WorksEngineering
Dynamic-Aware LC-DMA Beam Tracking for LEO Satellite Communications
Abstract
Liquid crystal (LC)-based dynamic metasurface antennas (DMAs) offer a power-efficient, low-cost, and scalable solution for low Earth orbit (LEO) satellite communications where large apertures are needed to compensate for large path loss in the Earth-satellite link. However, beam tracking with LC-DMAs is governed by two coupled dynamics: the required time-varying beam direction caused by user and satellite mobility, and the finite response time of the LC phase shifters. This raises the question of whether LC-DMAs can reconfigure sufficiently fast to support high-mobility LEO satellite links. To address this problem, we develop a dynamic-aware beam-tracking framework that explicitly accounts for both dynamics. In particular, we formulate an optimization problem to maximize the cumulative received signal-to-noise ratio (SNR) over a predictive tracking horizon subject to the physical dynamics of the LC unit cells as well as transmit power constraints. To solve this non-convex problem efficiently, we develop an alternating optimization framework combining a closed-form maximum ratio transmission (MRT) digital precoder and maximum ratio combining (MRC) with a parallel, closed-form dynamic-aware phase-shift update. Simulation results show that when the time response of LC-DMAs is properly accounted for, it can support practical mobility scenarios. Moreover, the proposed dynamic-aware algorithm maintains stable SNR and outperforms conventional designs, particularly at high vehicle velocities and low orbital azimuth angles.
Introduction
Non-terrestrial networks (NTNs) utilizing low Earth orbit (LEO) satellites have emerged as a key component for 6G wireless connectivity and providing continuous global coverage to remote areas and mobile devices [1], [2]. Nevertheless, establishing reliable uplink communication between LEO satellites and high-speed vehicles presents several challenges. In particular, high beamforming gains are required to compensate for the large path loss, and adaptive beamforming with continuous beam tracking is needed to accommodate rapid LEO satellite motion and vehicle mobility [3].
Conventional phased arrays rely on numerous high-resolution phase shifters and dedicated power amplifiers. This leads to prohibitive power consumption and hardware cost when scaled to large aperture sizes [4]. To overcome these limitations, dynamic metasurface antennas (DMAs) are considered as an energy-efficient alternative for holographic beamforming [5]. A DMA integrates dense sub-wavelength metamaterial radiating elements directly onto microstrip feed lines driven by a reduced number of radio frequency (RF) chains compared to the conventional phased arrays. By dynamically tuning the Lorentzian resonance of individual unit cells, DMAs achieve flexible spatial multiplexing and beam focusing directly in the electromagnetic domain and at the same time substantially reduce the RF hardware overhead [5], [6].
Among various reconfigurable technologies (e.g., PIN diodes and varactors), liquid crystal (LC) technology provides continuous phase tunability, high operating frequencies (into millimeter-wave and terahertz regimes), and near-zero static direct current (DC) power consumption [7]–[9]. In an LC-based DMA, applying a low-frequency bias voltage rotates the anisotropic nematic LC molecules, which alters the effective local dielectric permittivity and shifts the resonance frequency of the metamaterial element [5], [10]. These properties make LC-DMAs an attractive choice for large-aperture, energy-efficient satellite terminals.
Nevertheless, beam tracking with LC-DMAs is governed by two coupled dynamics: the mobility-induced dynamics of the vehicle-to-satellite link and the internal dynamics of the LC phase shifters. The former continuously changes the required beam direction due to vehicle and satellite mobility, while the latter limits how fast the beam can be reconfigured. In particular, the molecular reorientation speed of LC mixtures is fundamentally governed by rotational viscosity and elastic restoration forces, which results in switching response times on the order of milliseconds [11]. This mismatch between the required and achievable beam dynamics may become a performance bottleneck in LEO satellite systems. Conventional beam-tracking algorithms assume instantaneous phase transitions and aim to maximize the instantaneous signal-to-noise ratio (SNR). These algorithms may demand large phase changes across consecutive tracking steps that cannot be realized instantaneously with LC-DMAs, and hence result in transition periods during which the received SNR can severely degrade. This interplay between mobility and hardware dynamics raises the question of whether LC-DMAs can reconfigure sufficiently fast to support high-mobility LEO satellite links.
To address this challenge, we develop a dynamic-aware beam-tracking framework for LC-DMAs in vehicle-to-LEO satellite uplink communications that explicitly accounts for both mobility and LC hardware dynamics. The main contributions of this work are summarized as follows:
We establish an analytical framework for LC-based DMA transmitters (Txs) in satellite uplink channels that jointly models guided-wave propagation and the physical rise/decay transition dynamics of the LC phase shifters in Section II.
We formulate a joint digital beamforming and dynamic-aware LC phase-reconfiguration problem to maximize the cumulative received SNR over a predictive tracking horizon, subject to the physical dynamics and phase limits of the LC unit cells and the transmit power constraint in Section III.
We propose an alternative optimization (AO) algorithm that alternates between a parallelized, closed-form dynamic-aware phase-shift optimization with linear complexity in the number of DMA radiating elements in Section III-A, and a closed-form digital precoder and combiner in Section III-B.
In Section IV, numerical simulations show that, when the time response of LC-DMAs is properly accounted for, they can support practical mobility scenarios. Moreover, the proposed dynamic-aware design maintains stable SNR and outperforms conventional designs, e.g., [6], particularly at high vehicle velocities and low orbital azimuth angles.
Notation: Vectors and matrices are denoted by boldface lowercase () and boldface uppercase letters (), respectively, while scalars are represented by standard italic letters (). The sets of real and complex numbers are denoted by and , respectively, with and representing the spaces of -dimensional real vectors and complex matrices. The transpose, Hermitian, and Kronecker-product operators are denoted by , , and , respectively. The Euclidean norm of vector is denoted by , , and denote the absolute and conjugate values of scalar , respectively. The operator forms a diagonal matrix with the elements of vector along its main diagonal, and denotes the -th element of vector . The identity matrix, the all-zeros vector, and the all-ones vector are denoted by , , and , respectively. The statistical expectation is represented by , and denotes a circularly symmetric complex Gaussian random vector with mean and covariance matrix . Finally, and denote the natural logarithm and big-O notation, respectively.
System, DMA and Dynamic LC Models
In this section, we first present the considered system model, followed by the channel model. Finally, we discuss LC-based DMA time response model.
System model
We consider a narrowband uplink communication scenario where a moving vehicle equipped with an LC-based DMA transmits data to a LEO satellite. The satellite is equipped with an -element antenna array. The DMA Tx consists of microstrip/waveguide feed lines (each connected to a dedicated RF chain) driving a total of metamaterial radiating elements, where denotes the number of elements coupled to the -th feed line. The received baseband signal at the satellite is modeled as
where represents the wireless channel matrix from the metasurface elements to the satellite array, is the transmit signal vector, and , , denotes the additive white Gaussian noise (AWGN) at the satellite receiver (Rx) with variance . The transmit signal vector is expressed as , where is the baseband digital beamformer satisfying (with being the maximum transmit power), and is the transmitted symbol with normalized power . The spatial-reconfiguration and feed-waveguide response of the DMA is represented by the block diagonal matrix :
where contains the effective radiating weights along feed line . The response of the -th element on feed where is given by
where is the tunable response induced by the LC state, and is the waveguide response, which is given by
where is the waveguide attenuation coefficient (accounting for dielectric/ohmic losses and the average energy radiation from previous elements2), is the guided-wave propagation constant, is the position of the -th radiator, and represents the feed origin of line .
Mobility-Induced Channel Dynamics
Given the typical operating environment of satellite communications, the propagation channel is dominated by the far-field line-of-sight (LoS) path. Assuming uniform planar arrays (UPAs) at both the vehicle Tx and the satellite Rx, the channel matrix is modeled as
where denotes the complex channel attenuation factor of the LoS link (capturing free-space path loss), while and represent the array response (steering) vectors of the Tx and Rx, respectively. The term specifies the angle of departure (AoD) toward the satellite relative to the Tx array coordinate system, with and denoting the elevation and azimuth angles, respectively. Correspondingly, denotes the angle of arrival (AoA) from the vehicle defined within the satellite Rx array local coordinate system. Due to the motion of both the vehicle and the LEO satellite, and vary over time, which makes channel matrix time-varying and necessitates adaptive beamforming and beam tracking.
LC Hardware Dynamics
In contrast to the mobility-induced channel dynamics, the LC-DMA exhibits internal hardware dynamics that determine how quickly its radiating elements can be reconfigured. The response of the -th metamaterial element coupled to the -th feed line of a DMA, denoted by , is governed by a Lorentzian resonance profile [5, 10]:
where . Here, is the oscillator strength, is the time-varying resonance frequency tuned by the applied bias voltage , and denotes the damping factor. For LC elements, the temporal evolution of under an applied voltage can be experimentally characterized [7, 9]. Moreover, in a DMA, the amplitude and phase are inherently coupled. Let denote the phase of . Under resonance-tuning, narrow-band, and passivity conditions, the element response at the operating carrier frequency can be parametrized in terms of its phase response as [5]
where represents the controllable phase shift of the -th element at time . Due to the finite reorientation speed of the LC molecules under applied electric fields, the phase shift cannot switch instantaneously. The transition model for the LC phase shifts for the -th element to transition from an initial phase (the known current state of the elements at the current time slot) to a desired phase is given respectively for and by [8]
where and denote the rise (electric field-driven) and decay (relaxation-driven) reorientation time constants of the LC material, respectively, which depend on the LC physical parameters; see Figure 2. Furthermore, denotes the maximum achievable phase shift range. Note that for both equations in (8), , as soon as it reaches the desired phase shift since we used over- and undershooting technique [8].
Figure 2. Transient response (rise and decay dynamics) of the LC cell for various relaxation time constants resulting from differing LC parameters, where in all curves [8, 9].
Problem Formulation and Solution
We assume that the vehicle trajectory over the prediction horizon can be predicted using existing algorithms such as [12], while the satellite trajectory is known. Accordingly, the mobility-induced channel dynamics in (5) are determined by the predicted trajectories. Moreover, the LC hardware dynamics govern the achievable DMA configuration according to (8). As a quality of service (QoS) metric, we consider the receive SNR at the satellite, which captures both dynamics and is defined as
where is the receive combining vector. Assuming , we can drop the index from and denote it as , and . We then formulate the dynamic-aware beamforming problem over sampling instants as:
where and is the sampling time. Here, (10b) is the realizable phase-shift range with being the maximum unit cell phase-shift, and (10c) enforces the maximum transmit power budget and combiner constraint. The problem (10) is a non-convex problem due to the non-convexity of the objective function. Therefore, obtaining a global solution is challenging, and we optimize and using an AO framework.
LC-DMA Phase-Shift Optimization
Assuming a given fixed and , the time reconfiguration of the DMA phase configuration problem is:
s.t.
where with , defined in (7), and
with with , defined in (4) and .
Problem (11) is non-convex due to the non-linear coupling of and time in (11a). To solve this problem efficiently in real time with linear complexity, we first decouple the objective over the individual metamaterial elements across time snapshots in the following Lemma.
Lemma 1. For sufficiently large and under the phase alignment condition , the spatial SNR metric at snapshot can be decoupled across individual elements as,
where , , and is a constant independent of .
Proof. Expanding around element while fixing all other elements yields,
where . Letting and observing that under coherent alignment, (14) simplifies to . For large arrays (), the cross-term dominates the single-element self-radiation term , which completes the proof.
Using Lemma 1, maximizing (11a) over the prediction horizon separates into parallel subproblems. For each element , the optimization task is formulated as
where is the continuous physical phase reached at time when transitioning from the initial state toward the commanded phase according to the relaxation trajectory in (8). For any instantaneous snapshot , the unconstrained maximizer of is given in closed form by . Because the physical phase trajectory cannot switch instantaneously, the commanded phase target must steer the element toward the optimal future phase profile while respecting the finite response speed of the LC molecules. Over a transition window driven primarily by electric-field excitation , the unit step response of the LC phase at step exhibits an effective factor of , . Consequently, applying a second-order Taylor track approximation around the per-step optima yields the optimal commanded phase shift in closed form:
where the dynamic-aware temporal weights are given by
The closed-form update in (16)–(17) optimizes all elements in parallel with complexity per unit cell.
Tx Digital Precoder and Rx Combiner Optimization
For a fixed phase-shift profile , Problem (10) reduces to
s.t. $\lVert\mathbf{q}\rVert_2^2\leq P_t,\quad \lVert\mathbf{w}\rVert\leq 1,
Algorithm 1 Dynamic-Aware Predictive LC-DMA Tracking
Input: Current phase state , sampling duration , horizon , , and .
Predict look-ahead array responses and construct channel states .
Compute temporal weights , via (17).
Compute the ideal instantaneous phases , .
Compute the commanded phase vector in parallel across all elements via (16).
Construct metasurface matrix using current continuous phase state and update digital precoder via (19).
Update physical LC relaxation states for subsequent time steps according to (8).
which optimizes the baseband precoder and receive combiner under the transmit power budget. The optimal combiner and precoder that maximize the received summation over SNR are given by the AO over and [13]:
where , and denotes the unit-norm eigenvector associated with the maximum eigenvalue of the matrix.
Algorithm and Complexity Analysis
The overall procedure is summarized in Algorithm 1. In each reconfiguration period, the look-ahead channel steering vectors are predicted across the horizon . Each element’s target phase is then updated in parallel via the closed-form dynamic-aware estimator in (16). Finally, the digital precoder is updated via (19). Evaluating (16) across all elements requires operations, and computing requires . Thus, the proposed framework exhibits a strictly linear computational complexity of , making it well suited for real-time tracking in high-mobility satellite uplinks.
Performance Evaluation
Simulation Setup: We consider an Earth-centered coordinate system where the origin is at the center of the Earth and the Earth’s radius is km, as illustrated in Figure 1. The ground vehicle is initially located at and travels at a speed of m/s along a local circular path of radius m. We assume a LEO satellite orbiting in the - plane with a constant orbital speed of 7.5 km/s at an altitude of km. The trajectory of the satellite is described in Cartesian coordinates as , where denotes the orbital azimuth angle. The satellite is equipped with a element UPA with inter-element spacing
Figure 1. System model for satellite-to-vehicle communication using an LC-based DMA Tx: (left) orbital geometry and communication link, (middle) waveguide-fed DMA aperture connected to RF chains, and (right) large view of a voltage-tunable () LC metasurface unit cell.
operating at a carrier frequency of GHz, where is the wavelength and is the speed of light in vacuum. The vehicle Tx employs an LC-based DMA composed of feed lines, each driving radiating elements with an element spacing of . The guided-wave parameters are configured as Np/m and for all and [14]. The Rx noise power is calculated as , where dBm/Hz is the AWGN power spectral density, MHz is the channel bandwidth, and dB represents the Rx noise figure.
Benchmark Schemes: We compare the proposed method, which accounts for both channel and LC dynamics, against two benchmarks. Benchmark 1 (neglecting channel dynamics): Uses only the instantaneous channel state (), without predicting mobility-induced channel evolution. Benchmark 2 (neglecting LC dynamics): Predicts the channel evolution over the horizon , but designs the target phase shifts under the naive assumption of instantaneous LC reconfiguration (i.e., setting uniform weights and ignoring switching dynamics by considering during optimization). During evaluation, however, the phase transitions are executed subject to the realistic physical LC time response in (8).
Simulation Results: Figure 3 illustrates the achieved SNR over time for satellite azimuth angle . The dynamic phase reconfiguration is necessary to track the channel variations. While the Benchmark 1 scheme focuses solely on maximizing the instantaneous SNR without accounting for the LC switching delay, and Benchmark 2 assumes LC phase shifts can change instantly, the proposed transition-aware algorithm optimizes phase transitions that maximize the cumulative received SNR over the tracking horizon by considering the time response constraint of the LC. Consequently, the benchmark methods fail to attain the desired SNR within the channel coherence window of the moving vehicle, whereas the proposed method maintains stable connectivity. Figure 4 illustrates the average achieved data rate when the orientation time constants of the LC material change. As increases (with [9]), the performance decreases as the tracking becomes more challenging for LC.
Figure 3. Achieved SNR versus time for satellite azimuth angles at a vehicle speed of m/s where ms and .

Figure 4. Average data rate versus reorientation time constants of the LC materials when . and m/s.
Figure 5 and Figure 6 evaluate the achievable data rate for the proposed transition-aware and the benchmark approaches. Figure 5 shows that the performance gain of the proposed scheme becomes more pronounced as the vehicle velocity increases, as it penalizes slow phase transitions. Furthermore, Figure 6 illustrates the achievable rate across various satellite positions. As the satellite moves toward lower orbital elevation/azimuth angles relative to the vehicle, the required spatial phase gradients change more rapidly, thereby magnifying the necessity and performance advantage of the proposed algorithm.
Figure 5. Achievable data rate versus vehicle speed for satellite azimuth angle .

Figure 6. Achievable data rate versus satellite position at a vehicle speed of m/s.
Conclusion
In this paper, we addressed beam tracking for high-mobility vehicle-to-LEO satellite uplinks using LC-based DMAs. We proposed a dynamic-aware predictive beamforming framework that jointly accounts for mobility-induced channel dynamics and LC hardware dynamics. The proposed solution combines predictive beam tracking with dynamic-aware LC phase reconfiguration and achieves linear computational complexity. Simulation results demonstrate more stable SNR and higher achievable rates than schemes neglecting either channel or LC dynamics, particularly in high-mobility scenarios.
Acknowledgment
The authors would like to thank Prof. Anja Klein for her valuable comments on this work.
References
- [1]Y. Su et al., “Broadband LEO satellite communications: Architectures and key technologies,” IEEE Wireless Commun., vol. 26, no. 2, pp. 55–61, 2019.
- [2]O. Kodheli et al., “Satellite communications in the new space era: A survey and future challenges,” IEEE Commun. Surveys & Tutorials, vol. 23, no. 1, pp. 70–109, 2020.arxiv.org/abs/2002.08811
- [3]L. You et al., “Massive MIMO transmission for LEO satellite communications,” IEEE J. selected areas in commun., vol. 38, no. 8, pp. 1851–1865, 2020.DOI
- [4]N. Shlezinger et al., “Dynamic metasurface antennas for uplink massive MIMO systems,” IEEE Trans. Commun., vol. 67, no. 10, pp. 6829–6843, 2019.arxiv.org/abs/1901.01458
- [5]D. R. Smith et al., “Analysis of a waveguide-fed metasurface antenna,” Physical Review Applied, vol. 8, no. 5, p. 054048, 2017.DOI
- [6]S. F. Kimaryo and K. Lee, “Downlink beamforming for dynamic metasurface antennas,” IEEE Trans. Wireless Commun., vol. 22, no. 7, pp. 4745–4755, 2023.DOI
- [7]Y. Yoo et al., “Reconfigurable multibeam holographic antenna based on liquid-crystal technology for Ka-band LEO satellite application,” IEEE Access, vol. 13, pp. 27 141–27 149, 2025.DOI
- [8]M. Delbari et al., “Fast reconfiguration of liquid crystal-RISs: Modeling and algorithm design,” IEEE Trans. Wireless Commun., vol. 25, pp. 10 098–10 114, 2026.DOI
- [9]L. Xu et al., “Fully electrically driven liquid crystal reconfigurable intelligent surface for terahertz beam steering,” IEEE Trans. Terahertz Sci. and Technol., vol. 14, no. 5, pp. 708–717, 2024.DOI
- [10]J. Park et al., “Liquid-crystal-based beam-steering guided-wave metasurface antenna at millimeter-wave band,” IEEE Antennas and Wireless Propag. Lett., vol. 23, no. 12, pp. 4208–4212, 2024.DOI
- [11]M. Delbari et al., “Fast transition-aware reconfiguration of liquid crystal-based RISs,” in IEEE ICC Workshops, 2024, pp. 214–219.
- [12]W. Yuan et al., “Bayesian predictive beamforming for vehicular networks: A low-overhead joint radar-communication approach,” IEEE Trans. Wireless Commun., vol. 20, no. 3, pp. 1442–1456, 2021.arxiv.org/abs/2005.07698
- [13]D. Tse, Fundamentals of Wireless Communication. Cambridge University Press, 2005, vol. 2.DOI
- [14]Y. Sun et al., “Dynamic metasurface antennas assisted integrated sensing and communication,” IEEE Trans. Commun., vol. 74, pp. 3074–3087, 2026.DOI



