Optimization of RIS-Assisted 6G NTN Architectures for High-Mobility UAV Communication Scenarios

Abstract

The integration of reconfigurable intelligent surfaces (RISs) with non-terrestrial networks (NTNs), particularly those enabled by unmanned aerial vehicles (UAVs) or drone-based platforms, has emerged as a transformative approach to enhance 6G connectivity in high-mobility scenarios. UAV-assisted NTNs offer flexible deployment, dynamic altitude control, and rapid network reconfiguration, making them ideal candidates for RIS-based signal optimization. However, the high mobility of UAVs and their three-dimensional trajectory dynamics introduce unique challenges in maintaining robust, low-latency links and seamless handovers. This paper presents a comprehensive performance analysis of RIS-assisted UAV-based NTNs, focusing on optimizing RIS phase shifts to maximize the signal-to-interference-plus-noise ratio (SINR), throughput, energy efficiency, and reliability under UAV mobility constraints. A joint optimization framework is proposed that accounts for UAV path loss, aerial shadowing, interference, and user mobility patterns, tailored specifically for aerial communication networks. Extensive simulations are conducted across various UAV operation scenarios, including urban air corridors, rural surveillance routes, drone swarms, emergency response, and aerial delivery systems. The results reveal that RIS deployment significantly enhances the SINR and throughput while navigating energy and latency trade-offs in real time. These findings offer vital insights for deploying RIS-enhanced aerial networks in 6G, supporting mission-critical drone applications and next-generation autonomous systems.

1. Introduction

The integration of reconfigurable intelligent surfaces (RISs) into non-terrestrial networks (NTNs) represents a pivotal advancement for 6G wireless systems, particularly in UAV-enabled high mobility environments, characterized by dynamic channel fluctuations, spectral inefficiencies, and heterogeneous user demands [1,2]. UAV-assisted NTNs, encompassing drone swarms, aerial relays, and high-altitude platforms (HAPs), offer flexible deployments but remain challenged by rapid signal attenuation, latency variations due to mobility, and limited onboard processing and energy resources. These constraints are further exacerbated when serving dense ground or aerial user populations in dynamic 3D environments [3]. In this context, RISs can play a transformative role in improving coverage, steering beams dynamically and compensating for channel impairments inherent in UAV-based NTNs. RISs are made up of programmable passive elements, and they can address the aforementioned limitations by dynamically reshaping electromagnetic wave propagation through real-time adjustment of reflection coefficients, thereby enhancing signal integrity, extending coverage, and suppressing interference [4].

Recent innovations integrate RISs with non-orthogonal multiple access (NOMA), which enables spectral efficiency gains by multiplexing users within shared resources. However, interference management and real-time coordination challenges remain unresolved [5]. Later, space-time coding metasurfaces were introduced to enable adaptive beamforming and joint data modulation, resulting in reliable connectivity in mobility scenarios such as airborne or satellite-to-ground communications [6]. Recently, AI-driven frameworks have been used to optimize RIS configurations to align with fluctuating channel states and energy constraints, thus maximizing spectral and energy efficiency in resource-constrained NTN deployments [7]. These developments are augmented by intelligent resource management protocols, such as BLASTER, which holistically coordinate bandwidth allocation, power distribution, and user-RIS association to harmonize quality of service (QoS) with operational sustainability in hybrid terrestrial NTN ecosystems [8].

RIS deployment demonstrably increases system-level metrics, including SINR, throughput, and BER, by mitigating path loss and multipath fading, critical for seamless global connectivity [9]. Despite these developments, there are practical barriers, such as computational complexity in real-time RIS reconfiguration, hardware limitations in extreme mobility scenarios, and scalability in multi-RIS architectures. Expanding RIS applications to emerging NTN use cases, such as urban air mobility, intelligent reflecting satellite arrays, and integrated space–air–ground networks, as shown in Figure 1, will be pivotal in realizing scalable and adaptive 6G infrastructures capable of supporting next-generation mobility and connectivity demands [10].

An RIS enables enhanced signal coverage and quality in obstructed environments, which is a critical limitation in UAV-based communication due to dynamic channel blockages. The passive nature of RISs makes them a low-power, lightweight solution, which is essential for mobile UAV platforms that have stringent energy and payload constraints. The high speeds of UAVs and vehicles lead to rapid changes in channel conditions, characterized by significant Doppler shifts and short coherence times [11,12]. This requires channel estimation techniques capable of tracking these rapid variations, which is particularly challenging given the large number of passive RIS elements [13]. Traditional channel estimation methods often fail in these scenarios, and there is a need for efficient and accurate channel estimation, as imperfect channel state information (CSI) can significantly degrade the performance of RIS-assisted systems [14]. Static optimization techniques, which are calculated offline, are not suitable for high-mobility scenarios, and there is a need for real-time adaptation to changing channel conditions, which requires dynamic optimization algorithms [15]. This paper aims to address this gap by proposing a comprehensive framework for the optimization of RIS-assisted NTNs in high-mobility scenarios. We focus on the following key aspects:

• Joint optimization of RIS phase shifts: We develop an algorithm to optimize RIS phase shifts in real time, considering the dynamic nature of NTNs/UAVs and user mobility.
• Performance metrics: We evaluate system performance using multiple metrics, including SINR, throughput, energy efficiency, outage probability, and latency.
• Scenario-based analysis: We analyze the system performance under various real-world scenarios, serving dense ground, indoor, or aerial user populations in dynamic 3D environments.

The remainder of this paper is organized as follows. Section 2 presents related work and identifies the gaps in the field of UAVs that our work addresses. Section 3 describes the system model and the formulation of problems. Section 4 presents the proposed optimization framework. Section 5 discusses the simulation results and performance analysis. Finally, Section 6 concludes the paper and outlines future research directions.

2. Related Work

The use of RISs has emerged as a pivotal strategy for manipulating electromagnetic waves to improve signal quality and spectral efficiency. Zhang et al. [16] introduce AIRIS, an approach that integrates AI into RIS-assisted signal processing, demonstrating notable improvements in handling complex channel conditions. Complementing this perspective, Alexandropoulos et al. [17] used pervasive machine learning frameworks that dynamically adjust RIS configurations to create smart radio environments, thus emphasizing the potential of learning-based strategies in mitigating the adverse effects of multipath reflections and interference. Furthermore, Triwidyastuti et al. [18] extend this concept to the security domain by employing transfer learning to enhance physical-layer security in aerial RIS-based networks, while Khan et al. [19] push the boundaries of conventional RIS architectures by exploring “beyond diagonal” designs for 6G Internet of Things (IoT) applications. Although these works demonstrate significant advances in RIS design and optimization, they predominantly address terrestrial scenarios or isolated aspects of RIS operation without fully accounting for the dynamic and heterogeneous environments encountered in NTNs.

In parallel with RIS research, NTNs have received considerable attention as a means to extend connectivity to areas where traditional infrastructure is insufficient or hard to reach. Saleh et al. [20] offer a comprehensive review of integrated terrestrial and non-terrestrial localization challenges in 6G, while Tong et al. [21] provides a comprehensive review of NTN developments in 3GPP Release-18, emphasizing adaptive beamforming and Doppler compensation for high-mobility users. The inherent challenges of high-mobility scenarios are further addressed in studies such as those by Korikawa et al. [22] and Haq et al. [23]. In addition, efficient path selection and integrated network architectures are proposed by Haq et al. [24] to improve connectivity under dynamic conditions, prioritizing latency and reliability. The need for a robust handover scheme in rapidly changing environments is underscored by Junejo et al. [25] and Eydian et al. [26], whose investigations into adaptive handover management and low-Earth-orbit (LEO) satellite handover strategies provide critical insights for NTN deployments. Moreover, Jamshed et al. [27] and Hossain and Vera-Rivera [28] highlight the role of AI and emerging technologies in converging the terrestrial and non-terrestrial domains. Despite these advances, most of the literature focuses on individual aspects of NTN performance, such as localization accuracy, path selection, and handover efficiency, leaving a gap in comprehensive studies that address the joint impact of RIS optimization in high-mobility NTN scenarios.

Efforts to achieve a unified communication framework for 6G have also explored spectrum sharing, advanced antenna configurations, and edge computing. Shang et al. [29] discuss innovative spectrum-sharing mechanisms in satellite-terrestrial integrated networks, while Ekpe et al. [30] provide detailed work on massive MIMO techniques tailored for NTNs. Jiang et al. [31] introduce a multi-agent federated reinforcement learning approach for satellite edge computing, and Li et al. [32] examine the synergistic benefits of integrated environment sensing with green communication strategies in NTNs. These studies collectively underscore the importance of holistic system design; however, they tend to overlook the specific challenges associated with real-time RIS reconfiguration in high-mobility environments, which are critical when considering the diverse propagation characteristics and rapid channel variations typical of scenarios such as high-speed trains, aerial networks, and emergency communications, as shown in

Table 1. Comparison of related works with our contributions.

Category	Citations	Key Contributions	Limitations/Challenges	How Our Work Differs
Intelligent Signal Processing	Zhang et al. [16] Alexandropoulos et al. [17] Khan et al. [19]	AI-based adaptive phase-shift configuration. ML techniques for RIS reconfiguration.	Focus primarily on terrestrial or quasi-static channels. Limited exploration of dynamic NTN scenarios.	Proposes a joint optimization framework for RIS phase shifts in high-mobility NTN environments. Incorporates real-time reconfiguration tailored for 6G networks.
Integration of NTNs in 6G Architectures	Saleh et al. [20] Tong et al. [21] Haq et al. [23] Shang et al. [29] Ekpe et al. [30]	System-level integration of terrestrial and non-terrestrial networks. Spectrum sharing frameworks and massive MIMO techniques.	Limited emphasis on the interaction between RIS deployment and NTN performance under high-mobility conditions.	Provides a comprehensive simulation-based performance analysis across multiple scenarios (urban, rural, highway, indoor, emergency). Focuses on 6G-enabled drone and high-mobility communications.
Mobility Management	Korikawa et al. [22] Junejo et al. [25] Eydian et al. [26]	Advanced handover algorithms. Path selection techniques for NTN and LEO satellite networks.	Handover strategies are generally developed independently of RIS reconfiguration. Limited integration of physical-layer optimization.	Integrates real-time RIS reconfiguration with mobility management. Enhances handover performance while maintaining optimal signal quality in rapidly changing environments.
Security	Triwidyastuti et al. [18] Haq et al. [24] Jamshed et al. [27] Jiang et al. [31] Li et al. [32]	Physical-layer security enhancements. Use of transfer learning and edge computing for secure communications.	Typically focus on security and energy efficiency in isolation. Do not jointly optimize core performance metrics (e.g., SINR, throughput).	Simultaneously optimizes multiple performance metrics (SINR, throughput, energy efficiency, outage probability, latency). Addresses critical trade-offs for secure and efficient NTN operation in high-mobility scenarios.

.

Distinct Contributions and Novelty of This Work

While prior research has predominantly focused on RIS applications in static or terrestrial environments, this paper addresses the joint optimization of RIS phase-shift configuration and UAV mobility coordination for high-mobility 6G NTNs. Departing from studies limited to individual performance metrics or simplified channel assumptions, we propose a Doppler-resilient real-time optimization framework integrating the following:

(1)

A predictive RIS reconfiguration algorithm leveraging user trajectory forecasting and Doppler shift compensation;

(2)

Joint multi-objective optimization of SINR, throughput, outage probability, energy efficiency (EE), and latency within a unified formulation;

(3)

Comprehensive scenario-driven validation across urban, rural, emergency, highway, and indoor environments.

Extensive evaluations reveal critical performance trade-offs neglected in prior work and demonstrate significant improvements in weighted key performance indicator (KPI) satisfaction versus state-of-the-art baselines. This integrated solution establishes a scalable foundation for RIS-enhanced 6G NTN deployments under dynamic mobility constraints.

3. System Model and Problem Formulation

3.1. System Architecture for UAV-Enabled 6G NTNs

The proposed system comprises the NTN node, RIS, and a high-mobility user as depicted in the Figure 2. This architecture is designed to enhance connectivity in dynamic environments such as high-speed trains, airplanes, and autonomous vehicles. A control link facilitates logical signaling between the NTN node and the RIS, which is solely responsible for transmitting optimized phase-shift parameters and reconfiguration commands. This enables real-time RIS adjustments to accommodate changes in channel conditions, user mobility, and interference patterns, thereby ensuring coherent signal reflection and seamless handover in high-mobility scenarios.

An NTN node refers to a satellite or a UAV that transmits with power P. The RIS is modeled as a planar array consisting of M passive reflecting elements, each capable of imposing a reconfigurable phase shift ${\varphi }_m\in [0,2\pi )$. The user is a single-antenna receiver that operates under high-mobility conditions.

This hybrid NTN architecture, also known as the HAP model, achieves robust connectivity through mechanisms such as path diversity, redundancy, and extended coverage. Although the HAP offers low-latency localized service, a low-Earth-orbit (LEO) satellite ensures global coverage. The RIS intelligently optimizes the signal reflections from both the HAP and the LEO nodes to counteract the adverse effects of mobility-induced channel degradation, thus achieving effective load balancing and improved link reliability. This work assumes a passive, reflective RIS without embedded baseband signal processing or energy harvesting. Each element applies a discrete phase shift from a finite set (e.g., 4–6 bits of quantization). The RIS is assumed to be a non-amplifying, narrowband, continuous-phase metasurface with a shared controller via a backhaul or satellite/UAV-side interface. This assumption ensures low power consumption but requires accurate external CSI estimation and centralized control. Our model does not consider active RIS or hybrid reflective-transmissive metasurfaces, which are left for future work.

3.2. Received Signal

Let $h_d\in \mathbb{C}$ denote the direct complex channel gain from the NTN node to the user. Let $\mathbf{g}\in {\mathbb{C}}^M$ be the channel vector from the RIS to the user, where ${(·)}^H$ denotes the Hermitian (conjugate) transpose operation. This operation is used to compute the contribution of the reflected signal from the RIS. Similarly, let $\mathbf{h}\in {\mathbb{C}}^M$ represent the channel vector from the UAV to the RIS.

The RIS phase-shift matrix is defined as

$$\mathsf{\Phi }=diag\left(e^{j{\varphi }_1},e^{j{\varphi }_2},\dots ,e^{j{\varphi }_m}\right),$$

where ${\varphi }_m\in [0,2\pi )$ represents the phase shift induced by the m-th RIS element.

The received signal at the user is given by

$$y=\underset{\mathrm{Desired}\mathrm{signal}}{\underbrace{\left(h_d+{\mathbf{g}}^H\mathsf{\Phi }\mathbf{h}\right)s}}+\underset{\mathrm{Noise}}{\underbrace{n}},$$

(1)

where $s\sim \mathcal{CN}(0,P)$ is the transmitted symbol, modeled as a circularly symmetric complex Gaussian random variable with zero mean and power P; and $n\sim \mathcal{CN}(0,{\sigma }^2)$ denotes the additive white Gaussian noise (AWGN) with zero mean and variance ${\sigma }^2$.

3.3. Channel Modeling for UAV-Assisted Non-Terrestrial Networks

The channel model incorporates both the direct path between the NTN node and the user, and the reflected path through the RIS. These paths are subject to various wireless propagation phenomena, including path loss, shadowing, and small-scale fading.

3.3.1. Direct Path

Let $h_d\in \mathbb{C}$ denote the complex channel coefficient for the direct path between the NTN node and the user. This coefficient accounts for three key propagation effects: path loss, large-scale shadowing, and small-scale fading. It is modeled as

$$h_d=\underset{\mathrm{Path} \mathrm{loss}}{\underbrace{\sqrt{{\mathrm{PL}}_d}}}·\underset{\mathrm{Shadowing}}{\underbrace{\sqrt{{10}^{{\mathrm{Sh}}_d/10}}}}·\underset{\mathrm{Fading}}{\underbrace{{\tilde{h}}_d}},$$

(2)

where ${\mathrm{PL}}_d={\left({\displaystyle \frac{\lambda }{4\pi d_d}}\right)}^2$ represents the path loss for the direct link, with $\lambda$ denoting the carrier wavelength and $d_d$ the distance between the NTN node and the user. The variable ${\mathrm{Sh}}_d$ captures the shadowing effect, which is typically modeled as a log-normal random variable. Finally, ${\tilde{h}}_d$ represents the small-scale fading component, which is modeled as a complex Gaussian random variable to account for the multipath nature of the wireless environment.

3.3.2. Reflected Path

The reflected path consists of two segments: one from the NTN node to the RIS and the other from the RIS to the user. Let $\mathbf{h}\in {\mathbb{C}}^M$ denote the channel vector from the NTN node to the RIS, and $\mathbf{g}\in {\mathbb{C}}^M$ represent the channel vector from the RIS to the user, where M is the number of RIS elements. Both links are subject to path loss, large-scale shadowing, and small-scale fading.

The channel vector $\mathbf{h}$ from the NTN node to the RIS is modeled as

$$\mathbf{h}=\sqrt{{\mathrm{PL}}_h}·\sqrt{{10}^{{\mathrm{Sh}}_h/10}}·\tilde{\mathbf{h}},$$

(3)

where ${\mathrm{PL}}_h={\left({\displaystyle \frac{\lambda }{4\pi d_h}}\right)}^2$ denotes the path loss component, with $\lambda$ being the carrier wavelength and $d_h$ the distance between the NTN node and the RIS. The term ${\mathrm{Sh}}_h$ represents the shadowing effect, modeled as a log-normal random variable, and $\tilde{\mathbf{h}}\sim \mathcal{CN}(0,{\mathbf{I}}_M)$ accounts for small-scale fading, modeled as a complex Gaussian vector with zero mean and an identity covariance matrix.

Similarly, the channel vector $\mathbf{g}$ from the RIS to the user is given by

$$\mathbf{g}=\sqrt{{\mathrm{PL}}_g}·\sqrt{{10}^{{\mathrm{Sh}}_g/10}}·\tilde{\mathbf{g}},$$

(4)

where ${\mathrm{PL}}_g={\left({\displaystyle \frac{\lambda }{4\pi d_g}}\right)}^2$ is the path loss for the RIS–user link, with $d_g$ being the corresponding distance. The shadowing effect for this segment is represented by ${\mathrm{Sh}}_g$, and the small-scale fading is captured by $\tilde{\mathbf{g}}\sim \mathcal{CN}(0,{\mathbf{I}}_M)$, consistent with the fading model used for the NTN–RIS link.

The small-scale fading components ${\tilde{h}}_d$, $\tilde{h}$, and $\tilde{g}$ are modeled as circularly symmetric complex Gaussian random variables, i.e., ${\tilde{h}}_d,\tilde{h},\tilde{g}\sim \mathcal{CN}(0,1)$, corresponding to a Rayleigh fading environment. This assumption is appropriate for NLoS propagation scenarios, which are common in urban and indoor UAV deployments due to dense building obstructions and multipath reflections. However, in scenarios where a dominant LoS component is present, such as in rural or high-altitude NTN environments, a Rician fading model is more accurate. In such cases, the channel coefficients can be modeled as

$${\tilde{h}}_d=\sqrt{{\displaystyle \frac{K}{K+1}}}{h}_{\mathrm{LoS}}+\sqrt{{\displaystyle \frac{1}{K+1}}}{h}_{\mathrm{NLoS}},$$

(5)

where K is the Rician K-factor, and $h_{\mathrm{LoS}}$, $h_{\mathrm{NLoS}}\sim \mathcal{CN}(0,1)$. For simulation simplicity and consistency across diverse scenarios, we use the Rayleigh model as a baseline, while acknowledging that environment-aware selection of fading models is critical for real-world implementation.

3.3.3. Interference and Noise

The received signal at the user is affected by additive white Gaussian noise (AWGN) with variance ${\sigma }^2$. External interference is modeled as a random variable I, which depends on the scenario (e.g., high interference in urban environments).

Our system model assumes that the RIS controller estimates the relative propagation delays of the direct and reflected paths based on real-time location and channel measurements. The proposed RIS-assisted NTN architecture employs a single intelligent surface comprising passive reflecting elements subject to the unit-modulus phase constraint ${\varphi }_m\in [0,2\pi )$. Operating in the sub-6 GHz band (2.4 GHz), the system assumes block fading channels where coherence time exceeds the optimization interval. There is perfect synchronization between the NTN infrastructure and the RIS controller, with user trajectory and velocity obtained via onboard sensors or cooperative localization. Interference is modeled as a log-normal random variable $\mathcal{I}\sim \mathcal{LN}(\mu ,{\sigma }^2)$, where environment-dependent standard deviations follow empirically validated values. Doppler shifts are estimated through consecutive CSI samples at 1 ms intervals. These well-justified assumptions establish a reproducible evaluation framework for joint RIS–UAV optimization.

The feasibility of propagation delay estimation at the RIS controller stems from deterministic NTN geometry and modern localization capabilities. In UAV-/satellite-assisted deployments, RIS and NTN node positions are either fixed or tracked via GPS/inertial navigation systems. The location and velocity data of the users are acquired by the following:

• Direct feedback channels;
• Integrated sensor fusion;
• Cooperative localization protocols.

Leveraging this spatial awareness, geometric transmitter–RIS-receiver distances are approximated with sub-wavelength accuracy, enabling precise delay estimation. Further refinement is achieved through the following:

$$\widehat{\tau }=\underset{\tau }{argmin}{∥\mathbf{y}-\sum _{k=0}^{K-1}{\alpha }_k\mathbf{s}(\tau -{\tau }_k)∥}^2$$

(6)

which is applied during channel sounding, where $\mathbf{s}\left(\tau \right)$ denotes the pilot waveform. Recent RIS positioning studies report sub-nanosecond estimation errors (<1 ns) in LoS conditions, confirming practical viability for real-time adaptation. Residual errors remain within tolerable margins for sub-6 GHz NTN reconfiguration.

Given the deterministic geometry in NTN deployments (e.g., known locations of the RIS and the satellite), estimating path length differences and applying a timing advance or phase precompensation at the RIS or transmitter side is feasible.

4. Joint Optimization Framework for RIS Configuration and UAV Mobility

Joint optimization of RIS phase shifts in high-mobility NTN scenarios presents several significant challenges. Firstly, the inherently non-convex and unit-modulus-constrained nature of the optimization problem (that is, each RIS element must maintain a phase shift within $[0,2\pi )$) makes it computationally intensive and unsuitable for directly applying standard convex solvers. Secondly, in high-mobility environments such as UAV-assisted NTNs, CSI rapidly becomes outdated due to short coherence times and Doppler effects, thus demanding real-time or predictive optimization techniques. Furthermore, the dimensionality of the optimization problem scales linearly with the number of RIS elements, making it increasingly difficult to find optimal phase-shift configurations in large-scale RIS deployments. The coordination latency between the control node and RIS and the limited computational capacity of UAVs further constrain the feasibility of real-time reconfiguration. Furthermore, achieving joint optimization across multiple performance metrics, such as SINR, throughput, energy efficiency, and latency, often introduces conflicting trade-offs that complicate algorithm design. These challenges collectively require the development of lightweight, predictive, and adaptive algorithms capable of achieving near-optimal solutions in dynamic and resource-constrained NTN environments.

In this section, we present the proposed optimization framework for maximizing the performance of RIS-assisted UAVs in high-mobility scenarios. The framework focuses on optimizing the RIS phase shifts to enhance key performance metrics such as SINR, throughput, energy efficiency, and outage probability while minimizing latency. The framework consists of three main components:

• Real-time RIS phase-shift optimization.
• Predictive coordination between UAVs and RIS.
• Performance evaluation and trade-off analysis.

4.1. Real-Time RIS Phase-Shift Optimization

The heart of the proposed framework is the optimization of the RIS phase-shift matrix $\mathsf{\Phi }$ to maximize the SINR at the user. The corresponding optimization problem is formulated as follows.

4.1.1. Objective Function

The objective is to maximize the SINR at the user, which is defined as

$$\mathrm{SINR}={\displaystyle \frac{{\left|h_d+{\mathit{g}}^H\mathsf{\Phi }h\right|}^{2}P}{{\sigma }^2+I}},$$

(7)

where $h_d$ denotes the direct complex channel coefficient from the NTN node to the user, $h\in {\mathbb{C}}^M$ represents the channel vector from the NTN node to the RIS, and $g\in {\mathbb{C}}^M$ is the channel vector from the RIS to the user. The RIS phase-shift matrix $\mathsf{\Phi }\in {\mathbb{C}}^{M\times M}$ is a diagonal matrix given by

$$\mathsf{\Phi }=\mathrm{diag}\left(e^{j{\varphi }_1},e^{j{\varphi }_2},\dots ,e^{j{\varphi }_m}\right),$$

(8)

where ${\varphi }_M\in [0,2\pi )$ denotes the phase shift applied by the m-th RIS element. The variable P denotes the transmit power of the NTN node, and ${\sigma }^2$ is the noise variance experienced at the user.

The interference power I from other users or operators is modeled as a log-normal random variable, expressed as

$$I={10}^{{\displaystyle \frac{N(0,{\sigma }_I^2)}{10}}}·I_0,$$

(9)

where $N(0,{\sigma }_I^2)$ represents a zero-mean Gaussian random variable with variance ${\sigma }_I^2$, and $I_0$ is the average baseline interference power.

4.1.2. Optimization Problem

Maximize the SINR by optimizing the RIS phase shifts:

$$\underset{\mathsf{\Phi }}{max}{|h_d+g^H\mathsf{\Phi }h|}^2,$$

(10)

subject to

$${\varphi }_M\in [0,2\pi ),\forall m\in \{1,2,\dots ,M\}.$$

(11)

4.1.3. Solution

Step 1: Simplify the Objective Function

The objective function can be rewritten as

$$|h_d+g^H{\mathsf{\Phi }h|}^2={\left|h_d+\sum _{m=1}^M{g}_m^{*}{h}_m{e}^{j{\varphi }_m}\right|}^2,$$

(12)

where $g_m$ and $h_m$ are the m-th elements of $\mathit{g}$ and $\mathit{h}$, respectively. The optimal phase alignment is achieved when all components of the reflected path are coherently added to the phase of the direct path.

Step 2: Optimal Phase-Shift Design

To maximize the SINR, the RIS phase shifts should align the reflected signal with the direct signal. This is achieved by setting

$${\varphi }_m=-arg\left(h_d\right)+arg\left(g_m^{*}{h}_m\right),\forall m\in \{1,2,\dots ,M\}.$$

(13)

This ensures that the reflected signal is constructively added to the direct signal.

The proposed solution leverages constructive interference of direct and reflected signals to enhance the signal strength received at the user. This approach is particularly beneficial in environments with high mobility, where maintaining a reliable communication link is crucial.

4.2. Predictive Coordination Between NTN and RIS

In high-mobility scenarios, such as UAV or high-speed vehicular communication, the rapid variation in channel coefficients due to mobility-induced Doppler effects poses a significant challenge for the real-time configuration of the RIS. To address this, our framework includes a predictive coordination module that proactively tracks and forecasts channel behavior using both historical CSI and mobility information.

4.2.1. Channel Prediction with Mobility Awareness

We implement a recurrent neural network (RNN)-based predictor that leverages a sliding window of past channel estimates $\left(h(t-1),h(t-2),\dots ,h(t-n)\right)$, with the help of user trajectory, velocity vectors, and known RIS and NTN locations.

This model predicts the future states of the cascaded channels $h(t+1)$ and $g(t+1)$, as well as the direct path $h_d(t+1)$, allowing preoptimization of the RIS phase-shift matrix $\mathsf{\Phi }(t+1)$. The CSI prediction model is trained offline on synthetic and simulated NTN mobility data and updated online using observed channel behavior.

4.2.2. Doppler Effect Modeling

To capture the impact of high user velocity in UAV-assisted 6G non-terrestrial networks, Doppler shifts are incorporated into the channel model. The instantaneous complex baseband channel coefficients for each link, whether direct or RIS reflected, are expressed as time-varying functions.

Specifically, the direct link is modeled as

$$h_d\left(t\right)={\mathrm{PL}}_d·{10}^{{\mathrm{Sh}}_d/20}·{\tilde{h}}_d·e^{j(2\pi f_{D,d}t+{\varphi }_d)},$$

(14)

where ${\mathrm{PL}}_d$ is the path loss, ${\mathrm{Sh}}_d$ represents the shadowing component, ${\tilde{h}}_d\sim \mathcal{CN}(0,1)$ denotes the small-scale fading, $f_{D,d}$ is the Doppler shift, and ${\varphi }_d$ is the initial phase of the direct path.

For the reflected path via the RIS, each element contributes two separate Doppler-affected components: one from the NTN node to the RIS and one from the RIS to the user. The corresponding models are given by

$$\begin{array}{cc}\hfill h_m\left(t\right)& ={\mathrm{PL}}_h·{10}^{{\mathrm{Sh}}_h/20}·{\tilde{h}}_m·e^{j(2\pi f_{D,h_m}t+{\varphi }_{h_m})},\hfill \end{array}$$

(15)

$$\begin{array}{cc}\hfill g_m\left(t\right)& ={\mathrm{PL}}_g·{10}^{{\mathrm{Sh}}_g/20}·{\tilde{g}}_m·e^{j(2\pi f_{D,g_m}t+{\varphi }_{g_m})},\hfill \end{array}$$

(16)

where ${\tilde{h}}_m,{\tilde{g}}_m\sim \mathcal{CN}(0,1)$ represent the small-scale fading coefficients; $f_{D,h_m}$ and $f_{D,g_m}$ are the Doppler shifts for the NTN–RIS and RIS–user links, respectively; and ${\varphi }_{h_m}$, ${\varphi }_{g_m}$ are the corresponding initial phases.

These Doppler-induced components introduce time-varying phase rotations into the channel response. If left uncorrected, such rotations can cause misalignment of signal paths, resulting in destructive signal combining and an increased probability of outage.

4.2.3. Doppler-Aware RIS Phase Alignment

To maintain coherent combining of direct and reflected signals, the RIS phase shift for each element is adjusted to compensate for the Doppler-induced phase drift:

$${\varphi }_m(t+1)=-arg\left(h_d(t+1)\right)+arg\left(g_m^{*}(t+1)h_m(t+1)\right)-2\pi (f_{D,g_m}+f_{D,h_m})(t+1)$$

This ensures that the reflected signals remain phase-aligned with the Doppler-rotated direct signal. The Doppler shifts are estimated via differential phase tracking across successive CSI samples:

$${\widehat{f}}_{D,x}={\displaystyle \frac{1}{2\pi \Delta t}}arg\left({\displaystyle \frac{h_x\left(t\right)}{h_x(t-1)}}\right)$$

where $x\in \{d,h_m,g_m\}$, and $\Delta t$ is the channel sampling interval.

4.3. Performance Evaluation and Trade-Off Analysis

The performance of the proposed framework is evaluated using several key metrics.

4.3.1. Throughput

The system throughput is calculated using the Shannon capacity formula, which relates the achievable data rate to the available bandwidth and the received signal quality. It is expressed as

$$\mathrm{Throughput}=B·{log}_2(1+\mathrm{SIN}\mathrm{R}),$$

(17)

where B denotes the communication bandwidth, and SINR is the signal-to-interference-plus-noise ratio at the user.

The SINR for the considered system model is given by

$$\mathrm{SIN}\mathrm{R}={\displaystyle \frac{P{\left|h_d+{\sum }_{m=1}^M{g}_m^{*}{h}_m{e}^{j{\varphi }_m}\right|}^2}{{\sigma }^2+I}},$$

(18)

where P is the transmit power of the NTN node, ${\sigma }^2$ is the noise variance, and I denotes the aggregate interference power. The term $h_d$ represents the direct channel component, while the summation ${\sum }_{m=1}^M{g}_m^{*}{h}_m{e}^{j{\varphi }_m}$ accounts for the combined contribution of all M RIS elements. Here, $g_m^{*}$ and $h_m$ denote the m-th elements of the RIS-user and NTN-RIS channel vectors, respectively, and ${\varphi }_m\in [0,2\pi )$ is the phase shift applied by the m-th RIS element.

The phase shifts $\left\{{\varphi }_m\right\}$ are dynamically optimized to achieve constructive interference at the receiver by maximizing the magnitude of the overall signal. This coherent combination of the direct and reflected signal components significantly enhances the SINR, thereby improving the system throughput.

4.3.2. Energy Efficiency

Energy efficiency is a key performance metric that quantifies how effectively the system converts consumed power into useful throughput. It is defined as the ratio of the system’s throughput to its total power consumption:

$$\mathrm{Energy}\mathrm{Efficiency}={\displaystyle \frac{\mathrm{Throughput}}{P_{\mathrm{total}}}},$$

(19)

where $P_{\mathrm{total}}$ denotes the total power consumed by the system.

The total power consumption is modeled as

$$P_{\mathrm{total}}=P_{\mathrm{NTN}}+M·P_{\mathrm{RIS}},$$

(20)

where $P_{\mathrm{NTN}}$ is the transmit power of the NTN node, $P_{\mathrm{RIS}}$ is the power consumed per RIS element, and M represents the total number of RIS elements deployed in the system.

By intelligently optimizing the RIS phase shifts and efficiently managing power resources at both the NTN and RIS sides, the proposed framework significantly improves energy efficiency while maintaining high data rates.

4.3.3. Outage Probability

The probability of failure quantifies the likelihood that the quality of the received signal is insufficient to meet a predefined quality of service requirement. Specifically, it is defined as the probability that the instantaneous SINR falls below a given threshold $\gamma$. Mathematically, it is expressed as

$$P_{\mathrm{out}}=\mathbb{P}\left[\mathrm{SINR}<\gamma \right]$$

(21)

where $\gamma$ is the minimum SINR required to sustain reliable communication.

To reduce the risk of outages, the proposed system employs predictive coordination mechanisms and proactive RIS reconfiguration strategies. These techniques adaptively track channel variations and mobility dynamics, thus minimizing the probability of outages and maintaining a high level of communication reliability even in rapidly changing high-mobility environments.

4.3.4. Latency

Latency is the sum of the processing delay and transmission delay:

$$\mathrm{Latency}=T_{\mathrm{processing}}+T_{\mathrm{transmission}},$$

(22)

where $T_{\mathrm{processing}}$ depends on the number of RIS elements M, and $T_{\mathrm{transmission}}$ is a fixed delay. By optimizing the RIS configuration and predicting channel changes, we minimize the processing delay, thereby reducing the overall latency.

The proposed predictive coordination framework between the NTN and RIS shows significant improvements in throughput, energy efficiency, and reliability while maintaining low latency. The proactive approach to channel prediction, RIS reconfiguration, and handover management ensures that the system can adapt to rapid changes in channel conditions, making it highly suitable for high-mobility scenarios. The detailed step-by-step procedure to implement this predictive coordination mechanism is outlined in Algorithm 1.

Algorithm 1: Doppler-Aware RIS-Assisted NTN Optimization

Require:  Initial phase vector ${\mathit{\varphi }}^{\left(0\right)}$, step size $\eta \in (0,1]$, threshold $ϵ$, max iterations $K_{max}$ Ensure:  Optimized Doppler-compensated phase vector ${\mathit{\varphi }}^{*}$ and performance metrics  1: Inputs: Channels at $t-1,t$: $\{h_d(t-1),h_d\left(t\right)\}$, ${\{h_m(t-1),h_m\left(t\right)\}}_{m=1}^M$, ${\{g_m(t-1),g_m\left(t\right)\}}_{m=1}^M$; user trajectory and velocity; time step $\Delta t$; powers $P_{\mathrm{NTN}},P_{\mathrm{RIS}}$; noise variance ${\sigma }^2$; interference I; bandwidth B; SINR threshold $\gamma$  2: Initialize iteration counter: $k\leftarrow 0$  3: while true do  4:    Step 1: Channel and Doppler prediction  5:    Predict next-slot CSI: ${\widehat{h}}_d(t+1)$ and ${\{{\widehat{h}}_m(t+1),{\widehat{g}}_m(t+1)\}}_{m=1}^M$  6:    Estimate Doppler shifts: for each m, ${\widehat{f}}_{D,h_m}={\displaystyle \frac{1}{2\pi \Delta t}}arg\left({\displaystyle \frac{h_m\left(t\right)}{h_m(t-1)}}\right)$, ${\widehat{f}}_{D,g_m}={\displaystyle \frac{1}{2\pi \Delta t}}arg\left({\displaystyle \frac{g_m\left(t\right)}{g_m(t-1)}}\right)$, ${\widehat{f}}_{D,h_d}={\displaystyle \frac{1}{2\pi \Delta t}}arg\left({\displaystyle \frac{h_d\left(t\right)}{h_d(t-1)}}\right)$  7:    Step 2: Compute ideal phase shifts  8:    for $m=1$ to M do  9:      ${\varphi }_m^{\mathrm{ideal}}=-arg({\widehat{h}}_d(t+1))+arg({\widehat{g}}_m^{*}(t+1){\widehat{h}}_m(t+1))-2\pi ({\widehat{f}}_{D,g_m}+{\widehat{f}}_{D,h_m})\Delta t$ 10:    end for 11:    Step 3: Phase update (unit modulus) 12:    for $m=1$ to M do 13:      ${\varphi }_m^{(k+1)}=(1-\eta ){\varphi }_m^{\left(k\right)}+\eta {\varphi }_m^{\mathrm{ideal}}mod2\pi$ 14:    end for 15:    Step 4: Convergence check 16:    if $\parallel {\mathit{\varphi }}^{(k+1)}-{\mathit{\varphi }}^{\left(k\right)}{\parallel }_2\le ϵ$ or $k\ge K_{max}$ then 17:      break 18:    end if 19:    $k\leftarrow k+1$ 20: end while 21: Step 5: Performance evaluation at $t+1$ 22: $\mathrm{SINR}={\displaystyle \frac{|{\widehat{h}}_d(t+1)+{\widehat{\mathbf{g}}}^H(t+1)\mathrm{diag}(e^{j{\mathbf{\varphi }}^{(k+1)}})\widehat{\mathbf{h}}(t+1){|}^2·P_{\mathrm{NTN}}}{{\sigma }^2+I}}$ 23: $\mathrm{Throughput}=B{log}_2(1+\mathrm{SINR})$ 24: $\mathrm{EE}={\displaystyle \frac{\mathrm{Throughput}}{P_{\mathrm{NTN}}+M{P}_{\mathrm{RIS}}}}$ 25: $\mathrm{Outage}=Pr(\mathrm{SINR}<\gamma )$ 26: $\mathrm{Latency}=T_{\mathrm{proc}}+T_{\mathrm{tx}}$ 27: Set ${\mathit{\varphi }}^{*}={\mathit{\varphi }}^{(k+1)}$ 28: Output: ${\mathit{\varphi }}^{*}$ and performance metrics

5. Simulation Results and Discussion

All scenario-specific parameters were selected in a manner consistent with the principles outlined in the 3GPP TR 38.821 [33] ensuring that our simulation assumptions remain traceable to industry standards. We have carefully considered the recommended system-level assumptions, environmental characteristics, and propagation conditions provided in the report to design representative and practical simulation scenarios.

The selection of specific parameters as mentioned in

Table 2. Scenario-specific parameters.

Parameter	Value/Description
M_values	Number of RIS elements (10 to 50)
num_trials	1000 trials for statistical significance
Carrier Frequency ($f_c$)	2.4 GHz
Speed of Light (c)	$3\times {10}^8$ m/s
Transmit Power (P)	1 W
Noise Variance (${\sigma }^2$)	$1\times {10}^{-3}$ W
Bandwidth (B)	10 MHz
Power per RIS Element ($P_{\mathrm{RIS}}$)	0.1 W
NTN Power ($P_{\mathrm{NTN}}$)	10 W
Processing Time per Element (${\tau }_{\mathrm{RIS}}$)	1 μs
Transmission Delay ($T_{\mathrm{tx}}$)	1 ms
Scenarios	Urban, rural, highway, indoor, emergency
Scenario Variations	Different interference scales and shadowing standard deviations

and their technical justifications are as follows.

• RIS element count:
  $M\in \{10,20,30,40,50\}$ investigates performance–complexity trade-offs, with $M=50$ representing the current practical limit for real-time reconfiguration in existing RIS hardware.
• Carrier frequency:
  $f_c$ = 2.4 GHz operates in sub-6 GHz UAV bands, optimizing Doppler resilience and propagation characteristics relative to millimeter-wave alternatives.
• Power parameters:
  Transmit power $P=1$ W and per-element RIS consumption $P_{\mathrm{RIS}}=0.1$ W reflect energy-constrained NTN design principles established in the contemporary literature.
• Noise characterization:
  Variance ${\sigma }^2={10}^{-3}$ W; models thermal noise for a 10 MHz bandwidth at 300 K.
• Mobility modeling:
  User velocities $v\in [10,100]$ m/s follow scenario-specific UAV speed distributions representative of operational aerial platforms.
• Interference profile:
  Log-normal distribution with environment-dependent standard deviations (${\sigma }_{\mathrm{urban}}=6$ dB, ${\sigma }_{\mathrm{rural}}=4$ dB) that capture realistic NTN channel conditions.
• Processing latency:
  Per-element delay ${\tau }_{\mathrm{RIS}}=1$ μs; benchmarks state-of-the-art RIS controller architectures.
• Doppler compensation:
  The frequency shift $\Delta f=v·f_c·cos\theta /c$ allows for the predictive alignment of the phase, where $\theta$ denotes the angle of motion relative to the propagation of the signal.

Parameters not explicitly varied (e.g., Rician K factor, mobility patterns) remained scenario-constant to isolate RIS-scaling effects. Future work will investigate the adaptation of dynamic parameters under practical constraints for the deployment of NTNs.

The simulation results provide a comprehensive analysis of the performance of RIS-assisted NTNs in five scenarios (urban, rural, highway, indoor, emergency) with five metrics (SINR, throughput, energy efficiency, outage probability, latency), as shown in Figure 3. The following is a logical breakdown of the findings.

5.1. Analysis of Plotted Results Across Different Scenarios

The performance of each scenario is analyzed based on five key metrics: SINR, throughput, energy efficiency, outage probability, and latency. The following sections break down the trends observed for each metric.

5.1.1. Signal-to-Interference-Plus-Noise Ratio (SINR) [dB]

Trend: The urban scenario starts with a relatively high SINR, which fluctuates slightly and then gradually declines as M increases. The highway scenario shows a steady improvement in SINR, indicating that more RIS elements benefit the signal quality in this environment. The rural scenario exhibits a consistent decline in SINR, which may be attributed to increased interference or unfavorable propagation conditions. The indoor scenario remains mostly stable with slight variations, showing little sensitivity to RIS element growth. The emergency scenario follows a decreasing trend, with its SINR starting relatively high but dropping over time, possibly due to interference effects in dynamic emergency conditions.

Interpretation: Increasing the number of RIS elements M does not always lead to higher SINR in all environments. Although the highway scenario benefits from additional RIS elements due to relatively predictable signal propagation, other environments such as urban and rural environments experience limited improvements or declining SINR trends, likely due to interference and environmental obstacles. The emergency scenario, which starts with a high SINR, suggests an initial benefit from RIS deployment but suffers degradation at higher values of M, potentially due to complex signal interactions in critical situations. The stability of the indoor scenario indicates that RIS elements have minimal impact in controlled environments where multipath effects dominate.

This analysis highlights that the impact of the deployment of RIS is highly environment-dependent and simply increasing M may not always yield better signal quality.

5.1.2. Throughput [Mbps]

Trend: The urban scenario shows a steady but slow increase in throughput as M increases. The rural scenario initially remains stable but experiences a significant increase at higher M, indicating that more RIS elements improve performance in this environment. The highway scenario starts with the highest throughput but slightly declines as M increases, suggesting that additional RIS elements may not be as beneficial due to possible interference or diminishing returns. The indoor scenario exhibits a gradual increase in throughput, while the emergency scenario declines slightly with more RIS elements, likely due to interference effects or environmental challenges.

Interpretation: Increasing the number of RIS elements does not always lead to a uniform throughput improvement in all environments. While rural and indoor scenarios benefit from higher M, highway and emergency conditions show signs of saturation or interference at higher RIS deployments. Environmental factors such as interference, multipath propagation, and deployment conditions play a crucial role in throughput performance across different scenarios.

5.1.3. Energy Efficiency [Mbps/W]

Trend: The urban and indoor scenarios show a consistent increase in energy efficiency as the number of RIS elements increases, highlighting the effectiveness of RIS in these environments. The rural scenario starts by increasing while the highway scenario starts by decreasing but both scenarios later maintain steady energy efficiency across different M values, indicating stable power utilization. The emergency scenario starts with relatively high energy efficiency but this gradually declines, potentially due to increased power consumption outweighing performance gains.

Interpretation: The increase in energy efficiency for the urban and indoor scenarios suggests that RIS deployment can significantly improve power utilization in challenging environments. However, the decreasing trend in the emergency scenario emphasizes the trade-off between increasing RIS elements and diminishing energy returns under extreme conditions. Balancing performance and energy efficiency is critical, especially for NTN systems operating under diverse environmental constraints.

5.1.4. Outage Probability [%]

Trend: The urban scenario initially has a moderate probability of outages, but this improves steadily as the number of RIS elements M increases. The rural scenario starts with a lower probability of outage but it increases over time, surpassing the urban scenario in the higher-M range. The highway scenario remains relatively stable with slight variations. The indoor scenario exhibits a consistent probability of outage, showing little impact from increasing M. The emergency scenario begins with the highest outage probability which gradually decreases, indicating some improvement but still maintaining a relatively higher outage compared to other environments.

Interpretation: Increasing the number of RIS elements generally helps reduce outage probability in urban and emergency environments. However, rural environments experience an increasing outage trend, possibly due to changing propagation conditions or interference. The indoor environment remains unaffected, suggesting that RIS elements may have a limited impact in highly controlled settings.

It is important to note that the outage probability levels observed in the evaluated scenarios remain above 10%, which is not sufficient for ultra-reliable low-latency communication (URLLC) or other mission-critical applications. This is mainly due to practical system constraints, including imperfect CSI acquisition, rapid CSI aging caused by high Doppler shifts, limited RIS quantization resolution, and mobility-induced phase misalignment. Future work will focus on improving the accuracy of the real-time CSI estimation, integrating predictive Doppler compensation, and developing faster RIS reconfiguration strategies to reduce the probability of outages towards reliability-critical targets.

5.1.5. Latency [ms]

Trend: Latency generally decreases in the urban and emergency scenarios as M increases, indicating an improvement in response with more RIS elements. The rural scenario maintains a low latency for all M values, whereas the indoor and highway scenarios show a slight increase in latency as M increases, possibly due to processing overhead.

Interpretation: RIS-assisted NTN systems effectively manage latency, with notable improvements in the urban and emergency scenarios as RIS elements increase. While some scenarios (e.g., indoor) experience slight latency growth, the overall impact remains within acceptable limits for low-latency applications like vehicular and emergency networks.

The results of this study provide valuable insights for the design and implementation of RIS-assisted NTN systems, highlighting the potential of RIS technology to significantly improve communication performance in challenging environments.

5.2. Performance Trade-Offs and Design Implications

The integration of RISs with NTNs introduces critical trade-offs between performance metrics. These trade-offs are rooted in the mathematical relationships governing the system model and are evident in the simulation results. The analysis presented in

Table 3. Comparison of performance metrics.

Category	SINR vs. Energy Efficiency	Throughput vs. Latency	Outage Probability vs. Complexity
Observation	SINR (dB): Urban scenarios exhibit a stable yet slightly declining SINR beyond $M=30$. Highway scenarios show steady improvement in SINR with increasing M, while rural and emergency scenarios face SINR degradation due to interference. Indoor scenarios maintain a stable SINR irrespective of M. Energy Efficiency (Mbps/W): Urban and indoor scenarios show consistent increases in energy efficiency with M, while rural and highway scenarios stabilize beyond $M=30$. Emergency scenarios decline in energy efficiency at higher M due to increased power consumption.	Throughput (Mbps): Throughput improves across most scenarios with M, peaking in rural and indoor scenarios at higher M. Highway scenarios show diminishing returns after an initial rise, and emergency scenarios decline slightly with larger M. Urban throughput grows steadily but modestly. Latency (ms): Latency decreases in urban and emergency scenarios with increasing M. Rural latency remains low and stable, while indoor and highway latency slightly increases due to processing overhead.	Outage Probability: Urban scenarios see steady improvement in outage probability as M increases. Emergency scenarios start with high outage probability but gradually improve. Rural scenarios, however, face increasing outage probability at larger M, potentially due to interference. Indoor scenarios remain unaffected. Complexity: Higher M increases computational complexity, especially in highway and emergency scenarios where real-time adjustments are critical.
Trade-Off	Higher M improves the SINR and energy efficiency but leads to rising power consumption in emergency scenarios. Optimal M: Around $M=30$ balances the SINR and energy efficiency in most scenarios, except rural and highway.	Throughput grows in rural and urban scenarios but comes with processing overhead for highway and indoor scenarios. Low Latency: Emergency and urban scenarios require $M\le 30$ for responsiveness, while higher M (e.g., $M=50$) benefits throughput at the cost of latency.	Reducing outage probability in rural and emergency scenarios requires higher M (e.g., $M=50$) but increases complexity. Optimal Trade-Off: Low-complexity networks can limit M to $M=20$, while mission-critical systems balance complexity and performance at $M=40$.
Design Implications	Scenario-Specific Tuning: Urban and emergency scenarios favor $M=20–30$ for balance, while rural systems maximize $M=50$ for improved SINR and throughput. Power Optimization: Emergency networks should optimize RIS configuration for energy efficiency.	Real-Time Applications: Urban and emergency networks prioritize $M=30$ for latency-sensitive use cases. Bandwidth-Intensive Applications: Highway and rural scenarios can utilize $M=50$ to maximize throughput despite increased latency.	Mission-Critical Systems: Emergency scenarios require $M=40–50$ for reliability but need advanced algorithms to manage complexity. Algorithm Choice: Small M suits simple systems, while larger M necessitates machine learning-based optimization for dynamic configurations.

and

Table 4. Summary of trade-offs.

Trade-Off	Key Observation	Optimal Design Choice
SINR vs. Energy Efficiency	SINR saturates at $M=30$; energy efficiency peaks at $M=30$.	$M=30$ (peak efficiency).
Throughput vs. Latency	Throughput grows logarithmically; latency increases linearly.	$M=30–40$ (balance for real-time systems).
Outage vs. Complexity	Outage drops exponentially; complexity increases quadratically.	$M=40–50$ for reliability, $M=20–30$ for low complexity.

provides a detailed evaluation of the performance metrics and trade-offs for RIS-assisted 6G-enabled drone communication systems and other high-mobility NTN scenarios.

Urban systems tend to favor smaller M values, typically in the range of $M=20–30$, not only due to interference limitations but also due to deployment constraints such as restricted space for RIS placement and power limitations, as higher RIS element counts increase power consumption without proportional gains. Rural systems, on the other hand, prioritize the extension of coverage, making a larger M (e.g., $M=50$) more justifiable despite the increase in power consumption and diminishing returns in SINR and throughput beyond $M=30$. This highlights the importance of scenario-specific tuning for optimal performance. A key takeaway is the diminishing beamforming gains and increasing power consumption observed at higher RIS element counts, particularly in emergency scenarios, where energy efficiency starts declining at $M>30$. The balance between SINR, energy efficiency, and environmental constraints further underscores the need for adaptive RIS deployment strategies tailored to specific environments. Another critical trade-off identified is between the probability of outages and computational complexity. Mission critical systems, such as emergency networks, require $M=40–50$ to achieve ultra-reliability ($P_{\mathrm{out}}<1\%$). However, this increase in M significantly increases computational complexity, as RIS reconfiguration times increase quadratically with M (for example, approximately 50 ms for $M=50$). Although these systems can tolerate the additional complexity due to their need for reliability, simpler networks with $M=20–30$ achieve a better balance between the probability of outage and the computational overhead. However, they still experience higher outage probabilities (10–15%) in environments with unpredictable interference or limited propagation conditions, such as rural or emergency scenarios. For real-time applications sensitive to latency, such as drone communication or autonomous vehicular networks, urban and emergency scenarios benefit from $M\le 30$, which ensures reduced latency and maintains computational efficiency. In contrast, bandwidth-intensive applications, such as video streaming in rural or highway scenarios, can tolerate slightly higher latency in exchange for maximizing throughput at $M=50$.

In practical deployments, maximizing the SINR typically requires increasing M (the number of RIS elements) to improve beamforming gain. However, this introduces critical trade-offs: power consumption scales with M due to per-element requirements, while processing latency increases with real-time phase-shift computations, both adversely impacting energy efficiency and system responsiveness. Furthermore, environmental constraints, including obstacle density, user mobility patterns, signal blockage, and interference levels, fundamentally constrain the marginal utility of additional elements. For example, in urban environments with severe multipath propagation and dense interference, large RIS arrays exhibit diminishing returns due to constructive–destructive interference phenomena. In contrast, in rural/highway settings with dominant LoS paths, gains scale more linearly with M.

Consequently, adaptive deployment strategies are essential, requiring dynamic joint optimization of the following based on real-time measurements and environmental classification:

(i)

RIS element count (M);

(ii)

Physical placement geometry;

(iii)

Phase-shift update algorithms.

For latency-sensitive applications (e.g., emergency response, autonomous navigation), smaller arrays with computationally efficient heuristics prove optimal. Conversely, coverage-oriented deployments benefit from larger arrays employing offline-trained predictive optimization. This context-aware adaptation ensures Pareto-optimal balancing of the SINR, energy efficiency, and latency under practical constraints.

Overall, the analysis emphasizes the critical need for environment-aware RIS tuning and algorithmic approaches (e.g., gradient-based optimization for smaller M versus machine learning models for higher M) to address the performance trade-offs in diverse NTN and high-mobility scenarios. These findings underscore the importance of adaptive RIS configurations, where dynamic tuning of M and phase shifts is essential to address varying environmental conditions and application priorities. Furthermore, algorithm selection plays a crucial role, with a smaller M enabling gradient-based optimization and larger M necessitating machine learning for real-time operation. However, the results highlight the potential of RIS-assisted NTNs to enable ultra-reliable, low-latency communication in 6G networks, particularly for high-mobility use cases like drone swarms and autonomous systems.

6. Conclusions and Future Directions

This article investigates the integration of RISs with NTNs to address critical connectivity challenges in high-mobility scenarios, such as autonomous vehicles, drone swarms, and aerial networks. Our results demonstrate that RIS-assisted NTNs significantly enhance key performance metrics such as the SINR, throughput, and reliability in dynamic environments; however, their effectiveness depends on achieving an optimal trade-off among interdependent and often competing parameters. For instance, while increasing the number of RIS elements (M) improves signal quality (e.g., SINR rising to 15 dB in urban environments) and reduces outage probability (to <1% for $M=50$), it introduces trade-offs such as higher energy consumption (e.g., a 20% efficiency drop beyond $M=30$), increased latency (up to 0.05 ms for $M=50$), and quadratic growth in computational complexity. These findings underscore the necessity of scenario-specific adaptive configurations: urban interference-heavy environments favor a smaller M (e.g., $M=20$) to optimize energy efficiency, while rural areas benefit from a larger M (e.g., $M=50$) to maximize coverage and reliability. Similarly, mission-critical applications like emergency networks prioritize ultra-reliability ($P_{\mathrm{out}}<1\%$) at the cost of complexity, whereas latency-sensitive systems such as autonomous drones require lightweight RIS configurations ($M\le 30$) to meet real-time demands.

Future research will focus on real-time optimization of RIS-assisted NTNs using advanced machine learning techniques such as deep reinforcement learning and federated learning. The integration of RIS with IoT and smart city infrastructure will be explored to enhance connectivity in diverse environments. Channel modeling for mmWave/THz bands and 3D mobility will support more accurate performance analysis. In addition, hardware prototyping and standardization efforts will ensure scalable, energy-efficient, and interoperable RIS-NTN systems. The future development of RIS-assisted UAV networks can greatly benefit from the integration of intelligent multimodal sensing and communication frameworks. Emerging techniques, such as digital twins in real time [34] and LiDAR-assisted channel modeling [35] offer the potential to improve environmental awareness, user localization, and obstacle mapping, which are crucial for dynamic aerial deployments. By combining LiDAR sensing, GPS data, and visual sensors, UAVs can proactively detect changes in the propagation environment, enabling the RIS to adapt its phase shifts more effectively in real-time. Furthermore, multimodal sensing can support more accurate CSI prediction and trajectory-aware RIS configurations, thus mitigating the negative effects of Doppler shifts and fast fading in high-mobility scenarios. Incorporating these sensing-communication integrations into our proposed optimization framework would allow future aerial networks to achieve enhanced reliability, lower outage probability, and better support for URLLC applications.