Fast response bistable driving and low power scanning method, device and medium for liquid crystal phased array

Abstract

The application discloses a liquid crystal phased array fast-response bistable driving and low-power scanning method, equipment and medium, and relates to the technical field of liquid crystal phased array.The application compares the deviation matrix generated by real-time acquisition of liquid crystal molecule response data with the digital twin theory value, combines the deep reinforcement learning to output the electric field parameter, generates the non-uniform composite electric field through the reverse design of topological photonics, realizes the high-precision bistable steering of molecules, reconstructs the adaptive sparse scanning path based on the Bayesian compressed sensing, optimizes the energy distribution by matching the dynamic energy spectrum shaping strategy, predicts the molecule steering state through the physical information neural network embedded with physical constraints, generates the pre-control driving sequence combined with the event triggering mechanism, and outputs the optimal control instruction through the Lyapunov optimization, so that the steering response speed and stability of the liquid crystal phased array are effectively improved, and the energy utilization efficiency is greatly improved.

Description

Fast-response bistable driving and low-power scanning methods, equipment and media for liquid crystal phased arrays Technical Field

[0001] This invention relates to the field of liquid crystal phased array technology, specifically to a fast-response bistable driving and low-power scanning method, device, and medium for liquid crystal phased arrays. Background Technology

[0002] Liquid crystal phased array (LCD) technology, as a core beam control technology in modern wireless communication and radar systems, plays an irreplaceable role in key application areas such as 5G base stations, satellite communication, and phased array radar. Compared with traditional mechanically scanned antennas, LCD phased arrays can achieve electronic and rapid beam control, significantly improving the system's flexibility and reliability. The main obstacle currently facing LCD phased array technology lies in the fundamental flaws of its driving response mechanism. Existing driving schemes generally adopt a single electric field driving mode, relying solely on the longitudinal or transverse electric field acting on the liquid crystal molecules. This driving method cannot fully utilize the anisotropic characteristics of liquid crystal molecules, resulting in slow molecular turning processes and high energy consumption. More seriously, traditional scanning strategies use a fixed-step, point-by-point traversal method, performing the same scanning density and power configuration regardless of the target signal strength, resulting in a large number of invalid scans and energy waste.

[0003] The orientation response characteristics of liquid crystal molecules under an electric field directly determine the performance of the entire phased array system. Due to the significant anisotropy of liquid crystal molecules, their response speed and stability vary significantly under electric fields in different directions. A single-directional electric field cannot achieve rapid and stable orientation of the molecules. This physical limitation further leads to a deep contradiction between driving power consumption and response speed: achieving a fast response requires applying a high-intensity electric field, but maintaining a high electric field generates enormous continuous power consumption. For example, in satellite communication applications, when the system needs to complete a large-angle beam switch from east to west within milliseconds, traditional driving methods often require a response time of more than 10 milliseconds, while maintaining a high power consumption state throughout the communication process, placing enormous pressure on the satellite payload's power system. Summary of the Invention

[0004] The purpose of this invention is to provide a fast-response bistable driving and low-power scanning method, device and medium for liquid crystal phased arrays, so as to realize fast and stable turning of liquid crystal phased arrays, high energy utilization efficiency and fast beam switching.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] This application provides a fast-response bistable driving and low-power scanning method for liquid crystal phased arrays, including the following steps:

[0007] S1. Real-time acquisition of the dielectric tensor dynamic response spectrum and micro pointing angle distribution of liquid crystal molecules under the action of multi-dimensional electric field, comparison of the distribution with the theoretical pointing angle calculated by the digital twin constructed based on the liquid crystal continuum theory, and generation of real-time pointing deviation matrix of the entire array.

[0008] S2. The pointing deviation matrix is ​​fused with the channel state information of real-time communication. When the deviation value exceeds the first dynamic threshold, the pre-trained deep reinforcement learning decision-maker is activated and the multi-target electric field parameter combination is output.

[0009] S3. The multi-target electric field parameters are combined and applied to the liquid crystal phased array to generate a non-uniform composite electric field pattern designed by topological photonics inverse design, thereby obtaining a high-precision bistable steering response sequence.

[0010] The non-uniform composite electric field pattern actively compensates for the phase error caused by the boundary anchoring energy and material anisotropy, driving the liquid crystal molecules to complete the flipping and locking of the topological insulating state.

[0011] S4. Extract the spatial and frequency domain scanning sparsity features from the bistable turn response sequence, and use the Bayesian compressed sensing algorithm to reconstruct an adaptive sparse scanning path that only covers the key area based on the terminal distribution and signal strength.

[0012] S5. Based on the adaptive sparse scanning path, execute the dynamic energy spectrum shaping strategy to form a globally optimized energy distribution scheme;

[0013] The dynamic energy spectrum shaping strategy is used to identify weak signal regions and idle blocks in the path, dynamically reduce their scanning power and refresh rate, and redistribute the saved energy resources to beams with high communication load according to priority.

[0014] S6. Using historical driving parameters, ambient temperature and molecular response data, train a physical information neural network to obtain a highly reliable predictive steering model.

[0015] The loss function of the physical information neural network is embedded with the liquid crystal dynamics equation as a physical constraint, which is used to predict the probability distribution of molecular orientation state within a specific time window in the future.

[0016] S7. Extract the future beam switching demand sequence in the satellite communication scenario from the predicted turning model, integrate high-precision timestamps and task priorities, and generate a pre-control driving sequence through an event triggering mechanism;

[0017] S8. Based on the pre-control drive sequence, configure the system parameters of the phased array in real time, and use an online control algorithm based on the Lyapunov optimization framework to dynamically solve for the final control instruction set that satisfies multi-objective Pareto optimality.

[0018] A fast-response bistable driving and low-power scanning device for a liquid crystal phased array includes a processor, a memory, and computer program instructions stored in the memory. When the computer program instructions are executed by the processor, the fast-response bistable driving and low-power scanning method for the liquid crystal phased array described above is implemented.

[0019] A storage medium storing computer program instructions, which, when executed by a processor, implement the fast-response bistable driving and low-power scanning method for a liquid crystal phased array as described above.

[0020] The beneficial effects of this invention are as follows:

[0021] To address the problem that a single electric field driving method cannot fully utilize the anisotropy of liquid crystal molecules, resulting in slow molecular turning and significant phase errors, a real-time pointing deviation matrix is ​​generated by comparing digital twins. This matrix is ​​then combined with the output of a deep reinforcement learning decision-maker to create a multi-target electric field parameter combination. Finally, a non-uniform composite electric field is generated through topological photonics inverse design. This actively compensates for phase errors and drives liquid crystal molecules to achieve the flipping and locking of a topological insulating state, ultimately achieving a high-precision bistable turning response. This not only breaks through the limitations of the single driving mode but also improves the speed and stability of molecular turning.

[0022] To address the problem of traditional fixed-step point-by-point scanning using the same configuration regardless of signal strength, resulting in a large number of invalid scans and energy waste, this paper extracts scanning sparsity features, uses a Bayesian compressed sensing algorithm to reconstruct an adaptive sparse scanning path, and combines it with a dynamic energy spectrum shaping strategy to identify weak signal regions and idle blocks and dynamically adjust scanning power and refresh rate. This redistributes the saved energy to high-load beams, effectively avoiding invalid scans and significantly improving the system's energy utilization efficiency.

[0023] To address the deep-seated contradiction between driving power consumption and response speed, as well as the problem of beam switching response lag, a physical information neural network constrained by liquid crystal dynamics equations is used to predict molecular orientation states. Combined with satellite communication mission characteristics and event triggering mechanisms, a pre-controlled driving sequence is generated. Then, a multi-objective Pareto optimal control command is solved using a Lyapunov optimization framework to pre-adapt to future beam switching requirements, compensate for system nonlinearity and prediction errors, achieve rapid and stable beam switching, reduce steady-state maintenance power consumption, and resolve the conflict between response speed and power consumption. Attached Figure Description

[0024] To better understand and implement this application, the technical solution is described in detail below with reference to the accompanying drawings.

[0025] Figure 1 is a flowchart illustrating the fast-response bistable driving and low-power scanning method for a liquid crystal phased array provided in Embodiment 1 of this application.

[0026] Figure 2 is a flowchart of step S3 in the fast-response bistable driving and low-power scanning method for liquid crystal phased array provided in Embodiment 1 of this application.

[0027] Figure 3 is a flowchart of step S6 in the fast-response bistable driving and low-power scanning method for liquid crystal phased array provided in Embodiment 1 of this application. Detailed Implementation

[0028] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, exemplary embodiments will be described in detail below, examples of which are illustrated in the accompanying drawings. In the following description, when referring to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0029] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0030] The following detailed description of the specific implementation methods, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided in detail.

[0031] Example 1

[0032] Please refer to Figures 1-3. This embodiment provides a fast-response bistable driving and low-power scanning method for liquid crystal phased arrays, including the following steps:

[0033] S1. By integrating a micro-capacitor and photosensitive sensor unit into the array substrate, the dielectric tensor dynamic response spectrum and micro-pointing angle distribution of liquid crystal molecules under the action of a multi-dimensional electric field are collected in real time; the distribution is compared with the theoretical pointing angle calculated by the digital twin constructed based on the liquid crystal continuum theory to generate a real-time pointing deviation matrix of the entire array surface.

[0034] Further, step S1 specifically includes:

[0035] By using a miniature capacitor and a photosensitive sensor unit, the optical response signal of liquid crystal molecules under the action of electric field vectors in the X, Y, and Z axes is collected in real time to obtain the original response data containing spatiotemporal information. Based on the change in light intensity, the actual microscopic pointing angle distribution of the liquid crystal molecules in the whole field is calculated using a physical optical model.

[0036] The actual microscopic pointing angle distribution is input into a digital twin constructed based on liquid crystal continuum theory to obtain the corresponding theoretical pointing angle distribution; the measured distribution is compared with the theoretical distribution point by point to identify the regions that deviate from the theoretical value by more than a preset tolerance, and an abnormal region map is generated.

[0037] Machine learning algorithms are used to cluster the points in the abnormal area map according to their spatial distribution, electric field direction and deviation, identify the abnormal response pattern type, and calculate the difference between the micro pointing angle of each abnormal point and the theoretical value based on the identified pattern category, and quantify and generate the initial deviation field.

[0038] Spatiotemporal statistical analysis is performed on the initial deviation field to evaluate the persistence and severity of deviations in each electric field direction, and the final real-time pointing deviation matrix of the entire array is constructed. This matrix also contains dynamic stability evaluation information of the liquid crystal material under the current driving conditions.

[0039] Specifically, the actual microscopic pointing angle distribution of the liquid crystal molecules across the entire field is calculated using a physical optics model. This includes: constructing a Jones matrix transfer function with the pointing angle of the liquid crystal molecules as the variable based on the transmitted light intensity signal collected by the photosensitive sensor and the change in dielectric constant measured by the microcapacitor sensor; solving the inverse problem corresponding to this transfer function, substituting the measured optical parameters into the rigorously calibrated pointing angle-optical response mapping relationship, and using a nonlinear least squares optimization algorithm to iteratively invert and calculate the precise three-dimensional pointing angle of each pixel, ultimately reconstructing the full-field microscopic pointing angle distribution that reflects the actual spatial orientation of the liquid crystal molecules.

[0040] The actual microscopic pointing angle distribution is input into a digital twin constructed based on liquid crystal continuum theory. Specifically, this involves: establishing the dynamic equation of liquid crystal molecules under specific boundary conditions by solving the free energy functional that includes elastic deformation energy, electric field energy, and surface anchoring energy; taking the currently applied multidimensional electric field parameters, material property parameters, and boundary constraints as inputs, numerically solving the dynamic equation using the finite element method, and calculating the equilibrium orientation that liquid crystal molecules should achieve under ideal conditions, thereby obtaining the theoretical pointing angle distribution that completely corresponds to the measured conditions.

[0041] Machine learning algorithms are used to perform intelligent pattern recognition on points in the anomaly region map. Specifically, this includes: extracting the spatial coordinates, electric field vector direction, and angular deviation of each anomaly point to form a multi-dimensional feature vector; performing unsupervised learning on these feature vectors using a clustering algorithm based on a Gaussian mixture model, and automatically classifying anomaly pattern categories with different statistical characteristics based on feature similarity; simultaneously calculating the feature vector norm of the center point of each category to quantify the severity of different anomaly patterns, and finally establishing a mapping relationship between anomaly patterns and physical mechanisms to provide a classification basis for subsequent deviation compensation.

[0042] Spatiotemporal statistical analysis is performed on the initial deviation field, specifically including: calculating the time-series statistical characteristics of the deviation values ​​along the three electric field directions X, Y, and Z, including the mean, variance, and autocorrelation coefficient within the sliding window; extracting the spatial distribution characteristics of the deviation field in each direction using principal component analysis to identify clusters of correlated anomalous regions; and establishing a stability assessment method based on a hidden Markov model by combining time series characteristics and spatial distribution patterns to predict the evolution trend of the deviation state in each region, ultimately generating an enhanced deviation matrix containing spatiotemporal correlation characteristics.

[0043] Specifically, by constructing a monitoring system that deeply integrates multi-dimensional sensing and digital twins, the problem of inaccurate drive control caused by the lack of precise state sensing in traditional liquid crystal phased arrays has been solved. It has achieved a comprehensive state assessment from microscopic molecules to macroscopic array performance, laying a reliable foundation for subsequent precise control and effectively improving the system's control accuracy and response consistency.

[0044] S2. The pointing deviation matrix is ​​fused with the channel state information of real-time communication. When the deviation value exceeds the first dynamic threshold, the pre-trained deep reinforcement learning decision-maker is started. The deep reinforcement learning decision-maker takes the deviation matrix and the channel state as input and outputs a combination of multi-objective electric field parameters that simultaneously optimizes the response speed and steady-state retention rate, including electric field strength, direction and duty cycle of the driving waveform.

[0045] Further, step S2 specifically includes:

[0046] The real-time pointing deviation matrix of the entire array is spatiotemporally aligned and fused with the channel state information of the communication system. The channel state information includes at least the received signal strength indicator, the channel quality indicator, and the bit error rate. Based on the fused data, the comprehensive performance index of the current system is calculated. When the index or the average pointing deviation value of the key area exceeds the first dynamic threshold that is dynamically adjusted according to the communication service quality requirements, the pre-trained deep reinforcement learning decision-maker is immediately triggered to enter the working state.

[0047] The fused multidimensional feature vector is input into a decision-maker pre-trained using a deep deterministic policy gradient algorithm. The reward function of the decision-maker is specially designed to consider the weighted sum of pointing deviation, response speed, steady-state error, and communication interruption probability. Through the forward inference of the network, the optimal combination of multi-objective electric field parameters is output, including the electric field strength, direction vector, and duty cycle of the driving waveform of each phased array element.

[0048] Physical constraints are applied to verify the parameter combination output by the deep reinforcement learning decision maker. The constraints include: the electric field strength must not exceed the breakdown threshold of the liquid crystal material; the change of the direction vector must meet the requirements of continuity and smoothness to avoid mechanical stress; and the duty cycle must be kept within the effective driving range to ensure the complete response of the liquid crystal molecules. If any of the output parameters violates any constraint, a correction algorithm based on the Lagrange multiplier method is initiated to adjust it, ensuring its physical realizability and system security.

[0049] The verified parameter combination is encapsulated into a drive command frame according to the communication protocol of the phased array driver. The command frame contains the electric field intensity setting value, the three-dimensional coordinate components of the direction vector and the duty cycle parameter of each unit, and is sent to each drive unit of the phased array through a high-speed bus.

[0050] The calculation of comprehensive performance indicators specifically includes:

[0051] The average deviation of the region is calculated based on the deviation value of each pixel in the pointing deviation matrix. The real-time performance index is obtained by weighted summation, combined with the signal-to-noise ratio and bit error rate in the channel state information. The first dynamic threshold is dynamically adjusted according to the minimum signal-to-noise ratio and the maximum allowable bit error rate in the communication service quality requirements.

[0052] The training process of the deep reinforcement learning decision-maker specifically includes: establishing a reinforcement learning model with a state space, action space, and reward function as its core. The state space includes the eigenvalues ​​of the pointing bias matrix and channel state information, the action space is a combination of electric field parameters, and the reward function includes a pointing bias penalty term, a response speed reward term, and a communication quality reward term; using a deep deterministic policy gradient algorithm, offline training is performed through an actor-critic network structure to finally obtain a policy network that can output the optimal combination of electric field parameters.

[0053] The physical constraint verification specifically includes: setting the upper limit of the electric field strength to 80-90% of the breakdown threshold of the liquid crystal material, ensuring that the rate of change of the direction vector does not exceed the maximum allowable rate of change, and limiting the duty cycle to an effective range of 10%-90%; when the parameters exceed the constraint range, a constraint optimization algorithm is used to adjust them to the most recent effective value while maintaining the relative relationship between the parameters.

[0054] The aforementioned drive instruction encapsulation specifically includes: converting electric field parameters into binary data frames according to the phased array control protocol, adding frame headers, checksums, and timestamp information, and transmitting them to each drive unit via gigabit Ethernet or a high-speed serial bus to ensure that the instructions take effect synchronously within a specified time.

[0055] Each driving unit is the core terminal for performing electro-optic conversion in the liquid crystal phased array system. Each unit contains an independent electrode control circuit, a signal decoding module, and a power amplifier device. By analyzing the electric field intensity, direction vector, and duty cycle parameters in the driving command frame, it generates a driving signal with specific amplitude, phase, and timing, which precisely controls the molecular arrangement state of the corresponding liquid crystal unit, and ultimately achieves coordinated control of the beam shape and direction of the entire phased array.

[0056] Specifically, by constructing an intelligent driving mechanism that links perception and decision-making, the problem of response lag and low energy efficiency caused by the disconnect between driving strategy and real-time status in traditional liquid crystal phased arrays is solved. Dynamic matching between system status and driving parameters is achieved, which significantly improves response speed and energy utilization efficiency while ensuring beam control accuracy.

[0057] S3. The multi-target electric field parameters are combined and applied to a liquid crystal phased array to generate a non-uniform composite electric field pattern that has been reverse-designed by topological photonics.

[0058] The non-uniform composite electric field pattern actively compensates for the phase error caused by the boundary anchoring energy and material anisotropy, driving the liquid crystal molecules to complete the flipping and locking of the topological insulating state, thereby obtaining a high-precision bistable turning response sequence.

[0059] Further, step S3 specifically includes:

[0060] S31. Input the combination of multi-target electric field parameters into the topological photonics reverse design system, solve the continuum topology optimization problem based on the adjoint method, generate a non-uniform electrode pattern that can generate the target wavefront phase, and according to the optimization results, convert the combination of multi-target electric field parameters into a spatially modulated voltage signal through a multi-channel digital micromirror device or a programmable electrode array, and generate a non-uniform composite electric field corresponding to the non-uniform electrode pattern in the liquid crystal layer.

[0061] S32. The spatial distribution of the non-uniform composite electric field, which has been reverse-engineered by topological photonics, is pre-calculated so that the equiphase surface of the non-uniform composite electric field and the distribution of the inherent phase error caused by the non-uniformity of the boundary anchoring energy and the anisotropy of the material form a conjugate relationship in space; through this conjugate distribution relationship, the non-uniform composite electric field actively compensates for the inherent phase error in the spatial dimension.

[0062] S33. While applying the non-uniform composite electric field, the dynamic response of the liquid crystal molecules is monitored in real time using an integrated optical sensor. The orientation relaxation time, final pointing angle, and steady-state retention are collected to form a high-precision bistable orientation response sequence. The accuracy and stability of the orientation are verified by comparing the response sequence with the ideal bistable response calculated based on Landau-deGennes theory, ensuring that it meets the system's requirements for fast response and long-term steady-state retention.

[0063] The topological photonics reverse design system specifically includes: taking the target phase distribution as the optimization objective and the photoelectric properties and manufacturing process constraints of the liquid crystal material as the boundary conditions, constructing an optimization problem with the topological structure of the electrode pattern as the design variable; using gradient-based optimization algorithms (such as the moving asymptote method) to iteratively solve the problem, and finally obtaining a physically feasible and robust non-uniform electrode design scheme.

[0064] Locking under topological protection is specifically manifested as follows: the final orientation of liquid crystal molecules is located at a local minimum point in its free energy landscape, and this minimum point is surrounded by a sufficiently high energy barrier; the existence of the non-uniform recombination electric field is equivalent to modifying the free energy landscape, making the energy trap of the target steady state deeper and the energy barrier higher, thereby endowing the steady state with topological robustness, making it significantly suppressive of external perturbations. Here, the Landau-de-Gennes theory is used to describe the equilibrium properties and phase transition behavior of liquid crystals, which is suitable for the establishment and locking analysis of bistable states.

[0065] Active compensation for inherent phase errors is used to drive liquid crystal molecules to overcome the energy barrier during their bistable switching process, prompting the liquid crystal molecule group to undergo rapid cooperative flipping motion from the initial steady state to the target steady state. After the flipping is completed, the non-uniform composite electric field causes the liquid crystal molecules to enter the deep energy potential well formed by the modified system free energy landscape in the target orientation, realizing the directional locking of the topological insulating state. This locking state can effectively suppress the random molecular orientation caused by environmental thermal fluctuations or external disturbances, ensuring the long-term maintenance and angular accuracy of the bistable orientation sequence.

[0066] The quantitative evaluation method for steady-state retention includes: continuously monitoring the change in pointing angle within a specific time window after the driving electric field is removed; if the standard deviation is less than a preset threshold, it is considered that effective steady-state locking has been achieved.

[0067] Specifically, by constructing a non-uniform electric field driving mechanism based on topological photonics, the problem of insufficient beam pointing accuracy caused by boundary effects and material anisotropy in traditional liquid crystal phased arrays is solved. Active compensation for inherent phase error and stable locking of molecular orientation are achieved, which improves beamforming accuracy while ensuring long-term stability of the orientation state.

[0068] S4. Extract the spatial and frequency domain scan sparsity features from the bistable turn response sequence, and use the Bayesian compressed sensing algorithm to reconstruct an adaptive sparse scan path that only covers the key area based on the terminal distribution and signal strength.

[0069] The adaptive sparse scan path mainly addresses the issue of spatial coverage efficiency in which to scan, and its output provides a spatial framework for the dynamic energy allocation of S5.

[0070] Further, step S4 specifically includes:

[0071] The original data of the bistable turning response sequence is obtained, and the response sequence is segmented through a preset time window to obtain response data segments of multiple time periods. For each response data segment, its spatial sampling point density and frequency energy distribution are calculated, and spatial density features and frequency energy features characterizing the scan sparsity are extracted.

[0072] Based on the spatial density characteristics and frequency energy characteristics, the signal strength evaluation module calculates the comprehensive signal strength value of each region; the comprehensive signal strength value is compared with a preset strength threshold, and regions exceeding the threshold are marked as key signal regions, thus forming a signal strength distribution map.

[0073] Using the signal intensity distribution map as prior information, a scanning path optimization model based on Bayesian compressed sensing is constructed. The model aims to maximize the coverage of key signal areas and, under the condition of scanning resource constraints, obtains an adaptive sparse scanning path through iterative optimization. The path ensures that scanning energy consumption and time overhead are significantly reduced while meeting scanning performance requirements.

[0074] The method for calculating the spatial sampling point density includes: statistically analyzing the spatial distribution density of valid sampling points in each response data segment, and calculating the number of sampling points per unit area as the spatial density feature value.

[0075] The signal strength assessment module works by weighting and fusing the spatial density characteristic value and frequency domain energy characteristic value of each region. The weighting coefficients are dynamically adjusted according to communication quality requirements to obtain a comprehensive signal strength value that reflects the importance of the region.

[0076] The Bayesian compressed sensing model is constructed by using the signal intensity distribution as the prior probability distribution, the sparsity of the scanning path as a constraint, and reconstructing the optimal scanning path through maximum a posteriori probability estimation; the sparsity of the scanning path is achieved through L1 norm regularization.

[0077] Specifically, by constructing an intelligent path planning mechanism based on Bayesian compressed sensing, the problem of resource waste caused by uniform coverage of the entire area in traditional scanning strategies is solved, and accurate identification and efficient coverage of key signal areas are achieved, significantly improving scanning efficiency and resource utilization while ensuring communication quality.

[0078] S5. Execute the dynamic energy spectrum shaping strategy according to the adaptive sparse scanning path;

[0079] The dynamic energy spectrum shaping strategy, based on the spatial path provided by S4, addresses the energy efficiency issue of scanning at a certain power level. It is used to identify weak signal regions and idle blocks in the path, dynamically reduce their scanning power and refresh rate, and redistribute the saved energy resources to beams with high communication load according to priority, forming a globally optimized energy distribution scheme.

[0080] Further, step S5 specifically includes:

[0081] The signal power values ​​of each sampling point on the adaptive sparse scanning path are obtained, a signal intensity distribution matrix is ​​generated, and the scanning path is divided into strong signal area, weak signal area and idle area using a double threshold segmentation method; based on historical power consumption data, the initial power allocation benchmark value of each type of area is determined by cluster analysis.

[0082] The system monitors signal strength changes in each area in real time. When the area ratio of weak signal area or idle area exceeds the preset threshold, a dynamic power adjustment mechanism is activated. Based on the area type and communication demand priority, the scanning power and refresh rate of weak signal area and idle area are dynamically reduced, while the saved power resources are allocated to strong signal area according to the weight.

[0083] A constrained optimization algorithm is used to optimize the power allocation scheme globally, maximizing the overall energy utilization efficiency of the system while meeting the minimum communication quality requirements. The optimized power allocation parameters are then used to generate a standardized configuration file and deployed to the power control module of the scanning system.

[0084] The specific implementation of the dual-threshold segmentation method includes: setting two signal power thresholds, high and low; dividing the region with signal power higher than the high threshold into a strong signal region, the region with signal power lower than the low threshold into an idle region, and the region between the two thresholds into a weak signal region; the value of the threshold is dynamically adjusted according to the quality requirements of the communication system.

[0085] The dynamic power adjustment mechanism works by: establishing a correspondence between the power adjustment coefficient and the area ratio of the region; when the area ratio of the weak signal region or the idle region exceeds a preset threshold, dynamically adjusting the power allocation coefficient of these regions according to a pre-set functional relationship; the functional relationship ensures that the power is reasonably reduced while ensuring the basic scanning function.

[0086] The constrained optimization algorithm is constructed by using the goal of maximizing the overall energy efficiency of the system and the minimum power demand of each region as constraints, and employing linear programming to solve for the optimal power allocation scheme; the minimum power demand is determined based on the communication quality requirements and service priorities of each region.

[0087] As the core execution unit of the dynamic energy spectrum shaping strategy, the power control module establishes a dynamic mapping relationship between region type and power configuration by analyzing the signal strength distribution matrix and the preset power allocation benchmark value in real time. Based on the dual threshold segmentation results, it automatically identifies weak signal areas and idle blocks. When the area ratio of the above regions exceeds the set threshold, the power adjustment mechanism is immediately triggered. The scanning power and refresh rate are dynamically reduced according to the communication priority weight. At the same time, the surplus energy is redistributed to the strong signal area according to the optimization algorithm. Through the integrated constraint optimization solver, the energy efficiency configuration is continuously optimized under the premise of ensuring the bottom line of communication quality. The final parameter set is transformed into executable device control instructions, realizing a complete closed-loop management from energy strategy to hardware operation.

[0088] Specifically, by establishing a dynamic energy spectrum shaping mechanism, the problem of low resource utilization caused by the solidification of energy allocation in traditional scanning systems is solved, and power adaptive adjustment based on real-time signal characteristics is realized, which significantly improves the overall energy efficiency of the system while maintaining communication quality.

[0089] S6. Using historical driving parameters, ambient temperature, and molecular response data, a physical information neural network is trained. The loss function of the physical information neural network is embedded in the liquid crystal dynamics equation as a physical constraint to predict the probability distribution of molecular turning states within a specific time window in the future, thereby obtaining a highly reliable predictive turning model.

[0090] Further, step S6 specifically includes:

[0091] S61. Collect historical driving parameter sequences, environmental temperature time-series data and corresponding molecular response data to form the original training dataset; perform time alignment, outlier removal and standardization on the original training dataset to form a standard sample set for network training.

[0092] S62. Construct a physical information neural network containing prior knowledge of liquid crystal dynamics. The network takes driving parameters and ambient temperature as input and molecular orientation state as output. Embed physical constraint terms based on the Leslie-Erickson equation into the network loss function. By minimizing the weighted sum of prediction error and physical constraint residual, the end-to-end training of the network is completed.

[0093] S63. Utilize the trained network to probabilistically predict the molecular turning states within a specific future time window, and output the turning state distribution including confidence intervals; perform dual verification of the prediction results using offline test sets and online real-time data to ensure the reliability and accuracy of the prediction model in practical applications.

[0094] The data preprocessing method includes: segmenting the original data using a sliding time window, and calculating statistical characteristics for the data within each time window; the statistical characteristics include mean, variance, maximum value, and minimum value, which are used to characterize the system state within that time period.

[0095] The specific structure of the physical information neural network includes: using a long short-term memory network to extract temporal features, and a fully connected layer to realize feature mapping; the physical constraint term calculates the derivative of the network output with respect to time through automatic differentiation technology, so that it satisfies the evolution law described by the liquid crystal dynamics equation.

[0096] Model validation methods include: calculating the root mean square error between the predicted results and the measured data, and checking whether the predicted results meet the physical constraints; when the root mean square error is lower than the threshold and the physical constraint residuals meet the requirements, the model is deemed to have met the deployment standards.

[0097] The Leslie-Erickson equation is the governing equation of continuum theory describing the dynamic behavior of nematic liquid crystals. It characterizes the spatiotemporal evolution of the orientation field of liquid crystal molecules under the coupling of external field and flow through the constitutive relation of the viscous stress tensor and the velocity gradient. As a dynamic constraint in a physical information neural network, this equation is mainly used to predict the transient response and relaxation process of molecules during the driving process, complementing the Landau-de Genennes theory used in S3 for static energy landscape analysis.

[0098] Specifically, by constructing an intelligent prediction model that integrates physical laws, the problem of insufficient prediction accuracy caused by the lack of physical priors in traditional methods is solved. This enables accurate prediction of molecular orientation dynamics, provides the system with forward-looking decision-making capabilities, and significantly improves the accuracy and reliability of control.

[0099] S7. Extract the future beam switching demand sequence in the satellite communication scenario from the predicted switching model, integrate high-precision timestamps and task priorities, and generate a pre-control driving sequence only when the predicted switching angle exceeds the preset tolerance through an event triggering mechanism, so as to avoid invalid switching operations.

[0100] Further, step S7 specifically includes:

[0101] The beam pointing angle prediction sequence within a specific future time window is extracted from the predicted steering model; the prediction sequence is time-aligned with the satellite orbit ephemeris and communication mission timeline, and the switching requests are sorted according to the mission priority to generate an initial switching request queue with timestamps and priorities.

[0102] Establish a trigger judgment logic based on the predicted switching angle change, calculate the difference in predicted pointing angle between adjacent time steps, and mark the moment as a valid switching event only when the difference exceeds the tolerance band preset according to the system stability and communication quality requirements; perform conflict detection and elimination on the marked events to ensure that only the highest priority switching task is executed at the same time;

[0103] For each valid switching event, the driving waveform parameters are calculated in reverse based on the predicted target angle, system response characteristics, and link delay to generate a pre-control driving sequence containing amplitude, phase, and a precise switching timestamp. An online learning strategy based on historical performance data is used to dynamically optimize the driving parameters to compensate for model prediction errors and system nonlinearity. The switching timestamp is synchronized with the start time of the Lyapunov optimized control time slot in S8.

[0104] The timing alignment method includes synchronizing the time base of the predicted steering model with the timing signal of the satellite navigation system, and using a linear interpolation method to resample the predicted sequence to the system control cycle to ensure strict synchronization between the demand sequence and the system clock.

[0105] The tolerance band setting method includes: the lower limit of the tolerance band is determined by the system angular resolution, and the upper limit is determined by the communication link interruption risk threshold; the tolerance band is dynamically adjusted according to the signal-to-noise ratio, and is appropriately narrowed under low signal-to-noise ratio conditions to improve tracking accuracy.

[0106] The drive sequence optimization method includes: establishing a mapping relationship between drive parameters and switching accuracy and response speed, and searching for the optimal combination of drive parameters that satisfies the switching time constraint and minimizes energy consumption through gradient descent.

[0107] Calculate the difference in predicted pointing angles between adjacent time steps, including the predicted pointing angle at the previous time step t-1. Based on the current time t, the predicted pointing angle Perform the difference operation to obtain the instantaneous angle change. By comparing this change with a bilateral tolerance band preset according to system stability margin and communication link maintenance requirements. Real-time comparison is performed only when Only when a valid physical layer beam switching requirement is detected will the subsequent pre-control drive sequence generation process be triggered; among which, It represents the predicted pointing angle of the liquid crystal phased array beam at time t-1, and indicates the spatial pointing state of the beam at the previous time. The predicted pointing angle of the liquid crystal phased array beam at time t represents the spatial pointing state of the beam at the current time. It represents the change in beam pointing angle between adjacent time steps, indicating the instantaneous change in beam pointing. The minimum tolerance value for the change in beam pointing angle is determined by the system angular resolution and is a preset fixed threshold. The maximum tolerance value for the change in beam pointing angle is determined by the communication link interruption risk threshold, which is a preset fixed threshold. This indicates that the set does not belong to the symbol, signifying that the actual angle change exceeds the tolerance band range, triggering an effective beam switching event.

[0108] Specifically, by establishing an event-triggered intelligent switching mechanism, the problem of resource waste caused by frequent invalid operations in traditional beam switching systems has been solved. It enables accurate judgment of necessary switching times and proactive generation of pre-control commands, significantly improving the effective utilization rate of system resources while ensuring communication continuity.

[0109] S8. Based on the pre-controlled drive sequence, configure the system parameters of the phased array in real time, and adopt an online control algorithm based on the Lyapunov optimization framework to unify the instantaneous power consumption, beam switching delay and queue stability of the system in a stochastic optimization problem, and dynamically solve for the final control instruction set that satisfies the multi-objective Pareto optimality.

[0110] The execution cycle of step S8 is strictly aligned with the timestamp in the pre-control drive sequence output by S7. At the beginning of each control time slot, all valid pre-control drive instructions within that time slot are loaded first and used as the expected state input for the Lyapunov optimization problem, thereby ensuring seamless integration between the forward control instructions and the real-time system optimization.

[0111] Further, step S8 specifically includes:

[0112] The driving status, instantaneous power consumption, and task queue status of each element of the phased array are collected in real time, and a virtual queue system including power consumption queue, delay queue, and stability index is constructed. Based on the Lyapunov optimization framework, the multi-objective optimization problem is transformed into a single-step optimization problem of minimizing the Lyapunov drift plus penalty function.

[0113] In each control time slot, the single-step optimization problem is solved to dynamically determine the optimal driving parameters for each array element, including amplitude weighting coefficients, phase offset, and switching state. The optimization process simultaneously considers instantaneous power consumption constraints, switching delay upper limits, and system stability requirements to ensure optimal long-term performance indicators.

[0114] The optimized driving parameters are converted into a set of control instructions executable by the phased array hardware. The correctness and real-time performance of the instruction set are verified by hardware-in-the-loop simulation to ensure that it meets the requirements of beamforming accuracy and response speed.

[0115] The virtual queue system construction method includes: establishing a power consumption queue, a delay queue, and a stability queue to characterize the cumulative deviation of the system in terms of power consumption, delay, and stability, respectively; and transforming the multi-queue control problem into a single-objective optimization problem through a weighted summation method.

[0116] Methods for solving single-step optimization problems include: using stochastic gradient descent to find the optimal driving parameters, and improving system adaptability while ensuring algorithm convergence by adjusting the balance strategy between exploration and utilization.

[0117] The control command verification method includes: simulating the command execution effect through a digital twin system, comparing the similarity between the expected beam pattern and the actual beam pattern, and triggering a command re-optimization mechanism when the similarity is lower than a threshold.

[0118] The multi-objective optimization problem is transformed into a single-step optimization problem of minimizing Lyapunov drift plus a penalty function. Specifically, this involves: first, constructing a virtual queue vector Θ(t) containing a system power consumption queue, a time delay queue, and a stability queue, and calculating its Lyapunov function L(Θ(t)); then defining the Lyapunov drift ΔL(Θ(t)) = E{L(Θ(t+1)) - L(Θ(t))|Θ(t)}, and adding it to the penalty function V·E{C(t)|Θ(t)}, which characterizes the instantaneous operating cost of the system, to form the optimization objective function; finally, by minimizing the upper bound of this objective function, the long-term stochastic optimization problem is decomposed into a single-step optimization problem of solving the optimal control decision within each time slot, where the weight parameter V is used to adjust the balance between stability and efficiency in the optimization objective; where Θ(t): the system virtual queue vector at time t, a bold multi-dimensional column vector containing the state parameters of the power consumption queue, the time delay queue, and the stability queue;

[0119] L(Θ(t)): A Lyapunov function constructed based on the virtual queue vector Θ(t), a scalar that characterizes the stability state of the system;

[0120] ΔL(Θ(t)): Lyapunov drift, a scalar that represents the change in the Lyapunov function from time t to time t+1;

[0121] E{⋅}: Mathematical expectation operator, used to calculate the statistical expectation of a random variable;

[0122] V: Weight parameter, scalar, positive real number, used to adjust the balance between system stability and operational efficiency in the optimization objective;

[0123] C(t): The instantaneous operating cost function of the system at time t, a scalar that represents the combined operating cost of the system's power consumption and delay at that time.

[0124] Specifically, by constructing a system-level control architecture based on Lyapunov optimization, the complex decision-making problem of multi-objective control in traditional phased array systems, which is difficult to coordinate, was solved. This achieved a dynamic balance between system power consumption, latency, and stability, and significantly improved the long-term operating efficiency of the system while ensuring performance indicators.

[0125] A fast-response bistable driving and low-power scanning device for a liquid crystal phased array includes a processor, a memory, and computer program instructions stored in the memory. When the computer program instructions are executed by the processor, the fast-response bistable driving and low-power scanning method for the liquid crystal phased array described above is implemented.

[0126] A storage medium storing computer program instructions, which, when executed by a processor, implement the fast-response bistable driving and low-power scanning method for a liquid crystal phased array as described above.

[0127] Example 2

[0128] This embodiment provides another solution for achieving fast-response bistable driving and low-power scanning of liquid crystal phased arrays. It employs acoustic microfluidics technology to replace traditional electrode contact driving, precisely controlling molecular alignment through a non-contact method and avoiding electrode wear. A ferroelectric liquid crystal composite material system is introduced, utilizing its spontaneous polarization characteristics to improve the response speed to the microsecond level. Finally, a self-powered system is integrated, achieving system energy self-sufficiency through environmental energy harvesting. While maintaining fast response, this significantly improves equipment lifespan and environmental adaptability, making it particularly suitable for long-term stable operation in special scenarios such as outdoor base stations and mobile platforms. Specific details include the following:

[0129] Surface acoustic wave microfluidic molecular orientation control integrates a multilayer interdigital transducer array on the surface of an array substrate. A programmable surface acoustic wave field is generated by excitation with a radio frequency signal. This acoustic wave field acts directly on the liquid crystal molecules through acoustic radiation force. At the same time, the micro-eddy current generated by the acoustic flow effect is used to achieve precise control of molecular arrangement. By independently controlling the frequency (10-100MHz range) and phase (0-360° adjustable) of each transducer unit, a molecular arrangement pattern with a specific spatial distribution is formed in the liquid crystal layer.

[0130] The construction and driving optimization of the ferroelectric liquid crystal composite system adopts a composite material system of ferroelectric liquid crystal and nematic liquid crystal. The spontaneous polarization characteristics of ferroelectric liquid crystal under the action of electric field (response time <10μs) are used to achieve rapid switching of molecular orientation. At the same time, the continuous orientation characteristics of nematic liquid crystal are used to maintain the continuous scanning capability of the beam within ±60°. By optimizing the mixing ratio and arrangement parameters of the two liquid crystals, good optical uniformity is ensured while maintaining the fast response characteristics.

[0131] Photoacoustic coupling multi-physics field synergistic control involves arranging a pulsed laser array (wavelength 1064nm, pulse width 10ns) around the array to form a synergistic control system with the surface acoustic wave field. The laser pulse generates a transient thermal gradient field (temperature rise <2℃) in the liquid crystal layer through the photoacoustic effect. This thermal field couples with the acoustic field to produce an enhanced molecular driving effect. This multi-physics field synergistic effect can reduce the driving voltage requirement from >10V in traditional methods to <3V, while maintaining a sub-millisecond response speed.

[0132] Microencapsulation enhances bistable properties

[0133] Polymer microcapsules with diameters of 5-20 μm (wall thickness of 0.5-2 μm) were prepared using interfacial polymerization to encapsulate liquid crystal materials. The interfacial anchoring effect of the capsule wall provides additional orientation constraints for the liquid crystal molecules, significantly enhancing their bistable properties. By precisely controlling the capsule size distribution and wall thickness parameters, the retention time of the molecular orientation state was extended from several hours in traditional methods to tens of days, achieving true zero-power state maintenance.

[0134] The biomimetic retinal intelligent scanning strategy draws inspiration from the non-uniform sampling mechanism of vertebrate retinas, designing a zone-differentiated scanning scheme. A full sampling mode (60Hz refresh rate) is used within ±10° of the beam center, while event-triggered sparse sampling (refresh rate ≤5Hz) is employed in the edge regions (±10°-±60°). This strategy maintains scanning accuracy in the core region while reducing average system power consumption and increasing the probability of capturing burst signals.

[0135] The memristor array adaptive compensation system integrates a 32×32 memristor cross array (resistance range 10kΩ-1MΩ) in the control circuit and optimizes the driving parameters in real time through simulation calculations. Utilizing the conductance memory characteristics of the memristors, a dynamic model of the system's nonlinear characteristics is established to achieve automatic compensation for factors such as temperature fluctuations and material aging. This system can maintain beam pointing accuracy within ±0.1° over a long period.

[0136] The self-powered energy management system arranges a triboelectric nanogenerator array (output power density ≥ 0.5 mW / cm²) on the inner surface of the equipment casing. 2It achieves self-sufficiency in energy by collecting mechanical vibrations and environmental noise during equipment operation. Equipped with an intelligent energy management chip, it dynamically adjusts the system's operating mode (full power / energy saving / standby) according to the energy collection status, ensuring continuous operation for ≥72 hours without external power supply.

[0137] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A fast-response bistable driving and low-power scanning method for a liquid crystal phased array, characterized in that: The process includes the following steps: S1. Real-time acquisition of the dielectric tensor dynamic response spectrum and microscopic pointing angle distribution of liquid crystal molecules under the action of a multi-dimensional electric field. The distribution is compared with the theoretical pointing angle calculated by a digital twin constructed based on liquid crystal continuum theory to generate a real-time pointing deviation matrix for the entire array. S2. The pointing deviation matrix is ​​fused with the channel state information of real-time communication. When the deviation value exceeds the first dynamic threshold, a pre-trained deep reinforcement learning decision-maker is activated to output a combination of multi-objective electric field parameters. S3. The combination of multi-objective electric field parameters is applied to the liquid crystal phased array to generate a non-uniform composite electric field pattern designed by topological photonics inverse engineering, obtaining a high-precision bistable turning response sequence. The non-uniform composite electric field pattern actively compensates for the phase error caused by boundary anchoring energy and material anisotropy, driving the liquid crystal molecules to complete the flipping and locking of a topological insulating state. S4. From the bistable turning response sequence, the scanning sparsity features in the spatial and frequency domains are extracted. Using a Bayesian compressed sensing algorithm, an adaptive sparsity pattern covering only key areas is reconstructed based on terminal distribution and signal strength. S5. Based on the adaptive sparse scanning path, execute a dynamic energy spectrum shaping strategy to form a globally optimized energy distribution scheme. The dynamic energy spectrum shaping strategy is used to identify weak signal regions and idle blocks in the path, dynamically reduce their scanning power and refresh rate, and redistribute the saved energy resources to beams with high communication load according to priority. S6. Using historical driving parameters, ambient temperature, and molecular response data, train a physical information neural network to obtain a highly reliable predictive steering model. The loss function of the physical information neural network is embedded with the liquid crystal dynamics equation as a physical constraint to predict the probability distribution of molecular steering states within a specific future time window. S7. Extract the future beam switching demand sequence under the satellite communication scenario from the predictive steering model, integrate high-precision timestamps and task priorities, and generate a pre-control driving sequence through an event triggering mechanism. S8. Based on the pre-control driving sequence, configure the system parameters of the phased array in real time, and use an online control algorithm based on the Lyapunov optimization framework to dynamically solve for the final control instruction set that satisfies multi-objective Pareto optimality.

2. The fast-response bistable driving and low-power scanning method for a liquid crystal phased array according to claim 1, characterized in that: Step S1 specifically includes: acquiring the optical response signal of liquid crystal molecules under the action of a triaxial electric field vector in real time through a micro-capacitor and photosensitive sensor unit to obtain raw response data containing spatiotemporal information; calculating the actual microscopic pointing angle distribution of the liquid crystal molecules across the entire field using a physical optics model based on the change in light intensity; inputting the actual microscopic pointing angle distribution into a digital twin constructed based on the liquid crystal continuum theory to obtain the corresponding theoretical pointing angle distribution; comparing the measured distribution with the theoretical distribution point by point to identify regions that deviate from the theoretical value by more than a preset tolerance, generating an abnormal region map; using a machine learning algorithm to cluster the points in the abnormal region map according to spatial distribution, electric field direction, and deviation amount to identify the pattern type of abnormal response; calculating the difference between the microscopic pointing angle of each abnormal point and the theoretical value based on the identified pattern category, quantifying and generating an initial deviation field; performing spatiotemporal statistical analysis on the initial deviation field to evaluate the persistence and severity of deviation in each electric field direction, and constructing the final real-time pointing deviation matrix of the entire array.

3. The fast-response bistable driving and low-power scanning method for a liquid crystal phased array according to claim 1, characterized in that: Step S2 specifically includes: spatiotemporally aligning and fusing the real-time pointing deviation matrix of the entire array with the channel state information of the communication system; calculating the comprehensive performance index of the current system based on the fused data; when the index or the average pointing deviation value of the key area exceeds the first dynamic threshold, immediately triggering the pre-trained deep reinforcement learning decision-maker to enter the working state; inputting the fused multidimensional feature vector into the decision-maker pre-trained using the deep deterministic policy gradient algorithm, and outputting the comprehensive optimal combination of multi-objective electric field parameters through the forward inference of the network, including the electric field strength, direction vector, and duty cycle of the driving waveform of each phased array unit; applying physical constraints to verify the parameter combination output by the deep reinforcement learning decision-maker; if the output parameters violate any constraint, starting the correction algorithm based on the Lagrange multiplier method for adjustment; encapsulating the verified parameter combination into a driving command frame according to the communication protocol of the phased array driver, the command frame containing the electric field strength setting value, the three-dimensional coordinate components of the direction vector, and the duty cycle parameter of each unit, and sending it to each driving unit of the phased array through the high-speed bus.

4. The fast-response bistable driving and low-power scanning method for a liquid crystal phased array according to claim 1, characterized in that: Step S3 specifically includes: S31, inputting the combination of multi-target electric field parameters into the topological photonics reverse design system, generating a non-uniform electrode pattern that produces the target wavefront phase by solving a continuum topology optimization problem based on the adjoint method, and converting the combination of multi-target electric field parameters into a spatially modulated voltage signal through a multi-channel digital micromirror device or a programmable electrode array to generate a non-uniform composite electric field corresponding to the non-uniform electrode pattern in the liquid crystal layer; S32, the spatial distribution of the non-uniform composite electric field after topological photonics reverse design is pre-calculated so that the equiphase surface of the non-uniform composite electric field and the inherent phase error distribution caused by the non-uniformity of the boundary anchoring energy and the anisotropy of the material form a conjugate relationship in space; S33, while applying the non-uniform composite electric field, using an integrated optical sensor to monitor the dynamic response of the liquid crystal molecules in real time, collecting their turning relaxation time, final pointing angle and steady-state retention, forming a high-precision bistable turning response sequence.

5. The fast-response bistable driving and low-power scanning method for a liquid crystal phased array according to claim 1, characterized in that: Step S4 specifically includes: acquiring the original data of the bistable turning response sequence; segmenting the response sequence through a preset time window to obtain response data segments for multiple time periods; calculating the spatial sampling point density and frequency energy distribution of each response data segment; extracting spatial density features and frequency energy features characterizing scan sparsity; calculating the comprehensive signal strength value of each region based on the spatial density features and frequency energy features through a signal strength evaluation module; comparing the comprehensive signal strength value with a preset strength threshold, marking regions exceeding the threshold as key signal regions, and forming a signal strength distribution map; using the signal strength distribution map as prior information to construct a scan path optimization model based on Bayesian compressed sensing; the model aims to maximize the coverage of key signal regions, takes scan resource constraints as conditions, and obtains an adaptive sparse scan path through iterative optimization.

6. The fast-response bistable driving and low-power scanning method for a liquid crystal phased array according to claim 1, characterized in that: Step S5 specifically includes: acquiring the signal power values ​​of each sampling point on the adaptive sparse scanning path, generating a signal intensity distribution matrix, and dividing the scanning path into strong signal areas, weak signal areas, and idle areas using a dual-threshold segmentation method; determining the initial power allocation benchmark value for each type of area based on historical power consumption data through cluster analysis; monitoring the signal intensity changes of each area in real time, and activating a dynamic power adjustment mechanism when the area ratio of a weak signal area or idle area exceeds a preset threshold; dynamically reducing the scanning power and refresh rate of weak signal areas and idle areas according to the area type and communication demand priority, while allocating the saved power resources to strong signal areas according to weights; using a constrained optimization algorithm to perform global energy efficiency optimization on the power allocation scheme, maximizing the overall energy utilization efficiency of the system under the constraint of meeting the minimum communication quality requirements; generating a standardized configuration file for the optimized power allocation parameters and deploying it to the power control module of the scanning system.

7. The fast-response bistable driving and low-power scanning method for a liquid crystal phased array according to claim 1, characterized in that: Step S6 specifically includes: S61, collecting historical driving parameter sequences, ambient temperature time-series data, and corresponding molecular response data to form an original training dataset; performing time alignment, outlier removal, and standardization on the original training dataset to form a standard sample set for network training; S62, constructing a physical information neural network containing liquid crystal dynamics priors, wherein the network takes driving parameters and ambient temperature as inputs and molecular orientation states as outputs; embedding physical constraint terms based on the Leslie-Erickson equation into the network loss function, and completing end-to-end training of the network by minimizing the weighted sum of prediction error and physical constraint residuals; S63, using the trained network to predict the molecular orientation states within a specific future time window probabilistically, outputting an orientation state distribution containing confidence intervals; and performing dual verification of the prediction results using offline test sets and online real-time data.

8. The fast-response bistable driving and low-power scanning method for a liquid crystal phased array according to claim 1, characterized in that: Step S7 specifically includes: extracting the beam pointing angle prediction sequence within a specific future time window from the predicted steering model; aligning the prediction sequence with the satellite orbit ephemeris and communication mission timeline, sorting the handover requests based on mission priority, and generating an initial handover request queue with timestamps and priorities; establishing a trigger judgment logic based on changes in the predicted handover angle, calculating the difference in predicted pointing angle between adjacent time steps, and marking a valid handover event when the difference exceeds a tolerance band preset according to system stability and communication quality requirements, and performing conflict detection and elimination on the marked events; for each valid handover event, calculating the driving waveform parameters in reverse based on the predicted target angle, system response characteristics, and link delay, and generating a pre-controlled driving sequence containing amplitude, phase, and precise handover timestamps; and using an online learning strategy based on historical performance data to dynamically optimize the driving parameters to compensate for model prediction errors and system nonlinearity.

9. A fast-response bistable driving and low-power scanning device for a liquid crystal phased array, characterized in that: The system includes a processor, a memory, and computer program instructions stored in the memory. When the computer program instructions are executed by the processor, they implement the fast-response bistable driving and low-power scanning method for the liquid crystal phased array as described in any one of claims 1-8.

10. A storage medium storing computer program instructions thereon, characterized in that, When the computer program instructions are executed by the processor, the fast-response bistable driving and low-power scanning method for the liquid crystal phased array as described in any one of claims 1-8 is implemented.

Images