An unmanned aerial wide-angle scanning millimeter wave antenna adaptive regulation system

Abstract

This invention relates to the field of adaptive antenna control technology, specifically disclosing an adaptive control system for a wide-angle scanning millimeter-wave antenna on a UAV. The system includes a multi-source heterogeneous sensing module for collecting platform attitude data, airflow disturbance data, environmental attenuation data, and three-dimensional dynamic deformation data of the flexible array. Data fusion is performed through a filtering unit to output four-dimensional sensing results. The technical solution of this invention can effectively improve the bottleneck of communication link lockout in actual flight operations such as UAVs encountering complex airflow or severe maneuvers. When the flexible antenna array is about to deform or is undergoing deformation due to physical environmental influences, the system no longer passively waits for signal attenuation. Instead, it deeply integrates multi-dimensional environmental and flight maneuver commands to predict the deformation trend in advance and directly triggers a resource scheduling mechanism at the underlying level, rapidly allocating core computing power to ensure extremely low-latency computational output.

Description

Technical Field

[0001] This invention relates to the field of adaptive antenna control technology, specifically to an adaptive control system for a wide-angle scanning millimeter-wave antenna for unmanned aerial vehicles (UAVs). Background Technology

[0002] With the widespread application of UAVs in high-speed communication, millimeter-wave flexible array antennas with wide-angle scanning capabilities have become key components due to their lightweight design and conformal patch characteristics. In actual high-altitude flight operations, in order to maintain a stable air-to-ground communication link, the antenna system needs to adjust the beam pointing in real time according to the flight attitude. Current conventional antenna control schemes usually rely solely on static preset electromagnetic calibration matrices or simply on single fuselage attitude feedback provided by inertial navigation equipment for beam guidance and compensation.

[0003] However, under actual flight conditions involving complex airflow and fuselage vibration, existing control schemes face the following conflicts:

[0004] On the one hand, flexible antennas are prone to unpredictable dynamic physical deformation under aerodynamic loads and fuselage vibrations, which leads to complex coupling interference between the antenna's inherent electromagnetic properties, environmental attenuation and structural distortion, causing the original static calibration data to become invalid instantly.

[0005] On the other hand, traditional systems lack the ability to predict deformation trends in advance and the dynamic scheduling mechanism of underlying computing resources. When massive and sudden composite interference data floods in, the processor is likely to generate extremely high computing latency. This high computing latency creates a major contradiction with the requirement for ultra-fast beam alignment in the highly dynamic flight environment of UAVs. Summary of the Invention

[0006] This invention aims to at least partially address one of the technical problems in related technologies. Therefore, the objective of this invention is to propose an adaptive control system for a wide-angle scanning millimeter-wave antenna on a UAV, to enhance the reliability of communication systems in complex airspace.

[0007] To achieve the above objectives, a first aspect of the present invention provides an adaptive control system for a wide-angle scanning millimeter-wave antenna on a UAV, comprising:

[0008] The multi-source heterogeneous sensing module is used to collect platform attitude data, airflow disturbance data, environmental attenuation data and flexible array three-dimensional dynamic deformation data, and to perform data fusion through the filtering unit to output four-dimensional sensing results.

[0009] The electromagnetic-environment-structure coordinated control module is connected to the multi-source heterogeneous sensing module. It is used to separate the electromagnetic inherent gain loss, environmental gain loss and structural gain loss according to the four-dimensional sensing results, allocate compensation factor weights according to the loss ratio, and perform pre-distortion compensation in combination with the mapping table parameters to generate the final amplitude and final phase of the antenna excitation.

[0010] A low-latency parallel processing module, connected to the collaborative control module, is used to dynamically allocate processor computing resources according to the extracted deformation parameter level using a deformation-triggered resource scheduling mechanism, and output the compensation parameters corresponding to the pre-distortion compensation.

[0011] A multi-dimensional intelligent learning module is used to output deformation prediction values ​​in advance based on UAV maneuver commands and historical deformation data, and send them to the collaborative control module for pre-compensation.

[0012] The cross-system collaboration and anomaly handling module is used to monitor the rate of change of deformation parameters in real time. When the deformation parameters exceed the limit threshold, it triggers array reconstruction and coordinates the UAV flight control system and ground base station to adjust the attitude.

[0013] To achieve the above objectives, a second aspect of the present invention proposes an adaptive control method for a wide-angle scanning millimeter-wave antenna on a UAV, comprising:

[0014] The system collects platform attitude data, airflow disturbance data, environmental attenuation data, and three-dimensional dynamic deformation data of the flexible array, and then fuses the data through a filtering unit to output four-dimensional sensing results.

[0015] Based on the drone's maneuvering commands and historical deformation data, the deformation prediction values ​​are output in advance for pre-compensation.

[0016] Based on the extracted deformation parameter levels, a deformation-triggered resource scheduling mechanism is used to dynamically allocate processor computing resources in order to output the compensation parameters corresponding to the pre-distortion compensation.

[0017] Based on the four-dimensional perception results, electromagnetic inherent gain loss, environmental gain loss and structural gain loss are separated. Compensation factor weights are allocated according to the loss ratio, and the pre-compensation and pre-distortion compensation are performed in combination with the mapping table parameters and the deformation prediction values ​​to generate the final amplitude and final phase of the antenna excitation.

[0018] The system monitors the rate of change of the deformation parameters in real time, triggers array reconstruction when the deformation parameters exceed the limit threshold, and coordinates the UAV flight control system and ground base station to adjust the attitude.

[0019] To achieve the above objectives, a third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described adaptive control method for a wide-angle scanning millimeter-wave antenna on a UAV.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0021] The technical solution of this invention can effectively improve the bottleneck of communication link lockout in actual flight operation conditions such as UAV encountering complex airflow or violent maneuvers. When the flexible antenna array is about to or is undergoing deformation due to the influence of the physical environment, the system no longer passively waits for signal attenuation, but predicts the deformation trend in advance by deeply integrating multi-dimensional environment and flight maneuver commands, and directly triggers the resource scheduling mechanism at the bottom layer to quickly tilt the core computing power to ensure extremely low latency computing output;

[0022] Meanwhile, this scheme decouples hybrid interference into three independent dimensions: electromagnetic, environmental, and structural, and adaptively matches compensation weights and mapping parameters to perform precise amplitude and phase predistortion adjustments. This enables UAVs to proactively offset the effects of physical deformation even in extreme application scenarios involving severe turbulence and significant physical distortion of the antenna array, maintaining high beam pointing accuracy and communication link stability. It represents a leap from passive, delayed response to proactive, predictive, and adaptive defense, significantly enhancing the survivability and reliability of communication systems in complex airspace. Attached Figure Description

[0023] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0024] Figure 1 This is a schematic diagram illustrating the implementation of the adaptive control system for wide-angle scanning millimeter-wave antennas on unmanned aerial vehicles provided by this invention.

[0025] Figure 2 This is the three-dimensional dynamic deformation time-frequency domain separation spectrum of the flexible array in the adaptive control method for wide-angle scanning millimeter-wave antennas on UAVs provided by this invention;

[0026] Figure 3 This is a three-dimensional response surface plot of the multidimensional electromagnetic gain compensation quantity in the adaptive control method for wide-angle scanning millimeter-wave antennas on UAVs provided by the present invention;

[0027] Figure 4 This is the LSTM-based displacement prediction and measured residual convergence curve in the adaptive control method for wide-angle scanning millimeter-wave antennas on UAVs provided by this invention.

[0028] Figure 5 This invention provides a comparison of the far-field radiation pattern after low-frequency deformation distortion and radio frequency spatial compensation in the adaptive control method for wide-angle scanning millimeter-wave antennas on UAVs.

[0029] Figure 6 This is a comparison diagram of the 256-QAM communication constellation before and after high-frequency phase jitter compensation in the adaptive control method of wide-angle scanning millimeter-wave antenna for UAVs provided by this invention;

[0030] Figure 7 This is a timing diagram of the asymmetric hysteresis scheduling state transition under complex airflow disturbance in the adaptive control method for wide-angle scanning millimeter-wave antennas on UAVs provided by this invention;

[0031] Figure 8 This is a flowchart illustrating the adaptive control method for wide-angle scanning millimeter-wave antennas on unmanned aerial vehicles provided by the present invention.

[0032] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0033] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0034] The following description, with reference to the accompanying drawings, describes an adaptive control method, system, and electronic device for a wide-angle scanning millimeter-wave antenna on a UAV according to embodiments of the present invention.

[0035] Example 1:

[0036] This embodiment provides an adaptive control system for a wide-angle scanning millimeter-wave antenna on a UAV, which is mainly used to maintain the stability of the air-to-ground millimeter-wave high-frequency communication link under complex weather conditions and high-dynamic flight maneuvers.

[0037] This system relies on the efficient scheduling of multi-dimensional hardware sensors and underlying computing resources to construct a complete closed loop from environmental perception, feature extraction, prediction and compensation to anomaly protection. The system includes: a multi-source heterogeneous sensing module, an electromagnetic-environment-structure coordinated control module, a low-latency parallel processing module, a multi-dimensional intelligent learning module, and a cross-system collaboration and anomaly handling module.

[0038] Specifically, the multi-source heterogeneous sensing module is used to collect platform attitude data, airflow disturbance data, environmental attenuation data, and three-dimensional dynamic deformation data of the flexible array. Data fusion is performed through a filtering unit to output four-dimensional sensing results. The electromagnetic-environment-structure coordinated control module, connected to the multi-source heterogeneous sensing module, separates the electromagnetic inherent gain loss, environmental gain loss, and structural gain loss based on the four-dimensional sensing results. It allocates compensation factor weights according to the loss proportions and performs pre-distortion compensation in conjunction with mapping table parameters to generate the final amplitude and phase of the antenna excitation. The low-latency parallel processing module, connected to the coordinated control module, dynamically allocates processor computing resources based on the extracted deformation parameter levels using a deformation-triggered resource scheduling mechanism, and outputs the compensation parameters corresponding to the pre-distortion compensation. The multi-dimensional intelligent learning module outputs predicted deformation values ​​in advance based on UAV maneuver commands and historical deformation data, and sends them to the coordinated control module for pre-compensation. The cross-system coordination and anomaly handling module monitors the rate of change of the deformation parameters in real time. When the deformation parameters exceed the limit threshold, it triggers array reconstruction and coordinates the UAV flight control system and ground base station for attitude adjustment.

[0039] For example, the multi-source heterogeneous sensing module includes a flexible sensing-communication integrated substrate and a distributed strain sensor network. In practical applications, the fuselage surface of a drone typically presents an irregular curved surface. The flexible sensing-communication integrated substrate is made of dielectric materials with high mechanical toughness, such as polyimide, enabling it to be conformally mounted on the underside of the drone's fuselage or wings.

[0040] The flexible sensing-communication integrated substrate is divided into multiple finite element meshes, each corresponding to a millimeter-wave communication antenna element. Sensing nodes from the distributed strain sensor network are arranged at the vertices of each finite element mesh. This topology ensures that each antenna element radiating electromagnetic waves is within the mechanical monitoring envelope of four sensing nodes. When aerodynamic loads are applied to the flexible substrate, the sensing nodes at each vertex can synchronously output changes in resistance or capacitance, thereby characterizing the local stress state of the mesh region.

[0041] Optionally, during mission execution, the rotation of the UAV's rotor, engine vibration, and the effects of high-altitude crosswinds generate interference signals containing a large amount of high-frequency noise and low-frequency drift. Therefore, the filtering unit is designed as a Federated Extended Kalman Filter (FAF), comprising a platform disturbance sub-filter for fusing inertial navigation and magnetometer data, an airflow disturbance sub-filter for fusing radar micro-motion data, an environmental perception sub-filter for extracting attenuation data, and a structural deformation sub-filter for eliminating strain measurement noise. These four sub-filters output in parallel and are synthesized by the main filter into the four-dimensional perception result.

[0042] Specifically, the platform disturbance sub-filter addresses the integral drift phenomenon that inertial navigation units are prone to during long-term hovering or long-endurance flight. It uses absolute heading data from the magnetometer to update the state covariance, thereby outputting an accurate low-frequency attitude reference. The airflow disturbance sub-filter targets the Doppler frequency shift characteristics acquired by the airborne miniature radar, filtering out periodic clutter generated by rotor blade rotation and extracting airflow disturbance data characterizing external wind shear or turbulence intensity. The environmental perception sub-filter smooths the time-series received power of the reference signals in the transmit and receive links, eliminating power jumps caused by internal receiver thermal noise and extracting environmental attenuation data mainly caused by rain attenuation, cloud and fog absorption, or long-distance path loss. The structural deformation sub-filter addresses thermal noise and electromagnetic interference at the sensing nodes, performing measurement and time updates to ensure high signal-to-noise ratio in the output strain data. The main filter employs an information allocation principle, dynamically allocating federated information allocation coefficients based on the local state estimation covariance matrix of each sub-filter, ultimately outputting a comprehensive four-dimensional perception result including attitude, airflow, environment, and deformation.

[0043] It should be noted that the raw quantities output by the sensing nodes are only scalar strain values ​​and cannot be directly applied to electromagnetic phase compensation in three-dimensional space. Therefore, the multi-source heterogeneous sensing module also includes a dynamic deformation field calculation unit.

[0044] The dynamic deformation field calculation unit, based on the linear elasticity finite element method, transforms the measurement data collected by the distributed strain sensor network into the three-dimensional displacement and rotation angle of each finite element mesh node, reconstructing the three-dimensional dynamic deformation field. In actual calculations, this unit utilizes the Young's modulus and Poisson's ratio of the flexible substrate material, as well as a pre-constructed stiffness matrix, to map the local strain tensor into physical displacement in the global coordinate system.

[0045] Furthermore, the calculation unit performs time-frequency domain correlation separation between the three-dimensional dynamic deformation field and the platform vibration data, resolving the high-frequency deformation components caused by fuselage vibration and the low-frequency deformation components caused by aerodynamic loads. In applications, vibrations caused by fuselage engines or rotors typically manifest as periodic jitter in a specific frequency band, with limited deformation amplitude but extremely high frequency of change; while aerodynamic loads caused by high-altitude gusts or UAV acceleration and circling manifest as macroscopic bending or twisting of the entire substrate, with lower frequencies but larger deformation magnitudes. By setting low-pass and high-pass digital filters in the frequency domain, the system can separate these two deformation characteristics with completely different physical mechanisms, providing an independent data dimension for subsequent differentiated resource scheduling and compensation.

[0046] like Figure 2This is a time-frequency domain separation spectrum of a flexible array under three-dimensional dynamic deformation. The horizontal axis of the graph represents time in seconds, the vertical axis represents frequency in Hertz, and the color bars on the side of the graph correspond to the deformation amplitude in millimeters.

[0047] The dark red waveform energy band at the bottom of the figure is concentrated in the low-frequency region around 2 Hz, and its corresponding deformation amplitude reaches 10 mm, which objectively reflects the macroscopic bending phenomenon of the flexible antenna array caused by aerodynamic load during the flight of the UAV.

[0048] At the same time, a clear light blue waveform energy band exists near 150 Hz, with a deformation amplitude of about 0.5 mm, recording the high-frequency mechanical vibration transmitted to the array surface by the rotation of the UAV rotor and engine vibration.

[0049] These two deformations with different physical mechanisms exhibit significant vertical and horizontal stratification and spatial isolation on the frequency axis of the spectrum, proving that the dynamic deformation field solution unit inside the system can effectively separate the complex mixed deformation into two independent variable dimensions.

[0050] This separation of time and frequency domains directly connects to the subsequent control links of the system, enabling the system to send spatial compensation parameters to the RF phase shifter and amplitude controller for beam reshaping for low-frequency deformations at the 10 mm level, while extracting phase jitter envelope features for high-frequency deformations at the 0.5 mm level, and directly performing inverse phase rotation compensation calculations on the communication data stream in the digital baseband domain, thereby ensuring the gain stability and beam pointing accuracy of the millimeter-wave communication link in highly dynamic flight environments.

[0051] To reduce the real-time online computation load on the airborne processor, this system adopts a technical approach that combines offline calibration with online table lookup.

[0052] Specifically, the electromagnetic-environment-structure coordinated control module has a built-in five-dimensional stepped predistortion mapping table. The query dimensions of this mapping table consist of scanning angle, excitation amplitude, excitation phase, electromagnetic performance attenuation, and deformation parameter level. During the ground calibration phase before UAV takeoff, a known gradient of deformation stress is applied to the flexible array using a mechanical servo system in a microwave anechoic chamber, and the far-field radiation pattern of the entire airspace is recorded using a vector network analyzer. The system constructs a database covering all operating conditions.

[0053] To improve lookup efficiency, the deformation parameter levels are discretized from low to high according to the deformation quantification index into four levels: no deformation (level 0), slight deformation (level 1), moderate deformation (level 2), severe deformation (level 3), and extreme deformation (level 4). This discretized hierarchical processing maps the infinitely continuous deformation state to a finite set of nodes, greatly reducing the space occupied by the onboard memory. The five-dimensional ladder predistortion mapping table records the electromagnetic amplitude compensation value, environmental amplitude compensation value, structural amplitude compensation value, electromagnetic phase compensation value, environmental phase compensation value, and structural phase compensation value to be extracted under the corresponding scanning angle range and deformation parameter level.

[0054] like Figure 3 This is a three-dimensional response surface plot of the multidimensional electromagnetic gain compensation. The horizontal axis of the plot represents the scanning angle in degrees, the vertical axis represents the deformation parameter level, and the vertical axis represents the structural amplitude compensation value in decibels. The color bars on the side of the plot correspond to the magnitude of the structural amplitude compensation value.

[0055] The surface in the figure exhibits typical nonlinear spatial distribution characteristics. In the gentle region where the scanning angle is close to zero degrees and the deformation parameter level is between no deformation level (level 0) and slight deformation level (level 1), the surface appears dark blue, corresponding to a very small required structural amplitude compensation value, approximately 0 to 2 dB, indicating that the electromagnetic radiation performance of the flexible antenna array is in a stable state at this time.

[0056] When the scanning angle shifts to a large angle of -90 degrees or +90 degrees, and the deformation parameter level climbs to the range of severe deformation level (level 3) to extreme deformation level (level 4), the spatial height of the curved surface increases sharply, and the surface color gradually transitions from light blue to yellow-green and finally to dark red. At this time, the structural amplitude compensation value output by the multidimensional step-by-step pre-distortion mapping table reaches a peak value of 15 to 20 dB.

[0057] This sharp increase in color gradient and surface height intuitively reflects the objective physical law that when the projection aperture reduction caused by wide-angle scanning and the severe physical and mechanical bending cause superimposed interference, the amplitude adjustment requirements of the system exhibit a non-linear surge.

[0058] By constructing and storing the smooth three-dimensional response surface data during the offline calibration phase, the electromagnetic environment structure collaborative control module can directly obtain the corresponding compensation value through cross-domain query based on the values ​​extracted by the multi-source heterogeneous sensing module under the complex conditions of actual flight operations. This significantly reduces the real-time calculation load of the airborne processor and ensures the pre-execution efficiency of multi-dimensional interference compensation and the accuracy of beam energy refocusing under high-dynamic flight conditions.

[0059] In actual flight communication, the decrease in antenna gain is often the result of multiple factors coupled together. For example, when a UAV performs large-angle beam scanning in rainy weather with crosswinds, electromagnetic gain attenuation (large-angle scanning leads to a smaller projection aperture), environmental gain attenuation (raindrops scatter and absorb millimeter waves), and structural gain attenuation (array bending causes phase center shift) will occur simultaneously.

[0060] Specifically, the electromagnetic-environment-structure coordinated control module performs the following calculation steps for pre-distortion compensation: First, the system extracts the independent loss estimates caused by the above three types of factors and calculates the proportion of the structural gain loss in the total gain loss.

[0061] In the above steps, let the electromagnetic inherent gain loss value be... The environmental gain loss value is The structural gain loss value is Let the structural gain loss ratio be . The proportion is calculated as shown in formula (1):

[0062] (1)

[0063] In formula (1), all loss values ​​are expressed as linear power ratios to facilitate scalar summation and proportional allocation.

[0064] Subsequently, the system dynamically allocates the electromagnetic compensation factor weights based on this proportion. Environmental compensation factor weights and structural compensation factor weights And the constraint is In actual allocation, when When the value increases, for example when a drone encounters strong airflow, array deformation becomes the dominant factor in signal attenuation, and the system will adjust the frequency accordingly. The value of this makes subsequent compensation resources more focused on resisting physical deformation.

[0065] Next, based on the current four-dimensional perception results, the system retrieves the corresponding electromagnetic amplitude compensation value from the five-dimensional step-by-step predistortion mapping table. Environmental amplitude compensation value Structural amplitude compensation value and the corresponding electromagnetic phase compensation value Environmental phase compensation value Structural phase compensation value Generate the final amplitude. The calculation process is shown in formula (2):

[0066] (2)

[0067] in, The initial uncompensated reference amplitude represents the ideal radio frequency excitation amplitude required for the antenna to operate in ideal free space with an absolutely flat array surface. , , The dimensionless amplitude adjustment coefficient is extracted from the mapping table; the final amplitude This refers to the actual control quantity sent to the digitally controlled attenuators or RF amplifiers of each channel in the array. The practical function of this formula is to smoothly increase the radiated power reduction caused by multidimensional interference without causing nonlinear saturation of the RF front end, through the linear superposition of normalized weighting coefficients and independent compensation values.

[0068] Generate the final phase The calculation process is shown in formula (3):

[0069] (3)

[0070] in, The initial, uncompensated reference phase is calculated from the beam target pointing angle and the spatial physical coordinates of each antenna element based on the basic principles of phased array. , , The phase offset angle compensation amount obtained from the table lookup, in degrees or radians; final phase This refers to the actual dwell phase sent to each channel's digitally controlled phase shifter. The practical physical meaning of this formula is that, for phase wavefront distortion caused by different mechanisms, component-weighted digital calculations are used to reconstruct the equiphase surface of the electromagnetic wave, ensuring that the transmitted electromagnetic waves achieve in-phase superposition in the far-field target region, thereby maintaining the high fixed-point gain of the beam.

[0071] It is also important to note that in certain highly dynamic flight maneuvers, such as when a drone is attempting to avoid obstacles or birds, mechanical deformation often occurs suddenly. If compensation is based solely on current measurement data, the system's internal digital computation delays and the state transition times of radio frequency devices will cause the compensation action to lag behind the actual deformation, potentially leading to transient link interruptions.

[0072] Specifically, the multi-dimensional intelligent learning module includes a structural deformation prediction sub-model based on a long short-term memory neural network. The time-series input feature matrix of this sub-model includes UAV maneuver commands, multi-axis flight speeds and accelerations, and historical continuous deformation data within a sliding time window. Maneuver commands generated internally by the UAV flight control system, such as increasing the speed of a rotor to perform a yaw maneuver, produce electrical signals before the actual physical displacement occurs. By using these preceding control commands as input to the long short-term memory neural network, the network can learn the complex nonlinear time-delay relationship between commands and subsequent deformations within its internal cellular state units. The structural deformation prediction sub-model outputs predicted deformation values ​​for a predetermined look-ahead time period, enabling the antenna control system to anticipate the future.

[0073] Optionally, a pre-compensation unit is coupled into the electromagnetic-environment-structure coordinated control module. The pre-compensation unit receives the deformation prediction value, calculates the associated structural compensation parameters in advance within the preset look-ahead time period, and performs the pre-distortion compensation adjustment on the antenna array. This means that before the physical array surface bends, the phase shifter and amplitude controller have already been parameterized according to the expected bent state.

[0074] However, the prediction model cannot guarantee absolute accuracy under strong noise. Therefore, after the predicted time point is reached and the actual deformation occurs, the system uses the measured deformation data fed back from the four-dimensional perception results to calculate the residual, and uses the residual to generate a correction signal to update the compensation parameters executed in the previous stage with closed-loop error compensation.

[0075] In this process, the residuals and error correction factors are calculated as shown in formula (4):

[0076] (4)

[0077] in, This represents the actual deformation displacement vector of each node obtained in real time through finite element mesh calculation at a specified time point; This represents the predicted deformation displacement vector output by the Long Short-Term Memory Neural Network at a past time point for the current time. This represents the predicted physical displacement deviation vector. The pre-compensation unit uses this physical displacement deviation vector... The mapping is applied as additional phase and amplitude differences, which are superimposed on the current control cycle, thereby transforming the open-loop pre-prediction into a closed-loop dynamic approximation correction, ensuring extremely high compensation accuracy.

[0078] like Figure 4The graph shows the convergence curves of displacement prediction and measured residuals based on a long short-term memory neural network. The horizontal axis represents time in milliseconds, and the vertical axis represents displacement and residuals in millimeters.

[0079] The figure contains three curves: the blue solid line represents the measured deformation displacement obtained by real-time calculation from the finite element mesh; the red dashed line represents the predicted deformation displacement output in advance by the long short-term memory neural network; and the green solid line represents the physical displacement deviation residual obtained by subtracting the predicted displacement from the actual displacement.

[0080] At 10 milliseconds, influenced by sudden external airflow or UAV maneuver commands, the measured deformation displacement exhibits significant fluctuations, with a positive peak value reaching approximately 3.5 millimeters. The predicted deformation displacement curve, represented by the red dashed line, maintains a high degree of correlation with the measured deformation displacement curve during this period, objectively reflecting the neural network's ability to predict the deformation evolution trajectory in advance.

[0081] Meanwhile, the physical displacement deviation residual represented by the green solid line showed an initial error jump of about 0.5 mm in the early stage of the deformation change. However, as the pre-compensation unit used the feedback measured deformation data to calculate the residual and generate a correction signal, the residual curve quickly converged to the zero line in the subsequent 20 to 50 millisecond time interval, and finally stabilized in a very small range of less than 0.1 mm.

[0082] This waveform transformation process, where the residual curve rapidly decays from the initial fluctuations and converges to a near-zero state, intuitively demonstrates the technical effect of the system in updating the closed-loop error compensation parameters of the pre-executed structural compensation. It effectively suppresses the open-loop calculation deviation of the prediction sub-model under dynamic interference environment and ensures that the antenna array can still obtain high-precision pre-distortion control when encountering dynamic bending of the physical array surface.

[0083] The massive matrix operations, federated filtering, and neural network inference described above place stringent demands on the concurrent processing capabilities of the onboard processor. To ensure that the core algorithm is not blocked by the accumulation of operating system tasks, this embodiment employs non-linear management of the allocation of underlying computing resources.

[0084] Specifically, the low-latency parallel processing module includes a deformation-triggered resource scheduling unit. This unit performs the following resource allocation: when the current deformation parameter level is determined to be at or below the slight deformation level, processor computing power resources are evenly allocated according to an initially set ratio. At this time, the system considers the impact of deformation on the beam to be within a margin, and processor resources simultaneously handle routine tasks such as beam scanning, data caching, system log recording, and health status monitoring.

[0085] When the deformation parameter level reaches the moderate deformation level, processor computing resources exceeding the preset proportional threshold are preferentially allocated to the deformation calculation and structural compensation process. During this stage, the system actively suppresses the clock frequency of log read / write or low-priority environmental probing tasks, and assigns all interrupt priorities of the hardware multipliers of the multiply-accumulate units inside the field-programmable gate array (FPGA) and the digital signal processor (DSP) to the deformation data channel.

[0086] When the deformation parameter level reaches the severe deformation level or above, an emergency processing mode is triggered, suspending non-core task links within the system. This is a safety net defense mechanism. When facing extremely high-threat turbulence, the system deprives all task execution permissions unrelated to beam reconstruction of the authority to execute. All computation threads are dedicated to predistortion mapping and the solution of formulas (2) and (3), ensuring that the compensation parameters are output to the underlying physical link within an extremely low latency period.

[0087] For example, when the extreme stress has exceeded the physical limit that a single antenna element can repair through digital phase compensation, the electromagnetic-environment-structure coordinated control module also includes a deformation adaptive multi-subarray coordinated unit, and the flexible antenna array is divided into multiple independent subarrays with independent power control.

[0088] Specifically, when an assessment reveals that an independent subarray has reached the severe deformation level, its radiation impedance matching performance deteriorates drastically due to excessive distortion. Continuing to transmit at full power would result in strong standing wave reflections and could even burn out the back-end RF amplifier. In this case, the deformation-adaptive multi-subarray cooperative unit actively attenuates the transmit power of the independent subarray that has reached the severe deformation level to a preset power attenuation threshold. To compensate for the overall equivalent isotropic radiated power (EIRP) loss of the system, this unit simultaneously increases the transmit power of spatially adjacent normal subarrays for signal coverage compensation. When the ultimate deformation level is reached, the physical material may irreversibly yield. The system disconnects the excitation link of the independent subarray that has reached the ultimate deformation level and re-covers the target area using beamforming algorithms based on the normal independent subarrays. This method of local module shielding maintains the main communication link without interruption.

[0089] It should also be noted that the adaptive control capability of the airborne terminal is always limited by the physical limits of the UAV itself, and a joint defense line with external systems must be established.

[0090] Specifically, the cross-system coordination and anomaly handling module includes an air-to-ground coordinated deformation compensation unit and a deformation anomaly handling unit. The air-to-ground coordinated deformation compensation unit transmits the three-dimensional dynamic deformation field and its predicted data within the system to the ground base station, and receives feedback from the ground base station regarding the transmission power and modulation / coding adjustments. In actual communication networks, the transmission power and antenna aperture of the ground base station are typically much larger than those of the UAV. When the base station receives a high-risk deformation warning from the UAV, it can proactively increase its directional transmission power and reduce the order of the modulation / coding scheme (MCS), for example, from high-order quadrature amplitude modulation to low-order quadrature phase shift keying, sacrificing some data throughput in exchange for a higher received signal-to-noise ratio and improved bit error rate resistance, thus achieving joint updates of beam pointing and link status at both the air-to-ground transceiver ends.

[0091] Simultaneously, when the deformation anomaly processing unit detects that the array is at the irreversible extreme deformation level, it executes the array reconstruction protection mechanism and injects forced deceleration and attitude correction commands into the upper-level UAV flight control system. By directly intervening in the UAV's flight control laws, the UAV is instructed to reduce its flight speed, change its angle of attack, or stop high-overload hovering maneuvers, thereby weakening the aerodynamic load acting on the flexible substrate from the physical source, suppressing further aggravation of physical deformation, and achieving system-level safety protection.

[0092] Based on existing technology analysis, traditional airborne phased array antenna systems often employ rigid arrays, whose size and weight severely limit the payload capacity and flight time of UAVs. Although some existing technologies have proposed the concept of flexible antennas, their beam management mostly remains at the passive static calibration level, assuming that deformation is a slow and predictable single-dimensional change. Once a UAV encounters the dual physical interference of high-frequency fuselage vibration and sudden aerodynamic turbulence during actual operations, traditional control algorithms, lacking multi-dimensional high-frequency separation mechanisms and underlying dynamic resource tilt scheduling, are prone to computational resource congestion and compensation phase calculation overflow, leading to severe beam pointing deviations.

[0093] In contrast, the technical solution disclosed in this embodiment establishes a multi-dimensional joint control architecture covering the electromagnetic domain, structural domain, computing domain, and flight control domain through multi-source sensing of finite element meshes, intelligent advance prediction of time series, dynamic allocation formula of independent weights, and trigger-based suspension mechanism of underlying resources. This solution enables UAVs to replace passive response with forward closed-loop prediction when facing highly dynamic external interference, avoid computing power congestion by tilting resources on demand, significantly reduce the link interruption rate in wide-angle scanning scenarios, and greatly improve the survivability and beam pointing accuracy of millimeter-wave communication systems in complex aviation operating environments.

[0094] Example 2:

[0095] In the highly dynamic flight environment of UAVs, the mechanical stress experienced by flexible antenna arrays is not a single-dimensional static load, but a wideband complex hybrid deformation resulting from the superposition of multiple physical excitation sources. Traditional single-dimensional compensation methods are often limited by the physical response delay of RF hardware or the dynamic range bottleneck of digital baseband, making it difficult to simultaneously address large-scale physical distortions and minute high-frequency jitter. Therefore, this embodiment introduces a cross-domain differentiated compensation technology scheme targeting the deformation characteristics of different frequency bands.

[0096] Specifically, the electromagnetic-environment-structure coordinated control module of the UAV-borne wide-angle scanning millimeter-wave antenna adaptive control system in this embodiment further includes a dual-band cross-domain differential compensation unit. For the low-frequency deformation component, the dual-band cross-domain differential compensation unit extracts the corresponding array curvature change parameters, generates low-frequency spatial compensation parameters, and sends them to the RF phase shifter and amplitude controller to perform beamforming reshaping in the RF physical domain. For the high-frequency deformation component, its high-frequency phase jitter envelope features are extracted, and these features are converted into a digital phase rotation factor and input into the digital signal processor. Inverse phase rotation compensation calculations are then directly performed on the communication data stream in the digital baseband domain. The dual-band cross-domain differential compensation unit relies on a global hardware clock bus to force the activation time of the RF physical domain compensation command to be timestamped and synchronized with the execution time of the digital baseband domain compensation calculation.

[0097] For example, when a drone performs high-altitude communication relay or edge computing node tasks, its onboard flexible antenna array is typically subjected to two distinct types of mechanical disturbances:

[0098] The first category consists of aerodynamic loads caused by high-altitude atmospheric shear winds, gusts, and the drone's own macroscopic maneuvers such as climbs and dives. These aerodynamic loads act on the flexible conformal skin of the drone's wings or fuselage, and their mechanical transmission exhibits a certain degree of inertia and hysteresis, leading to macroscopic physical distortions of the antenna array, primarily characterized by bending and torsion. The physical characteristics of this type of distortion are characterized by large amplitudes, typically reaching millimeter or even centimeter-level displacements, but with relatively low frequencies, generally concentrated in the low-frequency band from 0.1 Hz to 10 Hz. The aforementioned dynamic deformation field calculation unit isolates and defines this as the low-frequency deformation component.

[0099] Regarding the aforementioned low-frequency deformation components, the significant shift in the physical position of the antenna array causes a severe alteration in the spatial aperture projection. Without physical phase reconstruction at the source of the electromagnetic field emission, electromagnetic waves cannot form an effective equiphase surface in free space, resulting in severe beam broadening, sidelobe level elevation, and main lobe gain reduction. This energy dissipation in physical space is irreversible by any subsequent digital baseband algorithm. Therefore, after acquiring the low-frequency deformation components, the dual-band cross-domain differential compensation unit first extracts the corresponding array curvature variation parameters. These parameters contain the three-dimensional spatial offset vector of each antenna element relative to an ideal flat reference plane.

[0100] Specifically, the system generates low-frequency spatial compensation parameters based on the array curvature variation parameters. The essence of this generation process is to calculate the radio frequency phase advance or hysteresis required to compensate for spatial position offset. During this process, the dual-band cross-domain differential compensation unit calculates the compensation value required for each radio frequency channel based on the preset scanning angle of the current beam and the low-frequency physical displacement of each array element. The specific analytical formula for its low-frequency spatial compensation parameters is shown below:

[0101] (5)

[0102] in, The coordinates of the flexible antenna array in the two-dimensional physical coordinate system are: The amount of radio frequency physical phase compensation required for a specific antenna element at a given location; Indicates the radio frequency operating center wavelength used in the current millimeter-wave communication link; This represents the low-frequency physical normal displacement of a specific antenna element in a direction perpendicular to the ideal reference plane, extracted by analyzing the low-frequency deformation components. This indicates the current beam main lobe scanning angle set by the system.

[0103] It should be noted that in practical applications, the calculated values... The signal is then routed to the RF phase shifter and amplitude controller to perform beamforming reshaping in the RF physical domain. Current RF front-ends typically employ multi-bit (e.g., 6-bit or 8-bit) numerically controlled microelectromechanical systems (MEMS) phase shifters or voltage-controlled semiconductor phase shifters. These RF physical devices require a Serial Peripheral Interface (SPI) bus to receive digital control words and drive internal physical circuitry, such as capacitor charging / discharging or diode bias state switching, to complete phase switching. The entire hardware-level response settling time is typically hundreds of nanoseconds to several microseconds, with inherent physical response delay limitations. Because the rate of change of low-frequency deformation components is much lower than the hardware settling rate of the RF phase shifter, the RF physical domain can fully and smoothly track and compensate for this macroscopic array curvature change, achieving beam energy refocusing at the physical space level.

[0104] like Figure 5 The graph compares the far-field radiation patterns after low-frequency deformation distortion and RF spatial compensation. The horizontal axis represents angle in degrees, and the vertical axis represents normalized gain in decibels.

[0105] The figure contains three curves. The blue solid line represents the radiation pattern of an ideal flat front surface, where the main lobe energy of the beam is highly concentrated in the zero-degree direction and the normalized gain reaches the reference peak of 0 dB.

[0106] The red dotted lines represent the low-frequency bending distortion pattern, reflecting the radiation state of the flexible antenna array after macroscopic physical distortion caused by aerodynamic loads such as high-altitude shear winds. At this time, the main lobe of the beam undergoes significant broadening, the peak gain of the main lobe drops sharply by about 4 dB, and the sidelobe level rises abnormally to about -12 dB. This directly reveals the risk of electromagnetic energy spatial dispersion and severe gain attenuation caused by the physical position shift of the array.

[0107] The green dashed line represents the radiation pattern after RF spatial compensation. It corresponds to the dual-band cross-domain differential compensation unit extracting array curvature change parameters and generating low-frequency spatial compensation parameters, which are then sent to the underlying RF phase shifter and amplitude controller to perform beamforming reshaping of the radiation state. After spatial phase modulation in the RF physical domain, the beam shape of the green dashed line converges again, the main lobe gain significantly recovers and approaches the ideal peak value of 0 dB, while the sidelobe level is also effectively suppressed.

[0108] This waveform transformation process, from beam widening and energy dissipation to the precise reconvergence of main lobe energy, objectively demonstrates that the cross-domain differentiated compensation architecture of this system can offset the negative impact of macroscopic bending of the flexible array from the source of physical electromagnetic field emission, ensuring that the airborne wide-angle scanning millimeter-wave antenna still has high fixed-point gain and extremely strong spatial pointing stability in complex and ever-changing aerodynamic interference environments.

[0109] Optionally, the second type of mechanical disturbance originates from the mechanical excitation of the UAV's propulsion system. When a UAV relies on a multi-rotor system to maintain hovering or high-speed propulsion, the high-speed rotating brushless motor and propeller blades inevitably generate broadband mechanical vibration waves. These vibration waves are transmitted to the flexible antenna array through the UAV's airframe, causing each antenna element to experience weak high-frequency jitter near its equilibrium position. The physical characteristics of this type of distortion are extremely small deformation amplitude, typically at the micrometer level, but extremely high vibration frequency, generally concentrated in the high-frequency band of 100 Hz to 2000 Hz. The aforementioned dynamic deformation field calculation unit isolates and defines this as the high-frequency deformation component.

[0110] To address the aforementioned high-frequency deformation components, attempting to use the aforementioned radio frequency (RF) phase shifter for real-time tracking compensation will face severe technical obstacles. High-frequency vibrations cause the required compensation phase to fluctuate wildly in milliseconds or even microseconds. Forcing the RF phase shifter to frequently switch its internal circuit states at extremely high frequencies would, on the one hand, cause severe congestion of the underlying communication bus due to a large number of serial control commands; on the other hand, the transient settling time of the RF phase shifter during frequent switching would increase dramatically, resulting in a significant amount of parasitic amplitude and phase modulation noise mixed into the RF output signal, which would severely degrade the error vector amplitude (EVM) of the communication signal.

[0111] To overcome this hardware challenge, the dual-band cross-domain differential compensation unit abandons the approach of tracking high-frequency jitter in the radio frequency physical domain, and instead extracts its high-frequency phase jitter envelope features. Although micrometer-level high-frequency physical displacement cannot significantly change the macroscopic beam spatial direction, it directly affects the shorter wavelength of millimeter waves, causing rapid pseudo-random jitter in the carrier phase of the transmitted signal. In communication principles, this is equivalent to introducing strong Doppler spread interference. The system converts this high-frequency phase jitter envelope feature into a digital phase rotation factor and inputs it into a digital signal processor.

[0112] Specifically, in the digital baseband domain, the communication data stream has not yet been up-converted to the radio frequency band by a digital-to-analog converter (DAC). Digital signal processors (DSPs) or field-programmable gate arrays (FPGAs) typically operate at clock frequencies of hundreds of megahertz, capable of processing massive multiply-accumulate operations within a single clock cycle. The system directly performs inverse phase rotation compensation operations on the communication data stream in the digital baseband domain. The analytical formula for rotation correction in the digital baseband layer is shown below:

[0113] (6)

[0114] in, Indicates discrete-time index in the baseband digital domain The complex baseband data symbols output after high-frequency compensation processing; Indicates discrete-time index in the baseband digital domain The raw input complex baseband data symbols are passed from the upper Media Access Control (MAC) layer to the Physical (PHY) layer for transmission. The imaginary unit; This represents the discrete-time index in the baseband digital domain based on the high-frequency phase jitter envelope characteristics. The digital phase rotation factor calculated at each moment has a sign that is strictly opposite to the sign of the additional physical phase caused by high-frequency physical displacement.

[0115] For example, through formula calculation, the digital signal processor utilizes its internal coordinate rotating digital computer (CORDIC) algorithm core to directly apply a reverse rotation bias to each digital baseband symbol. This mathematical operation in the digital domain does not rely on any physical RF devices with mechanical inertia or analog circuit delays, and its compensation rate is equivalent to the sampling rate of the baseband signal, for example, up to 100 Msps or more. This allows it to easily and with extremely high precision track and eliminate the phase noise caused by mechanical high-frequency jitter of several kilohertz, ensuring the convergence of the constellation diagram of high-speed communication modulation signals (such as 256-QAM high-order quadrature amplitude modulation).

[0116] like Figure 6 This is a comparison chart of the 256-QAM communication constellation before and after high-frequency phase jitter compensation. The horizontal axis represents in-phase amplitude, and the vertical axis represents quadrature amplitude. The chart contains two different colored scatter plots: red scatter plots represent constellation points with high-frequency jitter distortion, and blue scatter plots represent constellation points after digital baseband compensation.

[0117] Under the influence of high-frequency, weak physical jitter caused by the high-speed rotation of the UAV rotor and engine vibration, the millimeter-wave signal carrier generates rapid pseudo-random phase fluctuations. Reflected on the red scatter dots, the originally compactly arranged square array exhibits a significant annular angular dispersion phenomenon. In particular, the high-amplitude constellation points located in the outer region are prone to having their arc-shaped dispersion trajectories break through the decision boundary and overlap with each other, resulting in a significant deterioration of the error vector amplitude of the communication signal and increasing the risk of demodulation errors.

[0118] The blue scatter dots objectively demonstrate the actual control effect of the system extracting high-frequency phase jitter envelope features and converting them into digital phase rotation factors, and then directly performing inverse phase rotation compensation operations on the communication data stream in the digital baseband domain. After the intervention of the baseband digital domain, the originally divergent red arc trajectory converges again, restoring to a clearly defined and highly clustered blue dot matrix, and sufficient safety decision margins are regained between each constellation point.

[0119] This intuitive graphical comparison of the scattered distribution state from divergence to close convergence confirms that the cross-domain differentiated compensation architecture of this system can circumvent the hardware limitations of slow response of radio frequency physical devices, and effectively suppress phase interference caused by mechanical high-frequency jitter at the kilohertz level by utilizing the high efficiency of digital signal processing. This effectively ensures the convergence of the constellation diagram and the reliability of data link transmission of high-order modulation signals in complex high-dynamic aviation operating environments.

[0120] It is also important to note that the above scheme splits the composite deformation of the same physical flexible antenna array into two distinct system levels—the RF physical domain and the digital baseband domain—for asynchronous processing. In actual high-speed digital communication systems, the RF control link and the digital baseband processing link have completely different internal data pipeline structures and instruction execution cycles. If these two compensation mechanisms operate independently, it will cause serious cross-domain timing misalignment problems. For example, the digital baseband domain may have already applied a rotation factor for a high-frequency vibration state at a certain instant, but at this time, the adjustment command for the low-frequency state of the RF phase shifter is still queued on the serial bus and has not yet taken effect. This timestamp mismatch will cause the spatial state of digital rotation and physical phase to be misaligned and interfered with on the time axis. Not only will it fail to compensate for the error, but it will also cause the signal-to-noise ratio of the communication link to drop below the demodulation threshold instantaneously.

[0121] To this end, the dual-band cross-domain differential compensation unit relies on a global hardware clock bus to force the activation time of the RF physical domain compensation instruction to be timestamped and synchronized with the execution time of the digital baseband domain compensation operation. The global hardware clock bus obtains the pulse-per-second (1PPS) and high-frequency reference clock from a high-precision temperature-compensated crystal oscillator or satellite navigation timing module, and constructs a clock distribution network with equal-length traces, distributing them to all RF microcontrollers and baseband signal processor nodes.

[0122] Specifically, the system internally sets up a unified compensation triggering logic to ensure that the low-frequency physical space shaping and the high-frequency baseband data rotation are at the same time cross-section at the point where the electromagnetic waves converge at the antenna radiation port. The timing constraint formula for its synchronization alignment is shown below:

[0123] (7)

[0124] in, This indicates the unified absolute execution timestamp of the RF phase shifter's underlying hardware register latch taking effect and the digital baseband performing phase rotation operations on specific data packets after forced alignment and synchronization; This represents the current system absolute timestamp when the dual-band cross-domain differential compensation unit finishes the operation of the correlation matrix between the low-frequency spatial compensation parameters and the digital phase rotation factor; This represents the known inherent RF hardware transmission and setup delay period required for an RF control command to be issued and for the internal analog circuitry of the RF phase shifter to stabilize. This represents the known inherent baseband pipeline delay period required for the digital phase rotation factor to be injected into the baseband processing pipeline until the corresponding data symbol leaves the baseband chip pin. This represents the reserved protection margin time difference set to absorb bus clock jitter and temperature drift.

[0125] For example, in practical applications, due to the setup delay of the radio frequency physical devices... Typically greater than the processing latency of digital baseband. The global hardware clock bus scheduling mechanism requires that the digital baseband processor, after completing parameter generation, not immediately apply the latest rotation factor. Instead, it pushes the data stream to be sent, along with the new rotation factor, into a high-speed synchronous first-in-first-out (FIFO) buffer queue for manual delay. Only when the system time counter reaches the unified absolute execution timestamp... Only when the low-frequency beamforming reshaping in the radio frequency physical domain is confirmed to be established on the physical antenna surface will the baseband data in the buffer queue be synchronously released and superimposed with a high-frequency rotation factor for transmission. This highly rigorous cross-domain synchronization mechanism ensures high-precision time synchronization of software and hardware compensation operations.

[0126] Based on existing technologies, traditional UAV-borne antenna attitude compensation or phased array calibration schemes either rely solely on physical compensation using phase-shifting networks in the radio frequency domain or entirely on digital compensation using channel equalizers in the baseband domain. Pure radio frequency compensation schemes face hardware bottlenecks such as slow device response and inability to handle high-frequency rotor vibrations; while pure baseband digital compensation schemes are limited by a finite dynamic range. When encountering significant UAV maneuvers or severe array bending caused by strong winds, the main lobe of the beam shifts excessively in physical space, leading to a sharp attenuation of the receiving front-end energy, which is easily submerged in the system noise floor, exceeding the limits of baseband digital algorithms.

[0127] In contrast, the technical solution disclosed in this embodiment, which includes a dual-band cross-domain differential compensation unit, possesses extremely strong engineering applicability and system robustness. Starting from a physical mechanism, this solution decouples complex hybrid disturbances into low-frequency array curvature changes and high-frequency phase jitter envelopes. By entrusting large-amplitude low-frequency deformations to the RF physical domain for macroscopic beam reshaping to ensure basic link energy, and entrusting extremely high-frequency minute jitters to the digital baseband domain for microscopic phase inversion to ensure a high signal-to-noise ratio, it achieves microsecond-level timestamp synchronization fusion at the underlying global hardware clock bus.

[0128] This differentiated compensation architecture, characterized by hardware-software collaboration and cross-domain division of labor, effectively avoids the inherent performance limitations of a single control domain. While preventing control overload of the RF hardware, it significantly suppresses modulation errors introduced into the communication link by complex mechanical deformation. In actual high-dynamic UAV inspection, rescue, or air-to-ground broadband data transmission operations, this system ensures that the wide-angle scanning millimeter-wave beam is not only spatially accurate but also has stable link gain, significantly reducing the bit error rate and link interruption probability under flight flutter and atmospheric disturbance environments.

[0129] Example 3:

[0130] In the actual high-altitude operation environment of UAVs, airflow disturbances often exhibit strong irregularity, suddenness, and high-frequency oscillations. When the UAV is in the edge region of complex turbulence, or encounters gusts of wind shear between tall buildings and canyons, the aerodynamic load on the flexible antenna array will fluctuate violently in a short period of time. This boundary oscillation of the physical environment will directly cause the deformation parameter level extracted by the system to generate high-frequency cross-boundary fluctuations near the critical threshold.

[0131] If the system uses conventional single-threshold discrete triggering logic to allocate computing resources, the main control chip will get stuck in a high-frequency resource reallocation loop, causing severe scheduling congestion and loss of non-core task data. Therefore, this embodiment constructs an underlying system resource protection architecture based on asymmetric hysteresis decision-making and zero-copy task suspension.

[0132] Specifically, in the deformation-triggered resource scheduling unit of the UAV-borne wide-angle scanning millimeter-wave antenna adaptive control system in this embodiment, there is an embedded anti-shake hysteresis scheduling submodule and a pointer-level task snapshot submodule. The system is also equipped with a trigger threshold and a fallback threshold, and the fallback threshold is strictly less than the trigger threshold.

[0133] When the deformation parameter level exceeds the trigger threshold from low to high, the anti-jitter hysteresis scheduling submodule performs an escalation allocation of computing resources to core tasks; when the deformation parameter level falls back from high to low, it initiates a dynamic stabilization time window. The duration of the dynamic stabilization time window is calculated by positive adaptive scaling based on the variance of airflow disturbance within the historical sampling period. Only when the deformation parameter level remains below the fallback threshold for the entire dynamic stabilization time window is the degraded recovery of computing resources to non-core task links performed.

[0134] When the pointer-level task snapshot submodule triggers the emergency processing mode and requires suspending non-core task links, it starts a hardware-level interrupt interception processing flow. While keeping the physical address of the original memory data block unchanged, it extracts the program counter and register context state of the suspended task to generate a snapshot pointer and pushes the snapshot pointer into a dedicated static random access memory array for locking. When the degradation recovery conditions are met, it extracts the snapshot pointer from the memory array and performs lossless breakpoint recovery based on the snapshot pointer.

[0135] It is important to note that conventional single-threshold comparators are prone to generating high-frequency flip pulses at their output when the input signal is accompanied by random noise or environmental fluctuations. In computer operating system science, this phenomenon is known as the ping-pong effect or context switching thrashing. In UAV-borne antenna control systems, if the deformation parameter level fluctuates back and forth at a frequency of 100 Hz on the physical boundary between moderate and severe deformation, the system's task scheduler will continuously preempt and restore CPU time slices for non-core tasks (such as environmental detection caching and downlink system log packaging) at the same frequency. Each time slice preemption and restoration is accompanied by huge system bus addressing and memory movement overhead. This not only wastes valuable computing power that should be used for beam alignment and predistortion compensation, but also causes unpredictable jitter deviations in the system's internal clock synchronization.

[0136] For example, to address the aforementioned resource scheduling congestion problem, the anti-jitter hysteresis scheduling submodule introduces asymmetric hysteresis decision logic. The system hard-codes or dynamically writes two discrete numerical boundaries with a certain difference into its internal control register: the trigger threshold and the fallback threshold. Furthermore, the system's underlying logic relies on a hardware comparator to ensure that the fallback threshold is strictly less than the trigger threshold. This dual-threshold inequality design constructs a hysteresis interval immune to fluctuations.

[0137] Optionally, when a drone encounters a sudden strong gust of wind, causing the deformation parameter level to rise rapidly and exceed the trigger threshold from low to high, the system determines that the current antenna array faces an imminent beam shift threat. At this time, the anti-shake hysteresis scheduling submodule performs an escalation-level allocation of computing resources to core tasks. In specific applications, this escalation-level allocation manifests as the system task scheduler immediately increasing the thread priority of deformation calculation, electromagnetic predistortion mapping table lookup, and dual-band cross-domain differential compensation processes to the highest level, while forcibly releasing the computing cores in the multi-core processor originally used for background tasks and allocating them to the aforementioned core control tasks. This low-to-high determination process has extremely low latency, ensuring the system's agility in the face of deteriorating physical environments.

[0138] Specifically, when the external gusts weaken and the deformation parameter level drops from high to low and falls below the drop threshold for the first time, the anti-jitter hysteresis scheduling submodule does not immediately release the tilted computing resources. This is because in complex aviation meteorological environments, a brief weakening of wind force is often just a gap before the next stronger turbulence erupts. At this time, the module activates a dynamic stabilization time window. To ensure that this time window accurately reflects the degree of external airflow turbulence, the system continuously collects and calculates the airflow disturbance variance.

[0139] In this process, the variance of airflow disturbance is calculated as shown in formula (8):

[0140] (8)

[0141] in, It represents the variance of airflow disturbance within a specified sliding observation period, used to quantify the dispersion and intensity of external physical airflow. This represents the total number of airflow velocity samples acquired during the sliding observation period; This indicates the discrete time points extracted by the aforementioned airflow disturbance sub-filter in the airflow sensor sampling domain. The instantaneous relative velocity of the physical airflow at the location; Indicated in the The arithmetic mean of the relative airflow velocity within each sampling period. The practical function of this formula (8) is that, through statistical variance calculation, the system can accurately identify whether the UAV is currently in a stable laminar flow environment or in a turbulent boundary layer with drastic wind speed changes.

[0142] Furthermore, the duration parameter of the dynamic stabilization time window is calculated based on the positive adaptive scaling of the variance of airflow disturbance within the historical sampling period, and the analytical generation of its duration parameter is shown in formula (9):

[0143] (9)

[0144] in, This represents the actual physical duration of the dynamic stability time window calculated by the system. This represents the system's preset basic stability waiting time constant, which characterizes the shortest safe observation period required to confirm the fallback of the deformation state under ideal and stable airflow conditions. This represents the variance-sensitive scaling factor obtained through calibration experiments, used to linearly map the square dimension of velocity to the time dimension.

[0145] The practical significance of formula (9) lies in the fact that when the external airflow is extremely turbulent, i.e., the variance is extremely large, the system will automatically and significantly extend the observation and waiting time, forcibly maintaining a high-alert computing resource tilt state; while when the external airflow returns to stability, i.e., the variance approaches zero, the observation and waiting time is reduced to a basic constant, enabling the system to resume normal scheduling as soon as possible. Only when the deformation parameter level remains below the fallback threshold for an extended period within the complete dynamic stability time window will the computing resources be degraded and restored to non-core task links. This design eliminates scheduling oscillations from the root of the control logic.

[0146] like Figure 7 This is a timing diagram of the state transition of asymmetric hysteresis scheduling under complex airflow disturbances. The diagram contains three sub-diagrams sharing a horizontal time axis, with the time unit being seconds.

[0147] From top to bottom, the first subplot's vertical axis represents the variance of airflow disturbance, and it contains a magenta variance curve. The second subplot's vertical axis represents the deformation parameter level, and it contains a blue deformation level curve, a red dashed line indicating the trigger threshold, and a green dashed line indicating the fallback threshold. The trigger threshold is set to a value of 3 (corresponding to a severe deformation level), and the fallback threshold is set to a value of 2 (corresponding to a moderate deformation level). The third subplot's vertical axis represents the resource allocation status, and it contains a black state step curve. A state value of 1 represents tilted allocation, and a state value of 0 represents normal allocation.

[0148] At 3.1 seconds, affected by a sudden strong gust of wind, the blue deformation level curve of the middle subgraph suddenly rose and crossed the red trigger threshold. At this time, the black state step curve of the bottom subgraph showed a jump with extremely low latency, instantly jumping from the normal allocation to the tilted allocation state, indicating that the system rapidly tilted computing resources to the core task.

[0149] Subsequently, around 5 seconds later, the blue deformation level curve dropped and fell below the green drop threshold for the first time. However, at this time, the magenta variance curve of the top subgraph was at a high level. Based on this, the system adaptively extended the dynamic stabilization time window. During the waiting period, the deformation level curve fluctuated and rebounded again around 7 seconds. Therefore, the bottom state step curve remained in the tilted allocation state, avoiding premature release of computing resources.

[0150] Around 10 seconds later, the blue deformation level curve fell back below the green threshold, and the variance curve at the top had decreased significantly. The system shortened the dynamic stabilization time window accordingly. After the blue curve smoothly passed through the time window, the bottom state step curve accurately fell back to the normal allocation state at 11.5 seconds.

[0151] This dynamic judgment and state delay step process, reflected by the synergy of multiple curves, objectively verifies that the inequality dual threshold design and variance adaptive time window mechanism adopted by the system can effectively filter scheduling oscillations caused by high-frequency mechanical turbulence and significantly reduce the latency overhead of the underlying operating system frequently switching task contexts in critical states.

[0152] It's important to note that when the system must enter emergency mode and suspend non-core tasks due to exceeding a trigger threshold, traditional operating system suspension mechanisms also face significant performance bottlenecks. When suspending an executing task, the traditional mechanism requires kernel intervention to copy and push the entire memory data segment, stack space, and numerous variable entities of the task into a specific reserved area of ​​external dynamic random access memory (DRAM). Due to the limited read / write bus width and capacitor refresh latency of external DRAM, this massive data transfer process can consume tens or even hundreds of microseconds. In critical situations like high-speed drone flight where beam reconstruction requires microsecond-level precision, such low-level memory copying is unacceptably time-consuming.

[0153] For example, to minimize the time consumption caused by memory movement, the pointer-level task snapshot submodule initiates a hardware-level interrupt interception processing flow when an emergency processing mode is triggered and non-core task chains need to be suspended. At the hardware architecture level of the embedded microprocessor, hardware-level interrupt requests (IRQs) have a physical response priority that overrides all software scheduling algorithms. Within a few clock cycles of a hardware-level interrupt being triggered, the underlying machine instruction pipeline currently being executed by the processor's arithmetic logic unit (ALU) will be forcibly stopped.

[0154] Specifically, after stopping the processing flow, the submodule extracts the program counter and register context state of the suspended task to generate a snapshot pointer, while keeping the original physical address of the memory data block unchanged. This means that the system does not move any actual business data payload; for example, the unprocessed environmental awareness raw matrix data remains unchanged in its original static storage unit mapping area. Instead, it only extracts and saves the processor's current running context state. This state information includes the program counter indicating the physical address of the memory of the next instruction to be executed, as well as general-purpose registers and status registers storing local intermediate calculation results.

[0155] In this process, the generation structure and synthesis of the snapshot pointer are as shown in formula (10):

[0156] (10)

[0157] in, This represents the hash code of the fixed-length, fixed-width snapshot pointer generated by the submodule. This represents a non-cryptographic, high-efficiency hash mapping algorithm function that is embedded in the underlying system. This represents the unique identifier of a currently forcibly suspended non-core task in the operating system's process table. This indicates the physical address status value of the program counter at the moment of the current processor's internal suspension; This represents the data stream containing all general-purpose registers and carry / over flag status registers in the current processor core; This represents the concatenation operator for binary bit streams. The practical function of this formula is to generate a tiny, typically tens of bytes, identifier by concatenating core state data and performing hash mapping, which uniquely identifies the execution context of a processor breakpoint.

[0158] Optionally, the system pushes the snapshot pointer into a dedicated static random access memory array for locking. Compared to dynamic random access memory, which requires continuous periodic refreshing, the on-chip integrated static random access memory (SRAM) relies on flip-flop circuits to latch data. Its access latency is the same as the processor's core operating frequency, and it can complete data writing within a single instruction cycle. Since only a tiny snapshot pointer synthesized according to formula (10) is pushed in, this suspension action can be declared completed within a few nanosecond instruction cycles, thereby compressing the originally long task switching time to the limit. This is called a zero-copy pointer-level suspension mechanism. Subsequently, the processor is quickly cleared and released, and put into emergency beam compensation calculation in a state of extremely low latency.

[0159] Specifically, when external interference subsides and the system meets the degradation recovery conditions, i.e., the deformation level has smoothly passed the dynamic time window set by formula (9), the processor's computing power resources are allowed to be reallocated to background processes. At this time, the submodule retrieves a snapshot pointer from the memory array and performs lossless breakpoint recovery based on that snapshot pointer. The system pops the previously saved snapshot from the static random access memory. The program counter is obtained through reverse analysis. The address pointer is redirected back to the original physical memory execution segment before the interrupt occurred, and the register set is... The kernel is rewritten. Since the physical addresses of all background raw data have not been moved or overwritten during the suspension, non-core tasks can seamlessly resume their previous workflows within microseconds. This not only ensures the high real-time requirements of the communication baseband and servo control, but also guarantees the integrity of low-priority data such as environmental monitoring and flight logs.

[0160] Based on analysis of existing technologies, current airborne antenna control equipment often employs a general time-sharing operating system or a simple polling control mechanism for system task management. While this general scheduling mechanism can maintain system operation in mild, static environments, in specific high-dynamic aviation scenarios, when the fuselage experiences severe physical turbulence and aerodynamic deformation, massive compensation requests and sensor data can overwhelm the general operating system instantly. Due to the lack of dynamic anti-shake design in the time domain, existing technologies are prone to repeated context switching and memory swapping during task priority scheduling. This results in the crucial antenna compensation parameters being delayed due to internal computing power idleness, ultimately causing beamout and link interruption in millimeter-wave communication systems, which are extremely sensitive to phase.

[0161] In contrast, the technical solution disclosed in this embodiment filters out scheduling oscillations at the mathematical decision level by combining anti-jitter hysteresis scheduling with airflow variance adaptive scaling time windows; and eliminates system latency of context switching at the underlying memory access level by introducing a hardware-software combined zero-copy task snapshot mechanism. This overall solution enables the UAV-borne wide-angle scanning system to achieve efficient, accurate, and low-overhead tilting of computing resources when facing unpredictable high-frequency mechanical turbulence, significantly improving the equipment's anti-interference capability and resource utilization in harsh aviation weather environments, and ensuring reliable operation of high-bandwidth air-to-ground communication around the clock.

[0162] Example 4:

[0163] like Figure 8 As shown, this embodiment provides an adaptive control method for a wide-angle scanning millimeter-wave antenna on a UAV. This method aims to solve the technical problem of unstable communication links caused by multiple factors such as fuselage vibration, aerodynamic loads, and environmental attenuation in complex aviation operating environments.

[0164] In practical applications, this method relies on high-performance computing hardware, flexible antenna arrays, and multi-dimensional sensor integration systems deployed on UAV platforms. The method flow in this embodiment covers a complete logical closed loop, from raw data acquisition, intelligent prediction, resource scheduling to high-precision compensation execution and system-level anomaly protection, ensuring the pointing accuracy and gain stability of the millimeter-wave beam during wide-angle scanning.

[0165] Specifically, the adaptive control method for a wide-angle scanning millimeter-wave antenna on a UAV according to this embodiment includes the following steps:

[0166] First, the system uses a sensor network integrated on the UAV fuselage and flexible antenna substrate to collect platform attitude data, airflow disturbance data, environmental attenuation data, and three-dimensional dynamic deformation data of the flexible array. In actual operation scenarios, the platform attitude data is provided in real time by the inertial measurement unit (IMU), including the UAV's heading angle, pitch angle, and roll angle; the airflow disturbance data is acquired by an onboard miniature radar or barometer array to characterize the intensity of atmospheric turbulence in the space where the UAV is located; the environmental attenuation data is mainly obtained by monitoring the signal-to-noise ratio fluctuation and received signal strength (RSSI) of the current communication link to reflect the absorption of millimeter-wave energy by atmospheric molecules or the dissipation of energy by rainfall; the three-dimensional dynamic deformation data of the flexible array is collected in real time by distributed microelectromechanical systems (MEMS) strain sensors embedded inside the antenna substrate to accurately characterize the physical shape deviation of the flexible array under aerodynamic forces.

[0167] After the aforementioned multi-source heterogeneous data acquisition is completed, the system performs data fusion through a filtering unit to output four-dimensional perception results. The filtering unit eliminates the measurement noise and drift inherent in single-type sensors through Kalman filtering or complementary filtering algorithms, enabling the final generated four-dimensional perception results to fully reflect the complex characteristics of the UAV's current spatial attitude, airflow state, environmental links, and structural deformation under a unified timestamp.

[0168] Secondly, to overcome the lag between physical compensation actions and environmental changes, this method involves outputting predicted deformation values ​​in advance based on UAV maneuver commands and historical deformation data for pre-compensation. In practical applications, maneuver commands issued by the UAV flight control system, such as sharp turns, climbs, or accelerations, are the preceding causes of array deformation. This method analyzes the expected mechanics brought about by the maneuver commands, combines them with historical deformation sequences from several past sampling periods, and uses a time-series analysis model to predict the deformation trend within milliseconds. The obtained predicted deformation values ​​are transmitted to the compensation execution end in real time, enabling the antenna control system to pre-adjust the excitation parameters of each antenna element before the physical deformation is fully formed.

[0169] It is also important to note that computational efficiency and real-time performance are the core bottlenecks when dealing with complex millimeter-wave beamforming algorithms. Therefore, this method includes dynamically allocating processor computing resources based on the extracted deformation parameter levels using a deformation-triggered resource scheduling mechanism to output compensation parameters corresponding to pre-distortion compensation. The system automatically classifies deformation parameter levels based on the magnitude of the deformation in the four-dimensional perception results. When the deformation is within a small range, the processor (such as an airborne FPGA or DSP) only allocates basic computing power for maintenance; when the deformation level reaches preset medium or high thresholds, the scheduling mechanism actively suspends non-core task links and fully allocates the instruction cycles of the core computing units to the deformation compensation logic. This dynamic scheduling ensures that the system can still output accurate compensation parameters with extremely low latency even under extreme flight conditions.

[0170] Specifically, after obtaining the sensing data and computing power support, the method enters the core compensation stage, that is, based on the four-dimensional sensing results, the electromagnetic inherent gain loss, environmental gain loss and structural gain loss are separated, the compensation factor weights are allocated according to the loss ratio, and the pre-compensation and pre-distortion compensation are performed in combination with the mapping table parameters and the deformation prediction values ​​to generate the final amplitude and final phase of the antenna excitation.

[0171] In practice, the inherent electromagnetic gain loss is mainly caused by the reduction in effective aperture during wide-angle scanning; the environmental gain loss originates from the dielectric loss during path propagation; and the structural gain loss is caused by the phase center shift due to the physical deformation of the flexible array. After quantizing and separating these three factors, the system dynamically adjusts the compensation factor weights based on the current dominant loss factor.

[0172] For example, when strong winds cause severe bending of the array, the proportion of structural compensation weights will increase significantly. The system then retrieves the built-in pre-distortion mapping table, which stores electromagnetic calibration data under different scanning angles and deformation states. Combining the deformation prediction values ​​obtained from the previous steps, the system uses complex weighted operations to finally generate the final amplitude and phase of each element antenna excitation at the RF front end. This process not only corrects the static deviations that have already occurred, but also offsets the dynamic trend errors through pre-compensation, ensuring that the equiphase surface of the synthesized beam always maintains its optimal distribution in free space.

[0173] Finally, this method provides system-level safety redundancy through cross-system collaboration. Specifically, it involves real-time monitoring of the rate of change of the deformation parameters, triggering array reconstruction when the deformation parameters exceed the limit threshold, and coordinating with the UAV flight control system and the ground base station to adjust the attitude. When the deformation rate or amplitude of the flexible array exceeds the physical range that electromagnetic control can correct, i.e., when the limit threshold is reached, the system will proactively cut off the power supply to the damaged subarray or adjust the excitation distribution, performing array reconstruction to avoid physical damage to the hardware. Simultaneously, the control system sends an attitude intervention request to the UAV flight control system, instructing the UAV to reduce the aerodynamic loads acting on the array surface by changing the angle of attack or decelerating; at the same time, it notifies the ground base station to adjust the transmission power or beamwidth. Through closed-loop collaboration between the air and ground ends, the continuity of the millimeter-wave communication link is guaranteed to the greatest extent in extreme environments.

[0174] As can be seen from existing technologies, traditional control methods often treat the antenna as a rigid whole, ignoring the dynamic coupling relationship between UAV fuselage deformation and environmental losses. This leads to beam pointing drift or even disconnection under strong interference environments. The method provided in this embodiment, however, achieves a technological leap from single electromagnetic calibration to multi-domain fusion control through deep perception of four-dimensional features, intelligent prediction of future trends, dynamic allocation of computing resources, and cross-system collaborative protection.

[0175] In practical applications, this method significantly reduces the bit error rate of UAVs under high-dynamic flight conditions, greatly improves the adaptability and anti-interference robustness of millimeter-wave communication systems, and provides solid technical support for long-endurance, high-bandwidth remote operations of UAVs.

[0176] Example 5:

[0177] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0178] like Figure 9 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0179] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0180] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0181] The memory 103 stores a computer program corresponding to the adaptive control method for a wide-angle scanning millimeter-wave antenna on a UAV according to the above embodiments of the present invention. This computer program is executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0182] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 9 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0183] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An adaptive control system for a wide-angle scanning millimeter-wave antenna on a UAV, characterized in that, include: The multi-source heterogeneous sensing module is used to collect platform attitude data, airflow disturbance data, environmental attenuation data and flexible array three-dimensional dynamic deformation data, and to perform data fusion through the filtering unit to output four-dimensional sensing results. The electromagnetic-environment-structure coordinated control module is connected to the multi-source heterogeneous sensing module. It is used to separate the electromagnetic inherent gain loss, environmental gain loss and structural gain loss according to the four-dimensional sensing results, allocate compensation factor weights according to the loss ratio, and perform pre-distortion compensation in combination with the mapping table parameters to generate the final amplitude and final phase of the antenna excitation. A low-latency parallel processing module, connected to the collaborative control module, is used to dynamically allocate processor computing resources according to the extracted deformation parameter level using a deformation-triggered resource scheduling mechanism, and output the compensation parameters corresponding to the pre-distortion compensation. A multi-dimensional intelligent learning module is used to output deformation prediction values ​​in advance based on UAV maneuver commands and historical deformation data, and send them to the collaborative control module for pre-compensation. The cross-system collaboration and anomaly handling module is used to monitor the rate of change of deformation parameters in real time. When the deformation parameters exceed the limit threshold, it triggers array reconstruction and coordinates the UAV flight control system and ground base station to adjust the attitude.

2. The system according to claim 1, characterized in that, The multi-source heterogeneous sensing module includes a flexible sensing-communication integrated substrate and a distributed strain sensor network. The flexible sensing-communication integrated substrate is divided into multiple finite element meshes, each finite element mesh corresponds to a millimeter-wave communication antenna element, and the vertex of each finite element mesh is arranged with the sensing node of the distributed strain sensor network. The filtering unit is a federated extended Kalman filter unit, which includes a platform disturbance sub-filter for fusing inertial navigation and magnetometer data, an airflow disturbance sub-filter for fusing radar micro-motion data, an environmental perception sub-filter for extracting attenuation data, and a structural deformation sub-filter for eliminating strain measurement noise. The four sub-filters output in parallel and are synthesized by the main filter into the four-dimensional perception result.

3. The system according to claim 2, characterized in that, The multi-source heterogeneous sensing module also includes a dynamic deformation field calculation unit. The dynamic deformation field solution unit is based on the linear elasticity finite element method, which converts the measurement data collected by the distributed strain sensor network into the three-dimensional displacement and rotation angle of each finite element mesh node, and reconstructs the three-dimensional dynamic deformation field. The three-dimensional dynamic deformation field and platform vibration data are correlated and separated in the time and frequency domains to analyze the high-frequency deformation components caused by fuselage vibration and the low-frequency deformation components caused by aerodynamic loads.

4. The system according to claim 1, characterized in that, The electromagnetic-environment-structure coordinated control module has a built-in five-dimensional step-by-step predistortion mapping table. The query dimensions of this mapping table consist of scanning angle, excitation amplitude, excitation phase, electromagnetic performance attenuation, and deformation parameter level. The deformation parameter levels are discretized from low to high according to the deformation diffraction index into no deformation level, slight deformation level, moderate deformation level, severe deformation level and extreme deformation level. The five-dimensional step-by-step predistortion mapping table records the electromagnetic amplitude compensation value, environmental amplitude compensation value, structural amplitude compensation value, electromagnetic phase compensation value, environmental phase compensation value, and structural phase compensation value to be extracted under the corresponding scanning angle range and deformation parameter level.

5. The system according to claim 4, characterized in that, The calculation steps for pre-distortion compensation performed by the electromagnetic-environment-structure coordinated control module are as follows: Calculate the percentage of the structural gain loss in the total gain loss; The electromagnetic compensation factor weights are dynamically allocated based on this percentage value. Environmental compensation factor weights and structural compensation factor weights And the constraint is ; Based on the current four-dimensional perception results, the corresponding electromagnetic amplitude compensation value is retrieved from the five-dimensional step-by-step predistortion mapping table. Environmental amplitude compensation value Structural amplitude compensation value and the corresponding electromagnetic phase compensation value Environmental phase compensation value Structural phase compensation value ; Generate the final amplitude The formula is: ; Generate the final phase The formula is: ; in, This is the initial, uncompensated baseline amplitude. This is the initial, uncompensated reference phase.

6. The system according to claim 4, characterized in that, The low-latency parallel processing module includes a deformation-triggered resource scheduling unit. The deformation-triggered resource scheduling unit performs the following resource allocation: when it is determined that the current deformation parameter level is the slight deformation level or below, the processor computing power resources are evenly allocated according to the initial set ratio. When the deformation parameter level reaches the moderate deformation level, processor computing resources above the preset ratio threshold will be preferentially allocated to the deformation calculation and structural compensation process. When the deformation parameter level reaches the severe deformation level or above, the emergency processing mode is triggered, suspending non-core task links within the system.

7. The system according to claim 4, characterized in that, The electromagnetic-environment-structure coordinated control module also includes a deformation adaptive multi-subarray coordinated unit, and the flexible antenna array is divided into multiple independent subarrays with independent power control. When the assessment finds that an independent subarray has reached the severe deformation level, the deformation adaptive multi-subarray cooperative unit actively attenuates the transmission power of the independent subarray that has reached the severe deformation level to a preset power attenuation threshold, and simultaneously increases the transmission power of the spatially adjacent normal subarrays to compensate for signal coverage. When the specified limit deformation level is reached, the excitation link of the independent subarray that has reached the specified limit deformation level is cut off, and the beamforming algorithm is executed according to the normal subarray to re-cover the target area.

8. The system according to claim 1, characterized in that, The multi-dimensional intelligent learning module includes a structural deformation prediction sub-model based on a long short-term memory neural network. The time-series input feature matrix of the structural deformation prediction sub-model includes UAV maneuver commands, multi-axis flight speed and acceleration, and historical continuous deformation data within a sliding time window; the structural deformation prediction sub-model outputs deformation prediction values ​​for a future preset look-ahead time period.

9. The system according to claim 8, characterized in that, The electromagnetic-environment-structure coordinated control module is coupled with a pre-compensation unit; The pre-compensation unit receives the deformation prediction value, calculates the associated structural compensation parameters in advance within the preset look-ahead time period, and performs the pre-distortion compensation adjustment on the antenna array in advance. After the predicted time point is reached and the actual deformation occurs, the residual is calculated using the measured deformation data fed back from the four-dimensional perception results. The residual is then used to generate a correction signal to update the closed-loop error compensation parameters of the previous execution.

10. The system according to claim 4, characterized in that, The cross-system collaboration and anomaly handling module includes an air-ground collaborative deformation compensation unit and a deformation anomaly handling unit. The air-to-ground cooperative deformation compensation unit is used to send the three-dimensional dynamic deformation field and its prediction data inside the system to the ground base station, and to receive the transmission power and modulation coding adjustment feedback sent in reverse by the ground base station, so as to realize the joint update of the beam pointing at both ends of the air-to-ground transceiver. When the deformation anomaly processing unit detects that the array is at the irreversible extreme deformation level, it executes the array reconstruction protection mechanism and injects forced deceleration and attitude correction commands into the upper-level UAV flight control system to suppress further aerodynamic load deformation.

Images