile Aerial-STAR system shows superior performance over fixed deployed counterparts, and that the coupled phase STAR-RIS outperforms dual Transmit/Reflect RIS and conventional RIS setups (Rizvi et al., 30 Aug 2025).

RIS size is not monotone with airborne efficiency. In the Aerial-STAR energy-aware model, a larger RIS provides higher throughput but also increases drag, so efficiency can peak and then decrease as $N$ increases (Rizvi et al., 30 Aug 2025). In active STAR-RIS IoT NOMA, numerical results indicate that the UAV-mounted active STAR-RIS enhances the channel gain from the BS to the IoT devices by high-quality channel construction and power compensation, and that the performance gap widens as the number of STAR-RIS elements increases (Zhao et al., 5 Jan 2025). In a broader low-altitude wireless network setting, the average transmission rate of the overall system scales positively with both UAV count and STAR-RIS element numbers (Liang et al., 25 Oct 2025).

A more nuanced result emerges from direct aerial RIS versus aerial STAR-RIS comparison in 3D wireless environments. That study concludes that STAR-RIS outperforms RIS in low-altitude scenarios due to its full-space coverage capability, whereas RIS delivers better performance near the base station at higher altitudes (Yang et al., 9 Dec 2025). This qualifies any blanket claim of STAR-RIS dominance and indicates that deployment altitude and orientation are system-level variables, not merely implementation details.

6. Variants, misconceptions, and broader research directions

One common misconception is that all Aerial-STAR work assumes physically realizable STAR-RIS hardware. The literature itself distinguishes sharply between independent TRC models and coupled phase-shift models. The latter are presented as more accurate for passive surfaces, while the former offer greater flexibility and often higher upper-bound performance (Rizvi et al., 30 Aug 2025). A related practical distinction is between passive and active STAR-RIS. Active STAR-RIS-equipped UAVs introduce amplification coefficients and power compensation, and they are studied for IoT NOMA and covert satellite-terrestrial communications, but they also require additional per-element and total power constraints (Zhao et al., 5 Jan 2025).

A second misconception is that the aerial platform is only a placement mechanism. Multiple papers make the UAV state part of the optimization variables: trajectory, speed, altitude, orientation, propulsion energy, and even RIS-induced drag directly affect the achievable utility (Aung et al., 2023). In some formulations, the UAV must satisfy initial and final location constraints; in others, geofencing, velocity transition mechanisms, or collision-avoidance-style formation constraints appear. This suggests that Aerial-STAR is as much an aerial robotics control problem as it is an intelligent-surface communication problem.

The broader research frontier extends beyond single-UAV downlink sum-rate maximization. Reported directions include MEC energy minimization with task offloading (Aung et al., 2023), emergency communication networks with minimum average rate and maximum energy constraints (Lei et al., 2023), collaborative beamforming in low-altitude wireless networks using UAV swarms and STAR-RIS omnidirectional reconfigurable beamforming (Liang et al., 25 Oct 2025), and satellite-terrestrial covert communications assisted by AASTAR-RIS (Zhang et al., 22 Apr 2025). The STARS-enabled ISAC literature, although not specific to UAV mounting, also contributes a relevant systems perspective by framing STARS as a full-space platform for integrated sensing and communications, with both independent and coupled phase-shift models (Wang et al., 2022).

Taken together, these results indicate that Aerial-STAR has evolved from a throughput-oriented extension of RIS-assisted UAV communications into a broader design space centered on full-space coverage, air-ground mobility, hybrid continuous-discrete control, and physically constrained intelligent surfaces. A plausible implication is that future Aerial-STAR systems will be evaluated less by isolated sum-rate gains and more by how well they balance communication efficiency, fairness, mobility cost, hardware realizability, and deployment geometry.