Terahertz link pre-pointing control method based on link quality residual correction

Abstract

This invention provides a terahertz link pre-pointing control method based on link quality residual correction, relating to the technical field of terahertz communication. The method includes acquiring the ultra-narrow beam half-width, actuator feedforward compensation, historical deviation compensation, current actual beam pointing, measured link quality, and prior future LOS prediction angle; determining the predicted pointing error and predicted link quality; smoothing the link quality residual to obtain a smoothed link quality residual; determining the link quality angle gradient; determining the link residual correction angle; determining the link reliability based on the link quality angle gradient to obtain the final future LOS prediction angle; determining the future lock-out probability and deciding whether to trigger pre-pointing control; and generating a beam pre-pointing command when pre-pointing control is triggered. This application alleviates the technical problems of low pre-pointing accuracy and low link stability in ultra-narrow beam terahertz links under dynamic moving scenarios in the prior art.

Description

Technical Field

[0001] This invention relates to the field of terahertz communication technology, and in particular to a terahertz link pre-pointing control method based on link quality residual correction. Background Technology

[0002] Terahertz communication, with its abundant spectrum resources, extremely high transmission rates, and strong anti-interference capabilities, is considered one of the key enabling technologies for future ultra-high-speed wireless communication and space-based interconnection networks. However, terahertz signals suffer severe attenuation during atmospheric propagation and are extremely sensitive to directional deviations, typically requiring ultra-narrow beams for directional transmission to meet link budget and communication quality requirements.

[0003] In dynamic mobile scenarios, communication nodes may be mounted on high-speed moving platforms such as drones, vehicles, or satellites. The position, attitude, and relative line-of-sight (LOS) direction of the nodes are constantly changing. Since there are non-negligible time delays in sensor sampling, signal processing, control calculation, command transmission, and actuator response, if beam control only tracks the LOS direction at the current moment, the platform will have moved to a new position when the control command actually takes effect. The beam pointing will lag behind the true LOS direction, resulting in significant pointing errors and even link loss. Therefore, it is necessary to predict the LOS direction at several future moments to achieve pre-pointing control.

[0004] Existing methods generally treat link quality only as a post-event evaluation indicator of communication performance, without incorporating it into the closed-loop feedback loop of pre-pointing control. This lack of an effective correction mechanism leads to the continuous accumulation of deviations, ultimately resulting in alignment degradation.

[0005] However, the existing technology has the following problems: low pre-pointing accuracy and low link stability in dynamic moving scenarios. Summary of the Invention

[0006] In view of this, the purpose of this invention is to provide a terahertz link pre-pointing control method based on link quality residual correction, so as to alleviate the technical problems of low pre-pointing accuracy and link maintenance stability of ultra-narrow beam terahertz links in dynamic moving scenarios.

[0007] In a first aspect, embodiments of this application provide a terahertz link pre-pointing control method based on link quality residual correction, the method comprising: Acquire the ultra-narrow beam half-width, actuator feedforward compensation, historical deviation compensation, current actual beam pointing, measured link quality, and prior future LOS prediction angle; The prediction pointing error is determined based on the prior future LOS prediction angle and the current actual beam pointing, and the prediction link quality is determined based on the prediction pointing error. Based on the measured link quality and the predicted link quality, the link quality residual is determined, and the link quality residual is smoothed to obtain the smoothed link quality residual; A symmetrical perturbation is applied to the prior future LOS prediction angle to obtain a positive perturbation angle and a negative perturbation angle. The link quality angle gradient is determined based on the measured link quality corresponding to the positive perturbation angle and the negative perturbation angle. The link residual correction angle is determined based on the smoothed link quality residual and the link quality angle gradient. The link confidence level is determined based on the link quality angle gradient. The link residual correction angle is then weighted and superimposed onto the prior future LOS prediction angle based on the link confidence level to obtain the final future LOS prediction angle. The probability of future loss of lock is determined based on the final future LOS prediction angle, the current actual beam pointing, and the ultra-narrow beam half-width. Whether to trigger pre-pointing control is determined based on the future loss of lock probability, the predicted link quality, the final future LOS prediction angle, and the current actual beam pointing. When pre-pointing control is triggered, a beam pre-pointing command is generated based on the final future LOS prediction angle, the actuator feedforward compensation amount, and the historical deviation compensation amount.

[0008] In one implementation of the first aspect, the process of determining the prediction pointing error based on the prior future LOS prediction angle and the current actual beam pointing, and determining the prediction link quality based on the prediction pointing error, includes: The prediction pointing error is calculated based on the current actual beam pointing and the prior future LOS prediction angle. Represented as: , in, The current actual beam direction, For the prior future LOS prediction angle; The predicted link quality is determined based on the predicted pointing error. Represented as: , in, This serves as the reference link quality when the beam is aligned or in the reference direction. This is the descent coefficient obtained by fitting the curve near the main lobe of the antenna pattern.

[0009] In one implementation of the first aspect, the process of determining the link quality residual based on the measured link quality and the predicted link quality, and smoothing the link quality residual to obtain a smoothed link quality residual, includes: The link quality residual is obtained by subtracting the measured link quality from the predicted link quality. : , in, This is to measure the link quality. The smoothed link quality residual is obtained by performing exponentially weighted moving smoothing on the link quality residual. : , in, The residual smoothing coefficient is... The sampling period.

[0010] In one implementation of the first aspect, the process of applying a symmetrical perturbation to the prior future LOS prediction angle to obtain a positive perturbation angle and a negative perturbation angle, and determining the link quality angle gradient based on the measured link quality corresponding to the positive perturbation angle and the negative perturbation angle, includes: The candidate pre-pointing angle is obtained by initializing the prior future LOS prediction angle. Symmetrical perturbations are applied to both sides of the candidate pre-pointing angle to obtain a positive perturbation angle and a negative perturbation angle. The positive perturbation angle... and the negative perturbation angle Represented as: , , in, Candidate pre-pointing angle, The amplitude of the two-sided perturbation; The link quality angular gradient is determined by estimating the measured link quality corresponding to the positive and negative perturbation angles using the central difference method. Represented as: , in, and These are the measured link qualities corresponding to the positive and negative perturbation angles, respectively.

[0011] In one implementation of the first aspect, the process of determining the link residual correction angle based on the smoothed link quality residual and the link quality angle gradient includes: Based on the smoothed link quality residual and the link quality angle gradient, a link residual correction angle is generated. Represented as: , in, For residual correction gain, For the maximum allowable correction angle, For symbolic functions, surface The correction is only initiated when the measured link quality is lower than the predicted link quality.

[0012] In one implementation of the first aspect, the process of determining the link confidence level based on the link quality angle gradient, and then weighting and superimposing the link residual correction angle onto the prior future LOS prediction angle according to the link confidence level to obtain the final future LOS prediction angle includes: Link reliability is calculated based on the clarity of the link quality angle gradient. Represented as: , in, It is the gradient stability constant; The link correction values ​​are weighted according to confidence level and then superimposed onto the prior future LOS prediction angle to obtain the final future LOS prediction angle. Represented as: , in, This is to pre-point the target direction that is actually used for control.

[0013] In one implementation of the first aspect, the process of determining the future loss-of-lock probability based on the final future LOS prediction angle, the current actual beam pointing, and the ultra-narrow beam half-width includes: Based on the final future LOS prediction angle and the current actual beam pointing, the control command will take effect at the specified time. The future pointing error is considered a random variable. Represented as; , in, Let be the mean of the random variable pointing to future errors. The variance of the future-pointing random variable for error; The mean of the future pointing error random variable is expressed as: , in, This is for estimating the future beam direction when pre-pointing is not triggered or when the current control strategy continues to be executed; The variance of the future pointing error random variable is expressed as: , in, For the variance of future beam execution error, The variance of the future prediction error; , in, It is the standard normal distribution function. It is an ultra-narrow beam half-width.

[0014] In one implementation of the first aspect, the process of determining whether to trigger pre-pointing control based on the future loss-of-lock probability, the predicted link quality, the final future LOS prediction angle, and the current actual beam pointing includes: like ,but , or ,but , or , in, For pre-pointing the trigger flag, This is the threshold for the probability of losing the lock. This is the link quality threshold. This is the error margin coefficient.

[0015] In one implementation of the first aspect, the process of generating a beam pre-pointing command based on the final future LOS prediction angle, the actuator feedforward compensation amount, and the historical deviation compensation amount includes: When the pre-pointing trigger flag is 1, a beam pre-pointing command is generated based on the final future LOS prediction angle, the actuator feedforward compensation, and the historical deviation compensation. Represented as: , in, For the feedforward compensation amount of the implementing agency, This is the amount to compensate for historical deviations.

[0016] In one implementation of the first aspect, the actuator is driven to respond and the current actual beam pointing is updated according to the beam pre-pointing command.

[0017] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating the terahertz link pre-pointing control method based on link quality residual correction provided in an embodiment of the present invention; Figure 2 This is a geometric scene diagram of a two-dimensional mobile terahertz link provided in an embodiment of the present invention; Figure 3 The simulation results of pointing error comparison are shown in the embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 As shown, it is a flowchart illustrating the terahertz link pre-pointing control method based on link quality residual correction provided in an embodiment of the present invention.

[0023] To facilitate understanding of this embodiment, a detailed description of a terahertz link pre-pointing control method based on link quality residual correction disclosed in this embodiment of the invention will be provided first, including: Step S1: Obtain the ultra-narrow beam half-width, actuator feedforward compensation, historical deviation compensation, current actual beam pointing, measured link quality, and prior future LOS prediction angle. Step S2: Determine the prediction pointing error based on the prior future LOS prediction angle and the current actual beam pointing, and determine the prediction link quality based on the prediction pointing error; Step S3: Determine the link quality residual based on the measured link quality and the predicted link quality, and smooth the link quality residual to obtain a smoothed link quality residual; Step S4: Apply a symmetrical perturbation to the prior future LOS prediction angle to obtain a positive perturbation angle and a negative perturbation angle. Determine the link quality angle gradient based on the measured link quality corresponding to the positive perturbation angle and the negative perturbation angle. Step S5: Determine the link residual correction angle based on the smoothed link quality residual and the link quality angle gradient; Step S6: Determine the link confidence level based on the link quality angle gradient, and weight and superimpose the link residual correction angle onto the prior future LOS prediction angle according to the link confidence level to obtain the final future LOS prediction angle. Step S7: Determine the future loss of lock probability based on the final future LOS prediction angle, the current actual beam pointing, and the ultra-narrow beam half-width; determine whether to trigger pre-pointing control based on the future loss of lock probability, the predicted link quality, the final future LOS prediction angle, and the current actual beam pointing. Step S8: When it is determined that pre-pointing control is triggered, a beam pre-pointing command is generated based on the final future LOS prediction angle, the actuator feedforward compensation amount, and the historical deviation compensation amount.

[0024] Specifically, this invention calculates the prediction error and link quality by using a priori prediction angle and the current beam pointing, and then compares it with the measured link quality to obtain a smoothed residual. At the same time, it uses perturbation detection to obtain the gradient of link quality with respect to angle changes, and corrects the prediction angle accordingly and evaluates the link reliability for weighted fusion. Finally, it assesses the risk of loss of lock based on the reliable prediction angle and decides whether to trigger beam pre-pointing control. It fuses the open-loop geometric prediction with the closed-loop link quality feedback, and uses real-time measurement residuals to continuously calibrate the prediction angle. This overcomes the alignment degradation caused by the accumulation of errors in the pure prediction model in dynamic environments. The link reliability based on gradient representation enables adaptive adjustment of the correction weight. When the link quality is sensitive to angle, the reliability is high and the correction is strong. When entering a flat area or being blocked, the reliability is low and the correction is moderately suppressed to prevent noise from misleading beam deflection. This improves the pre-pointing accuracy and link holding stability of the ultra-narrow beam terahertz link in dynamic moving scenarios.

[0025] Specifically, the embodiments of the present invention include: an AoA angle measurement module, a platform motion state acquisition module, a beam execution feedback module, a terahertz link quality observation module, a future LOS prediction module, a link quality residual correction module, a two-sided perturbation gradient estimation module, a future loss-of-lock probability calculation module, and a beam pre-pointing control module.

[0026] In this embodiment of the invention, the AoA angle measurement module is used to obtain the current angle of arrival and its historical window; the platform motion state acquisition module is used to obtain the platform position, velocity, and acceleration; the beam execution feedback module is used to read the current actual beam pointing; the terahertz link quality observation module is used to obtain RSSI, SINR, or other link quality indicators; the future LOS prediction module is used to generate kinematic prediction angles and AoA sequence prediction angles; the link quality residual correction module is used to calculate the residual between the measured link quality and the predicted link quality; the two-sided perturbation gradient estimation module is used to determine the angle correction direction corresponding to the link quality change; the future loss-of-lock probability calculation module is used to determine whether advance control is needed; and the beam pre-pointing control module is used to output gimbal angle, phased array codebook, or array weight control commands.

[0027] The functional modules of this invention can be integrated into the beam pre-pointing controller, or implemented by a processor, FPGA, DSP, embedded industrial computer, baseband processing unit or a combination thereof. The beam pre-pointing controller is not additional, independently defined hardware, but a control computing unit used to perform prediction, correction, triggering and command generation processes. Its input terminals are respectively connected to angle measurement, integrated navigation, link monitoring and execution feedback units, and its output terminals are connected to a mechanical gimbal, a phased array beamforming network or a mechanical-electronic hybrid actuator.

[0028] Specifically, in this embodiment of the invention, the beam pre-pointing controller reads the system clock, sensor timestamp, communication link status record, actuator feedback record, antenna pattern calibration data, and link threshold calibration data to obtain the sampling period. Sensor sampling delay Signal processing delay Controlling computation delay Command transmission delay Response delay of implementing agencies Ultra-narrow beamwidth and bilateral perturbation amplitude .

[0029] In this embodiment of the invention, the beam pre-pointing controller can be an embedded processor, an FPGA (Field Programmable Gate Array), a DSP (Digital Signal Processor), a baseband processor, or a host computer control program. The beam pre-pointing controller is used to perform prediction, residual correction, loss-of-lock probability judgment, and command output. The beam pre-pointing controller sums the above delays to obtain the total execution delay.

[0030] Specifically, in this embodiment of the invention, the total execution delay is determined based on the sensor sampling delay, signal processing delay, control calculation delay, command transmission delay, and actuator response delay. The total execution delay is expressed as follows: , in, For sensor sampling delay, For signal processing delay, To control computational latency, For command transmission delay, This is to account for the response delay of the implementing agency.

[0031] Specifically, in this embodiment of the invention, given the total execution delay, the discrete prediction step size is determined based on the total execution delay and the sampling period. The discrete prediction step size is expressed as follows: , in, For discrete prediction step size, (·) is the floor function.

[0032] Specifically, the bilateral perturbation amplitudes in embodiments of the present invention satisfy the following constraints: , in, It is an ultra-narrow beam half-width.

[0033] In this embodiment of the invention, the sampling period can be obtained by controlling the cyclic timer or the difference between continuous sampling timestamps; the sensor sampling delay can be obtained by the difference between the sensor sampling trigger time and the data valid output time; the signal processing delay can be obtained by the baseband or processor input / output timestamps; the control calculation delay can be obtained by the start and end timestamps of the control algorithm; the command transmission delay can be obtained by the timestamps of command sending and execution mechanism receiving confirmation or by bus / link testing; the execution mechanism response delay can be obtained by the cloud step jump response, servo angle arrival time, phased array weight loading time or codebook switching confirmation time; the ultra-narrow beamwidth can be obtained by calibrating the half-power beamwidth of the terahertz antenna pattern, the allowable pointing error, or the link loss threshold; the bilateral perturbation amplitude is based on 0 < < The selection is based on the principle that the link quality degradation during the perturbation period should not exceed the allowable margin.

[0034] Understandably, by unifying the delays of various sensing, computing, transmission, and execution processes into the same prediction time domain, the control commands generated at the current moment will be... The prediction takes effect at any time, changing the LOS (Line-of-Sight) of the current target to the LOS at the future effective time, reducing pointing errors caused by control lag.

[0035] Specifically, this embodiment of the invention obtains the current angle of arrival through an AoA (Angle of Arrival) angle measurement module. The AoA angle measurement module can be a phased array angle of arrival estimation module, a terahertz angle measurement unit, a visual beacon, a photoelectric tracker, or other angle measurement sensors.

[0036] Specifically, the current angle of arrival in this embodiment of the invention It can be represented as: , in, For the true LOS direction, For random angle measurement noise, This is a deviation in a slowly varying system.

[0037] In this embodiment of the invention, It can be obtained from geometric position, simulation true value, or high-confidence angle measurement results. and It is obtained from angular residual statistics, calibration tests, or online deviation estimation.

[0038] Specifically, in this embodiment of the invention, the current arrival angle is written to a length of... The historical window is used for angular velocity and angular acceleration estimation. The historical window is represented as follows: .

[0039] In this embodiment of the invention, the history window length The number of samples should be greater than the order of the motion model to be fitted; for example, for a constant angular acceleration model, at least [number of samples required]. Increase when there is a lot of noise. To improve the smoothing effect, reduce [the impact of] strong maneuvers or rapid changes in attitude disturbances. To reduce lag; alternatively, offline simulation or online cross-validation can be used to select the method that minimizes the prediction residual. .

[0040] Understandably, converting instantaneous AoA observations into time series that can estimate angular velocity, angular acceleration, and deviation trends reduces the impact of single-point angle measurement noise on prediction.

[0041] Specifically, embodiments of the present invention obtain the platform position based on GNSS (Global Navigation Satellite System), IMU (Inertial Measurement Unit), visual odometry, flight control status, vehicle control status, encoder, or integrated navigation module. ,speed acceleration Optionally, the attitude angle can be obtained. and angular velocity Ω .

[0042] In this embodiment of the invention, if the positions of communication node A and communication node B are respectively and The relative position and geometric LOS direction are represented as follows: , , , in, It is a two-dimensional relative position vector. and In the global coordinate system and The directional component is represented as , This represents the link distance.

[0043] In this embodiment of the invention, communication node A is a mobile platform equipped with an AoA angle measurement module, and communication node B is a signal source to be aligned.

[0044] In this embodiment of the invention, in a two-dimensional scene, the geometric LOS direction is determined by... Function calculation.

[0045] In this embodiment of the invention, for the three-dimensional implementation, the relative position and link distance are expressed as follows: , , in, for The directional component.

[0046] In this embodiment of the invention, the unit LOS vector of the three-dimensional implementation Represented as: .

[0047] In this embodiment of the invention, the three-dimensional LOS can be represented as azimuth and elevation angles, with the azimuth angle expressed as: , The pitch angle is expressed as: , The three-dimensional LOS is represented as: .

[0048] In this embodiment of the invention, it relates to The formula can be executed separately for azimuth and pitch angles in three-dimensional implementation, or it can be fused, residual corrected and triggered in the form of two-dimensional angle vectors.

[0049] Understandably, by obtaining the link geometry and the relative motion states of the nodes, inputs are provided for future LOS kinematic predictions, and both two-dimensional and three-dimensional link scenarios have publicly available, executable computational processes.

[0050] Specifically, embodiments of the present invention obtain the current actual beam pointing through a mechanical gimbal encoder, servo motor angle feedback, phased array beamcode book feedback, phase weight feedback, or beam controller state variables. .

[0051] Specifically, in this embodiment of the invention, for mechanical pan-tilt units, the encoder angle can be directly read, and the encoder angle is the current actual beam pointing. , is represented as: , in, Indicates the encoder at time... Output mechanical angle.

[0052] Specifically, in this embodiment of the invention, for phased arrays, the current beamcodebook index can be mapped to an equivalent beam angle to obtain the current actual beam pointing. , is represented as: , in, Indicates the current codebook index of the phased array The corresponding equivalent main lobe angle.

[0053] It is understandable that the actual output of the actuator, not just the control command, should be obtained to avoid the prediction model mistaking the command angle as the current actual beam direction due to actuator lag, mechanical backlash, codebook switching delay, or array weight loading error.

[0054] Specifically, embodiments of the present invention acquire the received strength in real time through a terahertz receiver, a baseband demodulator, a link monitoring unit, or an internal statistical unit of the communication module. Signal-to-interference-to-noise ratio It can also be extended to signal-to-noise ratio. Modulation error ratio Bit error rate Packet error rate Or link margin.

[0055] Specifically, in this embodiment of the invention, the received strength and signal-to-interference-plus-noise ratio are normalized into link quality components, and the measured link quality is... Represented as: , in, , , in, and These are the weighting coefficients, and they satisfy... , , , , , , The normalized boundary values ​​can be determined by the receiver dynamic range, link budget, laboratory calibration, field statistical quantiles, or protocol thresholds. Indicates will Limited to between 0 and 1.

[0056] In this embodiment of the invention, the principle for determining the weighting coefficient is as follows: if the receiver... More stable, therefore increase If the interference changes significantly and It better reflects link availability, thus increasing If only one indicator is reliable, then set the weight of that indicator to 1 and the weight of the other indicator to 0.

[0057] In this embodiment of the invention, the calibration phase can also target the link packet error rate, retransmission rate, or lockout events, selecting the factor that minimizes the prediction error. and .

[0058] It is understandable that different units of measurement are used. The link quality is mapped to a unified dimension, allowing it to be incorporated into subsequent residual correction and judgment processes, rather than being used merely as a post-evaluation indicator.

[0059] Specifically, this invention obtains the true current actual beam pointing, avoiding mistaking the control command angle for the actual angle. Simultaneously, it collects multiple link indicators such as received signal strength and signal-to-interference-plus-noise ratio (SINR) in real time. After normalization and boundary limiting, these indicators are weighted and fused according to configurable weights to output the measured link quality. The encoder readback or codebook mapping obtains the true physical output of the actuator, avoiding interference from mechanical hysteresis, codebook switching delay, and weight loading errors on beam pointing judgment. The fusion of multi-dimensional link quality characteristics, including received signal strength and SINR, ensures robust link evaluation in weak signal and high interference scenarios, thereby improving the pre-pointing accuracy and link stability of ultra-narrow beam terahertz links in dynamic moving scenarios.

[0060] Specifically, the embodiments of the present invention will place the platform location ,speed and acceleration By incorporating the short-term acceleration kinematic extrapolation model, the effective time of the control command can be predicted. The platform position of this model belongs to the kinematic prediction branch in the prediction module. The model architecture includes a state input layer, a time delay extrapolation layer, a relative position calculation layer, and an angle output layer.

[0061] Specifically, embodiments of the present invention can be implemented in a short time. Assuming the acceleration remains approximately constant, the future platform position can be represented as: .

[0062] Specifically, in the two-dimensional case, the kinematic future LOS prediction angle of this embodiment of the invention is expressed as: , in, , These are two-dimensional coordinate components.

[0063] In this embodiment of the invention, if both communication node A and communication node B are mobile nodes, then relative motion is adopted, with relative speed... and relative acceleration They are represented as follows: , , The relative position at future moments is then represented as: .

[0064] Specifically, in the two-dimensional case, the kinematic future LOS prediction angle of this embodiment of the invention is expressed as: , Specifically, in a three-dimensional context, the embodiments of the present invention calculate the azimuth angle of the future LOS. and pitch angle .

[0065] Understandably, by using the platform's motion state to compensate for changes in LOS during execution latency, even if the current AoA has not yet shifted significantly, it is possible to generate an estimate of the future direction in advance based on the motion trend.

[0066] Specifically, embodiments of the present invention are based on the AoA history window. Estimate current angular velocity and angular acceleration .

[0067] Specifically, in this embodiment of the invention, before calculating angular velocity and angular acceleration, unwrap (phase unwrapping) or wrapToPi (angle normalization) processing can be performed on the angle sequence to avoid angle jumps. Mutations can occur at any time.

[0068] Specifically, the embodiments of the present invention provide two implementation methods: difference estimation and quadratic least squares fitting. The difference estimation implementation method is as follows: , , in, The current angular velocity, This represents the current angular acceleration.

[0069] Specifically, in another noise reduction implementation of this invention, a quadratic least squares fitting is performed within the historical window: , After fitting, take: , , in, , , The coefficients are for fitting the polynomial.

[0070] In this embodiment of the invention, Kalman filtering or Savitzky-Golay filtering can also be used to obtain the same angular velocity and angular acceleration estimates.

[0071] Specifically, after obtaining the current angular velocity and current angular acceleration, this embodiment of the invention uses a constant angular acceleration model to predict the future AoA within a short prediction window: , , in, Let AoA be the angular motion state vector. The future prediction angle for the AoA sequence.

[0072] Understandably, extracting angular motion trends from historical angular measurement sequences allows for the prediction of future LOS using AoA time series even when the platform location model is incomplete or there are installation deviations.

[0073] Specifically, this invention generates a priori future LOS prediction angles jointly from two paths. Kinematic extrapolation uses the platform's motion state for feedforward prediction, which can compensate for line-of-sight changes caused by high-speed platform movement or maneuvering, and make up for the lag caused by time delay in AoA tracking. Dynamic angle extrapolation extracts angular motion trends from historical AoA sequences, which can capture pointing changes that cannot be directly obtained through position kinematics, such as platform rotation and attitude changes. It complements the kinematic branch, enabling the priori prediction angles to maintain high accuracy in complex multi-dimensional maneuvering scenarios, thereby improving the pre-pointing accuracy and link stability of ultra-narrow beam terahertz links in dynamic moving scenarios.

[0074] Specifically, embodiments of the present invention are based on the kinematic prediction error variance AoA prediction error variance Calculate the fusion weights of the two prediction branches.

[0075] In this embodiment of the invention, the variance of kinematic prediction error and the variance of AoA prediction error can be estimated by rolling back the prediction residuals: at time... , will the previous The predicted value of time to the current time and the current reference angle By comparison, the prediction residuals of the two branches are obtained as follows: , , in, An angle normalization function is used to limit the angle difference to a certain value. Within the range, Indicates based on Time information for the current time The kinematic branch prediction value, This represents the corresponding AoA branch prediction value.

[0076] Specifically, embodiments of the present invention employ an exponentially weighted moving average to smoothly update the error variance: ; ; in, and These represent the prediction error variances of the kinematic branch and the AoA branch at the current time, respectively. High-confidence AoA, geometric LOS, link peak direction, or their fusion results can be used. , This is the variance smoothing coefficient.

[0077] Specifically, in this embodiment of the invention, the weights of each branch are calculated according to the minimum variance fusion criterion, and the branch with the smaller the prediction error variance is assigned a larger weight: , , in, and These are the fusion weights for the kinematics branch and the AoA branch, respectively.

[0078] Specifically, in this embodiment of the invention, the predicted values ​​of the two branches are weighted and fused according to the aforementioned weights to obtain the prior future LOS prediction angle. : .

[0079] In actual calculations, the embodiments of the present invention can... and Add a very small positive number to it. To avoid the denominator being zero.

[0080] It is understandable that adaptive fusion of two complementary predictive information, “platform kinematics” and “AoA sequence”, relies more on kinematics when the motion state is reliable and more on AoA prediction when the angle measurement sequence is more stable, thereby improving the robustness of the prior future LOS.

[0081] Specifically, embodiments of the present invention are based on the current actual beam pointing Compared with prior future LOS prediction angle Calculate the prediction pointing error, which is expressed as: , in, This represents the predicted pointing error calculated relative to the prior future LOS, based on the current actual beam direction.

[0082] Specifically, embodiments of the present invention calculate the predicted link quality based on the secondary characteristics of the antenna pattern near the main lobe, and the predicted link quality is expressed as: , in, This serves as the reference link quality when the beam is aligned or in the reference direction. This is the descent coefficient obtained by fitting the curve near the main lobe of the antenna pattern.

[0083] In this embodiment of the invention, the decreasing coefficient This can be obtained through antenna pattern calibration: recording different deflection angles within a small angular range of the main lobe. Normalized link quality ,use Perform a quadratic fit; if implemented in three dimensions, the descent coefficients in both azimuth and pitch directions or quadratic terms in matrix form can be used.

[0084] Understandably, mapping the angle prediction error to a quantity comparable to the measured link quality provides a model reference for residual construction, enabling RSSI / SINR to correct LOS predictions in reverse.

[0085] Specifically, the embodiments of the present invention will measure the link quality. With predicted link quality By subtracting the values, we obtain the link quality residuals: , in, This represents the current link quality residual.

[0086] Specifically, embodiments of the present invention perform exponentially weighted moving smoothing on the link quality residuals to obtain smoothed link quality residuals. : , in, is the residual smoothing coefficient, and .

[0087] In this embodiment of the invention, when When the value is negative, it indicates that the actual link quality is lower than the model prediction, which may indicate LOS prediction bias, antenna main lobe eccentricity, attitude estimation error, or actuator response error; when When the value is positive, it indicates that the actual link quality is better than the model prediction, and the correction intensity can be reduced or the current prediction can be maintained.

[0088] Understandably, since RSSI / SINR can only reflect the quality of the link and cannot directly indicate the direction of angle correction, it is important to consider the candidate pre-pointing angle. Apply small positive and negative angle perturbations nearby.

[0089] Specifically, this invention calculates the error angle between the current beam pointing and the prior predicted angle, and maps it to the predicted link quality using the secondary attenuation characteristics near the main lobe of the antenna pattern. This error is then subtracted from the real-time measured link quality to obtain the link quality residual. An exponentially weighted moving average is used to smooth the residual, suppressing transient noise interference, and finally outputting a smoothed link quality residual. The quadratic fitting model of the antenna pattern bridges the angle and signal domains mathematically, allowing purely geometric pointing deviations to be transformed into predicted link quality values. The exponentially weighted moving average, while preserving the residual trend, effectively filters out transient jitter caused by multipath fast fading and sudden interference, preventing drastic changes in the subsequent correction angle due to a single abnormal measurement. This ensures the smoothness and stability of system control, thereby improving the pre-pointing accuracy and link holding stability of the ultra-narrow beam terahertz link in dynamic moving scenarios.

[0090] Specifically, the candidate pre-pointing angle in the embodiments of the present invention It can be initialized as a priori future LOS prediction angle Alternatively, you can retrieve the valid command from the previous cycle. Or the weighted average of the two.

[0091] Specifically, in this embodiment of the invention, symmetrical perturbations are applied to both sides of the candidate pre-direction angle to obtain: , , in, and These are the positive and negative perturbation angles on either side of the candidate pre-pointing angle, respectively. The amplitude of the two-sided perturbation.

[0092] Specifically, embodiments of the present invention employ the central difference method to estimate the link quality angle gradient of the link quality versus the angle. : , in, The first-order differential gradient of link quality with respect to angle. and Beam pointing and The normalized link quality measured or estimated in real time can be approximated by short-stay measurements, phased array fast codebook switching, or pattern model.

[0093] In this embodiment of the invention, if This indicates that increasing the beam angle is beneficial to improving link quality; if This indicates that reducing the beam angle is more advantageous; if If the value is close to zero, it indicates that the link quality is not sensitive to the perturbation, and the link residual correction weight should be reduced.

[0094] Understandably, extracting the "correction to the left or right" direction information from the link quality that only indicates "good or bad" allows for obtaining the angle correction direction without extensive scanning, thus reducing disturbance to the communication link.

[0095] Specifically, this invention uses a candidate pre-pointing angle as a reference and applies symmetrical angular perturbations with an amplitude of δ on both sides to obtain the link quality corresponding to the positive and negative perturbation angles, respectively. The central difference method is used to calculate the first-order gradient of the link quality with respect to the angle, thereby obtaining information on the correction direction and intensity. Only a very small angle of perturbation detection is required near the pre-pointing angle, without the need for large-scale beam scanning, to obtain reliable gradient estimation. This minimizes the cost of obtaining the direction required for correction, reduces the risk of disturbance and interruption to normal communication links, and thus improves the pre-pointing accuracy and link maintenance stability of ultra-narrow beam terahertz links in dynamic moving scenarios.

[0096] Specifically, embodiments of the present invention are based on smoothing link quality residuals. and link quality angle gradient Generate link residual correction angle, link residual correction angle Represented as: , in, For residual correction gain, For the maximum allowable correction angle, For symbolic functions, This means that the correction will only be activated when the measured link quality is lower than the predicted link quality.

[0097] In this embodiment of the invention, and It can be determined through closed-loop stability requirements, simulation, or field calibration, and Usually not greater than Or its preset ratio.

[0098] It is understandable that by converting the magnitude of the link quality residual and the direction of the link quality gradient into specific angle corrections, when the model prediction is too optimistic, it can make limited corrections according to the direction of link quality improvement, thus avoiding excessive oscillation.

[0099] Specifically, embodiments of the present invention are based on the link quality angle gradient. The clarity of the link is used to calculate the link credibility. Represented as: , in, This is the gradient stability constant, used to prevent... When smaller Too large; This represents the link gradient confidence level, with a value ranging from 0 to 1.

[0100] Specifically, in this embodiment of the invention, the link correction amount is weighted according to confidence level and then superimposed on the prior future LOS prediction angle to obtain the final future LOS prediction angle. Represented as: , in, For the prior future LOS prediction angle, For link residual correction angle, It is the pre-pointing control that is actually used to determine the target direction.

[0101] Understandably, adding the link residual correction to the prior future LOS in a confidence-weighted manner preserves the main role of kinematics and AoA prediction, while allowing the actual link quality to make a small closed-loop correction to the prediction direction.

[0102] Specifically, this invention calculates a link residual correction angle based on the sign and magnitude of the smoothed link quality residual and the angle gradient. This correction angle is limited to a maximum magnitude and is activated only when the measured quality is worse than the prediction. Simultaneously, the correction angle is weighted and superimposed onto the prior future LOS prediction angle based on the reliability of the link feedback evaluated by the gradient magnitude. This generates the final prediction angle used for beam control. The correction amount is determined by the magnitude of the residual and the direction by the gradient sign. This allows the link quality feedback to be transformed into a specific correction action in the angle domain. This ensures that the system maintains the prediction direction when the link quality is better than expected, avoiding overcorrection and beam jitter caused by positive residuals. As a result, the pre-pointing accuracy and link holding stability of the ultra-narrow beam terahertz link in dynamic moving scenarios are improved.

[0103] Specifically, embodiments of the present invention are based on the kinematic prediction error variance AoA prediction error variance Link residual correction angle and actuator error Calculate the variance of future prediction errors: , in, This indicates the uncertainty of future LOS predictions. This indicates the uncertainty introduced by link residual correction. This indicates the uncertainty introduced by actuator lag, backlash, quantization codebook, or array weight error.

[0104] In this embodiment of the invention, It can be estimated from the rolling variance of the residuals after link correction. It can be estimated from the actuator calibration error or the rolling difference between the command angle and the feedback angle.

[0105] Understandably, describing both the forecast mean and forecast uncertainty simultaneously allows for the identification of potential lock-out risks due to excessive variance, even if the average pointing error is small.

[0106] Specifically, embodiments of the present invention will control the timing of the command taking effect. The future pointing error is treated as a random variable And based on the ultra-narrow beam half-width Calculate the probability of future lock loss.

[0107] The embodiments of the present invention make If the future beam direction estimate is given when pre-pointing is not triggered or when the current control strategy continues, then: , , , , in, Let the future point of error be a random variable. Let be the mean of the random variable pointing to future errors. The variance of the future-pointing random variable for error. For the variance of future beam execution error, It is the standard normal distribution function. This represents the probability of future lock loss.

[0108] Understandably, quantifying the relationship between future pointing error and ultra-narrow beam half-width as a probability can help identify high-risk conditions before the link is actually interrupted.

[0109] Specifically, this invention quantifies multiple uncertainties, such as kinematic prediction error, AoA prediction error, link residual correction error, and actuator error, into a unified future prediction error variance. It then treats the future beam pointing as a random variable with the prediction angle as the mean and the total variance as the statistical distribution. Combined with the ultra-narrow beam half-width, it calculates the future loss-of-lock probability by using a standard normal distribution function to determine the beam deviation exceeding the half-width range. This comprehensively reflects the potential loss-of-lock risk implied by the increase in prediction variance. The loss-of-lock probability provides a quantitative criterion for pre-pointing triggering decisions, avoiding potential omissions or misjudgments that might occur if relying solely on empirical thresholds. This improves the pre-pointing accuracy and link stability of ultra-narrow beam terahertz links in dynamic moving scenarios.

[0110] Specifically, embodiments of the present invention are based on the probability of future lock loss. Predicting link quality The final future LOS prediction angle and the current actual beam pointing determine whether pre-pointing control is triggered: like ,but , or ,but , or , in, For pre-pointing the trigger flag, This is the threshold for the probability of losing the lock. This is the link quality threshold. This is the error margin coefficient.

[0111] In this embodiment of the invention, the probability threshold of lock loss Link quality threshold and error margin coefficient It can be set according to business reliability requirements, allowable recapture time, target packet error rate, or system test results.

[0112] It is understandable that the trigger judgment is made by combining three types of indicators: probability risk, link quality, and angle deviation, to avoid misjudgment by a single indicator, and to enter pre-pointing control in advance when the risk increases but the link is still available.

[0113] Specifically, in this embodiment of the invention, when the pre-pointing trigger flag is 1, a beam pre-pointing command is generated based on the final future LOS prediction angle, the actuator feedforward compensation amount, and the historical deviation compensation amount. The beam pre-pointing command is expressed as follows: , in, To output beam pre-pointing commands to gimbals, phased arrays, or hybrid beam control systems, This is the feedforward compensation amount calculated based on the platform angular velocity, actuator dynamics, or phased array switching time. This is the deviation compensation amount estimated from long-term residuals, installation errors, or main lobe eccentricity.

[0114] In this embodiment of the invention, the beam pre-pointing command is directed towards a future effective time, rather than the current time.

[0115] Understandably, the final future LOS prediction result is converted into an executable control command. The output command already includes time delay compensation, link residual correction and actuator deviation compensation, which can directly drive different types of beam control systems.

[0116] Specifically, this invention triggers pre-pointing control based on the future loss-of-sight probability, predicted link quality, and the angular deviation between the final predicted angle and the current pointing angle. After triggering, the final future LOS prediction angle is combined with the actuator feedforward compensation and historical deviation compensation to generate a beam pre-pointing command for the future effective moment. This avoids misjudgment and omission caused by noise fluctuations or special operating conditions due to a single indicator, making the pre-pointing decision more robust and reliable. The output command not only includes the prediction of the future LOS, but also compensates for the dynamic characteristics of the actuator and the long-term installation or calibration errors of the system, ensuring that the beam can accurately align with the target direction at the actual effective moment. This improves the pre-pointing accuracy and link stability of the ultra-narrow beam terahertz link in dynamic moving scenarios.

[0117] Specifically, embodiments of the present invention are based on beam pre-pointing commands. This drives the actuator to respond and updates the current actual beam pointing. .

[0118] Specifically, the embodiments of the present invention employ the following methods depending on the type of actuator: If a mechanical gimbal is used, the current actual beam pointing can be updated according to the first-order actuator model. , in, For the response coefficient of the executing agency, In order to implement the delayed departure time, If a phased array is used, the target angle is mapped to the beam codebook or array weights. , in, This is the mapping function from the target angle to the array weights or codebook index. The weights are those of the phased array.

[0119] In this embodiment of the invention, for a mechanical-electronic hybrid system, coarse pointing can be performed first by a mechanical gimbal, and then fine pointing can be performed by phased array weights.

[0120] Understandably, the description of the actual response after the command enters the actuator goes beyond the level of prediction algorithms and also discloses the interface implementation with mechanical gimbals, phased arrays and hybrid beam systems.

[0121] Specifically, in each sampling moment, the embodiments of the present invention record the true LOS direction, AoA observation, kinematic prediction, AoA prediction, measured link quality, predicted link quality, link quality residual, link quality gradient, link residual correction angle, final future LOS prediction angle, future loss of lock probability, pre-pointing trigger flag, beam pre-pointing command, current actual beam pointing and loss of lock flag.

[0122] Specifically, the unlock flag in this embodiment of the invention is defined as: , in, This is an indicator function that takes the value 1 if the condition is true, and 0 otherwise.

[0123] Then, the average unlocking rate within the statistical period is: , in, This represents the average percentage of lockouts within the statistical period.

[0124] Specifically, this invention generates a pre-pointing command and then performs differentiated processing based on the type of actuator. Simultaneously, it records detailed key variables across the entire link, from LOS prediction and link quality gradient to correction angle and lock-out flag, at each sampling moment. It also defines the lock-out flag and the average lock-out ratio within the statistical period. Corresponding response models are adopted for the dynamic characteristics of different actuators to avoid pointing deviations caused by actuator modeling errors. The time-series recording of key variables across the entire link constructs a complete data loop, providing data support for system parameter tuning and fault diagnosis. Based on the average lock-out ratio statistically analyzed from the recorded data, the system's tracking performance is transformed into a quantifiable and comparable indicator, providing a clear evaluation benchmark for algorithm iterative optimization. This improves the pre-pointing accuracy and link holding stability of the ultra-narrow beam terahertz link in dynamic moving scenarios.

[0125] Please see Figures 2 to 3 The figures shown are a geometric scene diagram of a two-dimensional mobile terahertz link and a simulation result diagram comparing pointing errors.

[0126] Example 1, A two-dimensional mobile terahertz link simulation scenario is constructed, with the fixed terahertz node located at the origin of the coordinate system and the mobile node moving along a curved trajectory. This trajectory includes dynamic factors such as turning, curvature changes, and distance changes, which can simulate the change of the LOS direction over time in the terahertz link of the mobile platform.

[0127] like Figure 2 As shown, the fixed terahertz node is located at the origin of the coordinate system, and the mobile node moves along the curved trajectory. The color represents the change over time. The gray dashed line in the figure represents the LOS direction between the fixed node and the mobile node at different times. It can be seen that the LOS direction changes continuously during the movement of the mobile platform. If there is an execution delay in the beam control command, tracking only the current LOS will produce a significant lag. Therefore, it is necessary to predict the LOS direction at the future effective time.

[0128] Figure 3This paper compares the pointing errors of traditional delay tracking methods, motion-AoA pre-pointing methods, and the LQRC-LOSP method of this invention. The red dashed line in the figure represents the lock-out threshold, and the red shaded area represents the lock-out region after exceeding the threshold. The traditional method shows a significant increase in pointing error during periods of intense target movement, and repeatedly exceeds the lock-out threshold. The motion-AoA pre-pointing method can reduce some errors, but error fluctuations still exist in complex dynamic stages. This embodiment, through link quality residual correction and execution delay compensation, keeps the pointing error at a low level for most of the time and effectively avoids entering the lock-out region.

[0129] The simulation results show that unifying the platform motion state, AoA history window, RSSI / SINR link quality residuals, and bilateral perturbation gradients into the future LOS prediction control framework can improve the pre-pointing accuracy and link holding stability of the ultra-narrow beam terahertz link in dynamic moving scenarios.

[0130] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present application should be included within the scope of protection of the present application.

Claims

1. A method for terahertz link pre-pointing control based on link quality residual error correction, characterized in that, include: Acquire the ultra-narrow beam half-width, actuator feedforward compensation, historical deviation compensation, current actual beam pointing, measured link quality, and prior future LOS prediction angle; The prediction pointing error is determined based on the prior future LOS prediction angle and the current actual beam pointing, and the prediction link quality is determined based on the prediction pointing error. Based on the measured link quality and the predicted link quality, the link quality residual is determined, and the link quality residual is smoothed to obtain the smoothed link quality residual; A symmetrical perturbation is applied to the prior future LOS prediction angle to obtain a positive perturbation angle and a negative perturbation angle. The link quality angle gradient is determined based on the measured link quality corresponding to the positive perturbation angle and the negative perturbation angle. The link residual correction angle is determined based on the smoothed link quality residual and the link quality angle gradient. The link confidence level is determined based on the link quality angle gradient. The link residual correction angle is then weighted and superimposed onto the prior future LOS prediction angle based on the link confidence level to obtain the final future LOS prediction angle. The probability of future loss of lock is determined based on the final future LOS prediction angle, the current actual beam pointing, and the ultra-narrow beam half-width. Whether to trigger pre-pointing control is determined based on the future loss of lock probability, the predicted link quality, the final future LOS prediction angle, and the current actual beam pointing. When pre-pointing control is triggered, a beam pre-pointing command is generated based on the final future LOS prediction angle, the actuator feedforward compensation amount, and the historical deviation compensation amount.

2. The method of link quality residual error corrected terahertz link pre-pointing control according to claim 1, characterized in that, The process of determining the prediction pointing error based on the prior future LOS prediction angle and the current actual beam pointing, and determining the prediction link quality based on the prediction pointing error, includes: calculating a predicted pointing error based on the current actual beam pointing and the prior future LOS predicted angle, the predicted pointing error is represented as: ， in, The current actual beam direction, For the prior future LOS prediction angle; The predicted link quality is determined based on the predicted pointing error. Represented as: ， in, This serves as the reference link quality when the beam is aligned or in the reference direction. This is the descent coefficient obtained by fitting the curve near the main lobe of the antenna pattern.

3. The terahertz link pre-pointing control method based on link quality residual correction according to claim 1, characterized in that, The process of determining the link quality residual based on the measured link quality and the predicted link quality, and smoothing the link quality residual to obtain a smoothed link quality residual, includes: The link quality residual is obtained by subtracting the measured link quality from the predicted link quality. : ， in, This is to measure the link quality. The smoothed link quality residual is obtained by performing exponentially weighted moving smoothing on the link quality residual. : ， in, The residual smoothing coefficient is... The sampling period.

4. The terahertz link pre-pointing control method based on link quality residual correction according to claim 1, characterized in that, The process of applying a symmetrical perturbation to the prior future LOS prediction angle to obtain a positive perturbation angle and a negative perturbation angle, and determining the link quality angle gradient based on the measured link quality corresponding to the positive perturbation angle and the negative perturbation angle, includes: The candidate pre-pointing angle is obtained by initializing the prior future LOS prediction angle. Symmetrical perturbations are applied to both sides of the candidate pre-pointing angle to obtain a positive perturbation angle and a negative perturbation angle. The positive perturbation angle... and the negative perturbation angle Represented as: ， ， in, Candidate pre-pointing angle, The amplitude of the two-sided perturbation; The link quality angular gradient is determined by estimating the measured link quality corresponding to the positive and negative perturbation angles using the central difference method. Represented as: ， in, and These are the measured link qualities corresponding to the positive and negative perturbation angles, respectively.

5. The terahertz link pre-pointing control method based on link quality residual correction according to claim 1, characterized in that, The process of determining the link residual correction angle based on the smoothed link quality residual and the link quality angle gradient includes: Based on the smoothed link quality residual and the link quality angle gradient, a link residual correction angle is generated. Represented as: ， in, For residual correction gain, For the maximum allowable correction angle, For symbolic functions, surface The correction is only initiated when the measured link quality is lower than the predicted link quality.

6. The terahertz link pre-pointing control method based on link quality residual correction according to claim 1, characterized in that, The process of determining the link confidence level based on the link quality angle gradient, and then weighting and superimposing the link residual correction angle onto the prior future LOS prediction angle according to the link confidence level to obtain the final future LOS prediction angle includes: Link reliability is calculated based on the clarity of the link quality angle gradient. Represented as: ， in, It is the gradient stability constant; The link correction values ​​are weighted according to confidence level and then superimposed onto the prior future LOS prediction angle to obtain the final future LOS prediction angle. Represented as: ， in, This is to pre-point the target direction that is actually used for control.

7. The terahertz link pre-pointing control method based on link quality residual correction according to claim 1, characterized in that, The process of determining the future loss-of-lock probability based on the final future LOS prediction angle, the current actual beam pointing, and the ultra-narrow beam half-width includes: Based on the final future LOS prediction angle and the current actual beam pointing, the control command will take effect at the specified time. The future pointing error is considered a random variable. Represented as; ， in, Let be the mean of the random variable pointing to future errors. The variance of the future-pointing random variable for error; The mean of the future pointing error random variable is expressed as: ， in, This is for estimating the future beam direction when pre-pointing is not triggered or when the current control strategy continues to be executed; The variance of the future pointing error random variable is expressed as: ， in, For the variance of future beam execution error, The variance of the future prediction error; ， in, It is the standard normal distribution function. It is an ultra-narrow beam half-width.

8. The terahertz link pre-pointing control method based on link quality residual correction according to claim 1, characterized in that, The process of determining whether to trigger pre-pointing control based on the future loss-of-lock probability, the predicted link quality, the final future LOS prediction angle, and the current actual beam pointing includes: like ,but , or ,but , or , in, For pre-pointing the trigger flag, This is the threshold for the probability of losing the lock. This is the link quality threshold. This is the error margin coefficient.

9. The terahertz link pre-pointing control method based on link quality residual correction according to claim 1, characterized in that, The process of generating beam pre-pointing commands based on the final future LOS prediction angle, the actuator feedforward compensation amount, and the historical deviation compensation amount includes: When the pre-pointing trigger flag is 1, a beam pre-pointing command is generated based on the final future LOS prediction angle, the actuator feedforward compensation, and the historical deviation compensation. Represented as: ， in, For the feedforward compensation amount of the implementing agency, This is the amount to compensate for historical deviations.

10. The terahertz link pre-pointing control method based on link quality residual correction according to any one of claims 1 to 9, characterized in that, It also includes driving the actuator to respond and updating the current actual beam pointing according to the beam pre-pointing command.

Images