A control method, device and storage medium for a tethered drone against wind load

Abstract

The present invention relates to a control method, device and storage medium for a tethered unmanned aerial vehicle against wind loads. The method includes: collecting and modeling the tethered unmanned aerial vehicle to obtain a system dynamics model; performing state estimation to obtain system wind load disturbance data and desired attitude data; solving for optimal control of multiple tether points to obtain tension distribution control data for each tether point; inputting into a discrete-time neural network controller for attitude compensation calculation to obtain thrust vector compensation control data; performing adaptive robust control on the attitude of the tethered unmanned aerial vehicle to obtain rotational speed control commands for each motor of the quadrotor; and performing a safety threshold judgment. When any monitored data exceeds a preset threshold, a working mode switching instruction is output and a safe landing control is executed. The present invention establishes a multi-level safety protection mechanism, and through real-time monitoring and mode switching, ensures the reliable operation of the tethered unmanned aerial vehicle under extreme conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles, and particularly to a control method, device and storage medium for the wind load resistance of a tethered unmanned aerial vehicle. Background Art

[0002] In a strong wind environment, a tethered unmanned aerial vehicle faces serious wind load interference problems, which not only affect its flight stability, but may also cause excessive or uneven distribution of forces on the tether rope, endangering flight safety.

[0003] Currently, traditional control methods for the wind load resistance of tethered unmanned aerial vehicles mainly adopt PID control with fixed gains or simple adaptive control strategies, which are difficult to effectively cope with strong non-linear disturbances in complex wind field environments. At the same time, due to the strong coupling relationship between the forces on the tether points and the body attitude, a single control strategy is difficult to simultaneously take into account the attitude stability and the optimization of the tension distribution of the tether rope. In addition, existing control methods often neglect the safeguard mechanism for system safety and lack the ability to promptly respond to and handle abnormal situations such as sudden changes in wind load and overload of the tether points, which can easily lead to safety accidents in practical applications. Summary of the Invention

[0004] The main purpose of the present invention is to provide a control method, device and storage medium for the wind load resistance of a tethered unmanned aerial vehicle. The present invention establishes a multi-level safety protection mechanism, and through real-time monitoring and mode switching, ensures the reliable operation of the tethered unmanned aerial vehicle under extreme conditions.

[0005] To achieve the above object, the present invention provides a control method for the wind load resistance of a tethered unmanned aerial vehicle, including the following steps:

[0006] Collect and model the force data, body attitude data and environmental wind field data of multiple tether points of the tethered unmanned aerial vehicle to obtain a system dynamics model;

[0007] Based on the system dynamics model, perform state estimation to obtain system wind load disturbance data and desired attitude data;

[0008] According to the system wind load disturbance data and the desired attitude data, perform optimal control solution for the multiple tether points to obtain the tension distribution control data of each tether point;

[0009] Input the system wind load disturbance data and the tension distribution control data into a discrete-time neural network controller for attitude compensation calculation to obtain thrust vector compensation control data;

[0010] According to the thrust vector compensation control data, perform adaptive robust control on the attitude of the tethered unmanned aerial vehicle to obtain the rotational speed control commands of each motor of the quadrotor;

[0011] Perform a safety threshold judgment on the body attitude data, the force data of the mooring points, and the environmental wind field data. When any monitored data exceeds the preset threshold, output a working mode switching instruction and execute a safe landing control.

[0012] The present invention also provides a control device for a moored unmanned aerial vehicle against wind loads, including:

[0013] A modeling module for collecting and modeling the force data, body attitude data, and environmental wind field data of multiple mooring points of the moored unmanned aerial vehicle to obtain a system dynamics model;

[0014] A state estimation module for performing state estimation based on the system dynamics model to obtain system wind load disturbance data and desired attitude data;

[0015] A solution module for performing optimal control solution on the multiple mooring points according to the system wind load disturbance data and the desired attitude data to obtain the tension distribution control data of each mooring point;

[0016] A calculation module for inputting the system wind load disturbance data and the tension distribution control data into a discrete-time neural network controller for attitude compensation calculation to obtain thrust vector compensation control data;

[0017] A control module for performing adaptive robust control on the attitude of the moored unmanned aerial vehicle according to the thrust vector compensation control data to obtain the rotational speed control instructions of each motor of the quadrotor;

[0018] A judgment module for performing a safety threshold judgment on the body attitude data, the force data of the mooring points, and the environmental wind field data. When any monitored data exceeds the preset threshold, output a working mode switching instruction and execute a safe landing control.

[0019] The present invention also provides a computer device, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps of the method described in any one of the above are implemented.

[0020] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in any one of the above are implemented.

[0021] In summary, the technical solution provided by the present invention realizes an accurate description of the motion characteristics of a tethered unmanned aerial vehicle (UAV) in a strong wind environment by establishing a system dynamics model for multi-source sensing data fusion. The distributed model reference adaptive event-triggered control architecture is adopted to reduce the computational burden of the system and improve the compensation accuracy for wind load interference at the same time. The Hamilton-Jacobi-Bellman equation is introduced for the optimal distribution of the tether point tension, effectively solving the problem of balanced force on multiple tether points and avoiding local overload phenomena. A multi-level discrete-time neural network controller is designed, and through online learning and adaptive mechanisms, the adaptability of the system to unknown wind load interference is enhanced. The adaptive robust control strategy is adopted to improve the control stability of the system under large-range wind speed changes. A multi-level safety protection mechanism is established, and through real-time monitoring and mode switching, the reliable operation of the tethered UAV under extreme conditions is ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a schematic diagram of the steps of a control method for a tethered UAV against wind loads in an embodiment of the present invention;

[0023] Figure 2 is a block diagram of the structure of a control device for a tethered UAV against wind loads in an embodiment of the present invention;

[0024] Figure 3 is a schematic block diagram of the structure of a computer device in an embodiment of the present invention.

[0025] The implementation, functional features, and advantages of the objectives of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] In order to make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0027] Referring to Figure 1 , this embodiment provides a control method for a tethered UAV against wind loads, including the following steps:

[0028] S1, collect and model the force data, body attitude data, and environmental wind field data of multiple tether points of the tethered UAV to obtain a system dynamics model;

[0029] Among them, the force signals of the torque sensors installed at multiple mooring points of the tethered UAV are sampled in real time to obtain the real-time force data of each mooring point. These data include the magnitude and direction of the tension, and the timeliness and accuracy of the data are ensured through high-frequency sampling. While collecting the mooring point data, the body motion data is collected by using the triaxial acceleration sensor and angular velocity sensor on the body of the tethered UAV. Through sensor fusion technology, the acceleration and angular velocity data are comprehensively processed to eliminate the errors caused by sensor noise or drift, and accurate body attitude motion data is generated to reflect the attitude changes of the UAV in space, including the dynamic responses of the pitch, roll, and yaw angles. At the same time, an array of wind speed and direction sensors is arranged in the UAV operation area. Through the collection of local wind speed and direction data by each sensor in the array, the data is processed by using the spatial interpolation algorithm to generate a refined model of the environmental wind field distribution. This wind field model can reflect the distribution of wind speed and direction in space and can take into account the turbulence characteristics in the wind field to provide accurate wind load information for subsequent calculations. Based on the real-time collected mooring point force data and environmental wind field distribution data, tethered mechanics calculations are carried out to establish a force model of the tether rope. In this process, the nonlinear characteristics of the tether rope, including its stiffness, damping, and bending effects, are considered to construct a physical model that can accurately describe the dynamic behavior of the rope. This model can predict the tension changes and deformation conditions of the tether rope under the action of wind load. Through the kinematic analysis of the body attitude motion data and the real-time mooring point force data, a body attitude motion model is constructed. This model is based on the inertial coordinate system of the UAV and describes the motion law of the body under the mooring point constraints and external disturbances. The model integrates the rotational dynamics and linear kinematic equations to capture the attitude changes and displacement characteristics of the UAV. According to the environmental wind field distribution data, the aerodynamic moment of the body is calculated to obtain a wind load disturbance model. The wind load disturbance model can reflect the forces and moments exerted on the UAV by the wind field, including lift, drag, and lateral forces, as well as the aerodynamic moments generated by these forces. This model comprehensively considers the effects of wind speed, body geometry, and attitude to ensure that the description of wind load disturbance has physical significance. The tether rope force model, body attitude motion model, and wind load disturbance model are combined to form a comprehensive dynamic model of the system. The force conditions of the mooring points, the motion state of the UAV, and the comprehensive action of the environmental wind field are incorporated into a unified framework to completely describe the dynamic behavior of the tethered UAV. To improve the accuracy and adaptability of the model, the parameters in the comprehensive dynamic model of the system are optimized through parameter identification technology, and the model parameters are corrected by using experimental data or field operation data, so that the finally obtained system dynamic model can better match the actual application scenario.

[0030] S2. Based on the system dynamic model, state estimation is carried out to obtain the system wind load disturbance data and the desired attitude data;

[0031] Specifically, the system dynamics model is input into the distributed control network, and the entire dynamics system is decomposed through a network partitioning algorithm to obtain multiple subsystem control units. The complex global dynamics behavior is dispersed into different sub-units to achieve the efficiency and modular management of distributed control. On the basis of decomposition, an event-triggering threshold is set for the state data of each subsystem, and by setting the trigger condition criterion, it is ensured that each subsystem updates its state only when necessary, thus effectively reducing the computational overhead and communication burden. The system dynamics model is input into the reference model for response characteristic calculation. The reference model generates the expected dynamic response data of the system according to the design goal to ensure the stability and wind resistance of the UAV under various disturbances. By analyzing the dynamic response characteristics of the system under ideal conditions, the expected behavior of each subsystem in different states is obtained. These expected dynamic response data provide a reference benchmark for the evaluation of the actual state. By comparing the actual states of multiple subsystem control units with the expected dynamic response data provided by the reference model, the state tracking error data of each subsystem is calculated. These error data reflect the deviation between the actual system and the ideal response. The adaptive law is applied to update the state tracking error data to estimate the wind load disturbance data of the system. The adaptive law dynamically estimates the magnitude and direction of the wind load disturbance by adjusting the control gain and parameters in real time and using the information hidden in the error data, thereby providing accurate data support for compensation control and attitude optimization. This process can not only effectively capture the changes in the wind load disturbance, but also improve the adaptability of the system to complex wind fields, enabling the UAV to maintain stable flight under high wind load conditions. After the estimation of the wind load disturbance data is completed, the state tracking error data and the preset trigger condition criterion are input into the event judgment module for comparison. When the trigger condition is met, the module generates a state update trigger instruction. These instructions are used to activate the state update mechanism of the subsystem control unit to respond in a timely manner to changes in external disturbances. At the same time, according to the generated state update trigger instruction and the wind load disturbance data, the upwind attitude of the UAV is optimized. In this process, the optimal attitude angle data of the UAV is calculated through an optimization algorithm, and its upwind direction relative to the wind field is adjusted to minimize the influence of wind resistance and unstable torque. The wind resistance characteristics of the optimal attitude angle data are analyzed to evaluate the wind load performance of the UAV in this attitude, including key parameters such as wind resistance and aerodynamic torque. According to these analysis results, the final expected attitude data is generated.

[0032] S3. According to the system wind load disturbance data and the expected attitude data, perform optimal control solution for multiple mooring points to obtain the tension distribution control data of each mooring point;

[0033] It should be noted that the system wind load disturbance data and the desired attitude data are input into the HJB equation. Through its solution process, the time interval of the entire control problem is divided, the change trend of the system state in different stages is identified, and the originally complex dynamic problem is split into several more manageable small time intervals, which is convenient for subsequent control target setting and calculation accuracy optimization. The time interval division data is refined, and clear local control objectives and constraint conditions are set within each time sub-interval. These conditions include the balance requirements for attitude maintenance, the force range of each mooring point, and the dynamic impact of wind load disturbance on the mooring ropes. In this way, each sub-interval is given a clear physical meaning, providing a constraint basis for subsequent tension distribution calculation. In this process, the design of local control objectives will comprehensively consider the change trend of the desired attitude data to ensure the attitude stability of the UAV in each stage, while minimizing the energy consumption of the system and the non-uniformity of the tension distribution. According to the set local control constraint conditions, the position distribution data of each mooring point relative to the wind direction is calculated. Based on the relative relationship between the real-time wind direction and the UAV position, through geometric models and mechanical analysis, the distribution of each mooring point in three-dimensional space is determined, thus providing accurate input parameters for tension calculation. On this basis, the position distribution data of the mooring points is input into the tension calculation module, and combined with the wind load disturbance and the desired attitude requirements, the stress state of the mooring ropes is accurately analyzed. By establishing a correlation model of tension-position-wind load, the ideal tension data of each mooring point is calculated, and these data reflect the optimal force distribution of the mooring ropes under the current wind load and attitude requirements. After obtaining the ideal tension data, the HJB equation is used again to optimize and solve the tension distribution scheme. The solution process of the HJB equation can dynamically balance the forces of all parts of the system and ensure the calculation of the global optimal solution. Through this process, the generated optimal tension distribution scheme can effectively reduce the structural instability problem caused by uneven tension distribution in the system, and at the same time resist the interference of external wind loads to the greatest extent. The generated optimal tension distribution scheme is input into the position actuator to drive each mooring point to adjust the tension and achieve precise control. The position actuator ensures that the actual tension is consistent with the target tension by adjusting the length and angle of the rope in real time. During the execution of the tension control scheme, in order to prevent the tension from exceeding the limit caused by extreme external wind loads or accidental disturbances, the tension control amount of the mooring points is protected from exceeding the limit. By setting the tension protection threshold data, the tension change of each mooring point is monitored in real time, and the tension exceeding the threshold range is limited. The limiting operation can effectively avoid the risk of rope breakage or UAV attitude out of control caused by excessive local tension. The tension distribution control data after the limiting adjustment is transmitted to each mooring point.

[0034] S4. Input the system wind load disturbance data and the tension distribution control data into the discrete-time neural network controller for attitude compensation calculation to obtain the thrust vector compensation control data;

[0035] Specifically, the system wind load disturbance data and the tension distribution control data are input into the input layer of the discrete-time neural network controller. The input layer consists of 12 neuron nodes, and each node corresponds to three-dimensional wind load data (three components of wind speed and direction) and the tension data of four mooring points. The data of the input layer is transmitted to the first hidden layer for processing. The first hidden layer contains 24 neuron nodes and uses the ReLU activation function to achieve non-linear mapping. Preliminary feature extraction is performed on the input data to convert the complex input relationship into the first-layer feature mapping data representing the interaction between wind load disturbance and tension distribution. Through the non-linear transformation of the ReLU function, the response ability of the neural network to disturbances of different intensities is enhanced, enabling it to identify potential non-linear features. The first-layer feature mapping data is input into the wind load compensation control layer for processing. The wind load compensation control layer is a key layer in the neural network, containing 36 neuron nodes, and uses the Tanh activation function to perform non-linear transformation on the data, which is used to calculate the required thrust compensation amount of the system. This layer generates wind load compensation data that matches the current wind field conditions according to the dynamic changes of wind load disturbance and tension distribution. These data directly reflect the external force compensation requirements for the attitude stability of the UAV. In this process, the use of the Tanh activation function ensures that the output data has a smoother range of changes and can effectively process the complex features in the wind load data. The wind load compensation data is input into the second hidden layer for processing. The second hidden layer also contains 24 neuron nodes and uses the ReLU activation function for feature extraction. At this stage, the network refines the implicit features in the wind load compensation data to generate the second-layer feature mapping data. Through the multi-layer feature extraction structure, the deep learning ability of the neural network to disturbance data is ensured, enabling it to effectively capture the dynamic changes of the system. The second-layer feature mapping data is input into the attitude tracking control layer for processing. The attitude tracking control layer contains 18 neuron nodes and uses the Sigmoid activation function to calculate the desired attitude angle through the transformation of the feature mapping data. The output of this layer is the attitude compensation data, which represents the attitude adjustment information required by the UAV under the combined action of wind load and tension distribution. The use of the Sigmoid activation function in this layer maps the attitude compensation data to a limited range, ensuring the stability and controllability of the calculation results. The generated attitude compensation data is input into the third hidden layer for processing. The third hidden layer contains 12 neuron nodes and uses the ReLU activation function to complete the dimensionality reduction mapping, compressing the high-dimensional features into low-dimensional features suitable for thrust allocation. Through this dimensionality reduction process, the network simplifies the complexity of subsequent calculations while ensuring that the important information of the data is not lost. The dimensionality-reduced feature data is input into the thrust allocation layer for processing. The thrust allocation layer contains 8 neuron nodes and uses the linear activation function to directly calculate the thrust components of each rotor. Through the processing of this layer, the thrust components of each rotor can match the wind load compensation requirements and attitude compensation requirements, achieving precise control of the UAV attitude.Transfer the thrust allocation data to the output layer for processing. The output layer contains 4 neuron nodes, and the Tanh activation function is used to normalize the data, and finally generate the thrust vector compensation control data.

[0036] S5. Perform adaptive robust control on the attitude of the tethered UAV according to the thrust vector compensation control data, and obtain the rotational speed control commands of each motor of the quadrotor.

[0037] Among them, the thrust vector compensation control data is input into the nominal controller for decomposition calculation. The nominal controller calculates the desired angle data of the roll angle, pitch angle, and yaw angle according to the characteristics of the compensation data and the attitude requirements. These desired angle data indicate the attitude targets that the UAV should reach under ideal conditions, providing the direction and amplitude of attitude adjustment. The desired angle data is compared with the actual attitude data to calculate the three-axis attitude angle error and the corresponding angular velocity error. These error data reflect the deviation between the current attitude state of the UAV and the target attitude. The attitude angle error and angular velocity error are input into the adaptive robust controller for tracking control calculation. The adaptive robust controller dynamically adjusts the control parameters according to the real-time state and error characteristics of the system, and through the estimation and compensation of disturbances and uncertainties, outputs the attitude angular acceleration control quantity that can respond quickly and has high robustness. These control quantities can correct the attitude error in real time and can cope with the influence of external disturbances such as wind load disturbances, thereby improving the attitude stability of the UAV. Perform a dynamic inverse mapping on the attitude angular acceleration control quantity to convert it into the total thrust control quantity of the quadrotor and the three-axis torque control quantity that is more suitable for physical execution. Using the dynamic model of the UAV, map the output of the controller from the form of angular acceleration to the form of thrust and torque to achieve matching with the actual forces of the rotors. The total thrust control quantity reflects the lift demand of the UAV in the vertical direction, while the three-axis torque control quantity determines the distribution of the rotational torque of the UAV in the roll, pitch, and yaw directions. Solve the thrust mapping matrix according to the total thrust control quantity of the quadrotor and the three-axis torque control quantity. Through mathematical transformation, the thrust mapping matrix distributes the demands of the total thrust and torque to the thrusts of the four rotors, ensuring that the four rotors can meet the balance requirements and attitude control requirements of the system under their combined action. This process not only needs to consider the linear distribution of thrust, but also needs to comprehensively consider factors such as rotor position, lever arm length, and relative angle, so that the thrust distribution has physical feasibility and mechanical balance. Input the thrust distribution data of the four rotor motors into the characteristic curve of the rotor for speed mapping. The characteristic curve describes the non-linear relationship between the motor speed and the output thrust. Through this mapping process, the thrust demand is converted into the corresponding equivalent motor speed data to ensure the consistency between the controller output and the actual motor drive parameters. At this stage, dead zone compensation and saturation limiting are performed on the equivalent speed data to eliminate the non-linear response of the motor in the small thrust range and prevent the motor speed from exceeding the safe range, resulting in mechanical failures or reduced efficiency. Input the safe speed range data processed by dead zone compensation and saturation limiting into the electronic speed controller (ESC) drive module, and generate specific control signals for driving the motor through PWM signal conversion. The ESC drive module converts the input speed command into the PWM signal required by the motor, and these signals finally achieve precise speed control of the four rotors by adjusting the supply voltage and frequency of the motor.

[0038] S6. Perform a safety threshold judgment on the aircraft attitude data, the force data of the mooring point, and the environmental wind field data. When any of the monitored data exceeds the preset threshold, output a working mode switching instruction and execute a safe landing control.

[0039] Specifically, set safety thresholds for the aircraft attitude data, define the safe area within the attitude angle range, such as the limit ranges of roll angle, pitch angle, and yaw angle, as well as the maximum allowable value of the angular velocity, to generate attitude threshold range data. At the same time, set safety thresholds for the force data of the mooring points to monitor the tension range that the mooring ropes bear during wind loads and the dynamic adjustment of the UAV. According to the rope material properties, mooring point distribution, and dynamic load calculation, set the tension threshold range data for each mooring point, including the maximum tension limit to prevent breakage and the minimum tension limit to avoid slack or failure. Similarly, set safety thresholds for the environmental wind field data, and set the safe ranges of wind speed, wind direction, and turbulence intensity according to the wind resistance ability of the UAV design and the requirements of the mission scenario to generate wind field threshold range data. These threshold range data together constitute the safe operation boundary of the UAV. Input the attitude threshold range data, mooring tension threshold range data, and wind field threshold range data into the monitoring module for multi-threshold judgment. The monitoring module compares the real-time collected UAV state data and external environment data with the preset threshold ranges. The multi-threshold judgment result data can identify whether the current system is in a safe state. If any data is detected to exceed the threshold, a working mode switching instruction will be triggered to switch the UAV from the normal operation mode to the emergency control mode. In the emergency control mode, input the multi-threshold judgment result data into the attitude protection module for calculation. The attitude protection module calculates the desired attitude angle and angular velocity through the self-stabilization algorithm to generate attitude self-stabilization data. These data ensure that the UAV can maintain basic stability under extreme conditions, such as restoring excessive pitch or roll angles to the safe range, while reducing the drastic change of the angular velocity to avoid attitude out of control. The self-stabilization data provides a reliable attitude reference for the subsequent landing trajectory planning. Based on the attitude self-stabilization data, plan the landing trajectory and determine the descent path and speed control strategy of the UAV. At this stage, comprehensively consider the current wind field conditions, UAV altitude, and dynamic characteristics, and generate descent speed control data through an optimization algorithm to ensure that the UAV can gradually approach the ground at a controllable speed within the safe range. The descent speed control data also needs to adapt to the dynamic disturbances suffered by the UAV during the landing process to ensure the smoothness and accuracy of the landing trajectory. Input the descent speed control data into the motor control distribution module to calculate the rotation speed adjustment instructions for each motor of the quadrotor, so as to achieve safe landing control. The motor control distribution module adjusts the rotation speed of each motor in real time according to the dynamic model of the UAV and the current attitude adjustment requirements to balance the attitude of the UAV and control the descent speed. At the same time, perform dead zone compensation and amplitude limiting operations to ensure that the motor operates within the safe range and prevent overload or failure.

[0040] In one example, collect and model the force data, aircraft attitude data, and environmental wind field data of multiple mooring points of the moored UAV to obtain a system dynamics model, including:

[0041] Sample the force signals of the torque sensors at multiple mooring points to obtain the real-time force data of the mooring points, and fuse the data of the triaxial acceleration sensor and angular velocity sensor of the moored UAV to obtain the body attitude motion data;

[0042] Perform spatial interpolation on the data collected by the wind speed and direction sensor array in the operation area of the moored UAV to obtain the environmental wind field distribution data;

[0043] Based on the real-time force data of the mooring points and the environmental wind field distribution data, conduct mooring mechanics calculations to obtain the mooring rope force model, and perform kinematic analysis on the body attitude motion data and the real-time force data of the mooring points to obtain the body attitude motion model;

[0044] Perform aerodynamic moment calculations on the environmental wind field distribution data to obtain the wind load interference model;

[0045] Combine the mooring rope force model, the body attitude motion model and the wind load interference model to obtain the system comprehensive dynamics model, and perform parameter identification on the system comprehensive dynamics model to obtain the system dynamics model.

[0046] In this example, the force signals of the mooring points are sampled by torque sensors installed at multiple mooring points, and these signals include the tension magnitude of the mooring ropes and direction vectors , where represents the number of the mooring point, and there are a total of mooring points. Using these sensor data, the force conditions of each mooring point are obtained in real time to form the real-time force data of the mooring points. At the same time, the triaxial acceleration sensor and angular velocity sensor on the UAV body collect data through the inertial measurement unit, and respectively record the linear acceleration of the body and angular velocity . These data are processed by a sensor fusion algorithm, such as Kalman filtering or extended Kalman filtering, to combine the acceleration and angular velocity data, filter out noise and drift, and obtain the dynamic motion data of the body attitude, including the attitude angles (pitch angle , roll angle , yaw angle ) and the attitude change rate. After completing the acquisition of the body attitude and mooring point force data, the wind field data in the operation area of the UAV is processed. Through the wind speed and direction sensor array arranged in the operation area, the local wind speed and wind direction angle are collected in real time. These data are processed by a spatial interpolation algorithm (such as Kriging interpolation or inverse distance weighted interpolation) to generate a wind field distribution model , where are spatial coordinates. This model describes the wind speed and direction at any point and captures the changing characteristics of the wind field. Based on the real-time force data of the mooring points and the environmental wind field distribution data, mooring mechanics calculations are performed to obtain the mooring rope force model. The mooring rope force model is based on the tension and direction vector of each rope, and reflects the rope's acting force by calculating the resultant force and resultant moment exerted on the UAV by each mooring point. Its basic formula is:

[0047] ;

[0048] ;

[0049] where, is the total mooring force; is the total moment; is the displacement vector of the th mooring point relative to the UAV's center of mass; represents the cross product operation. At the same time, kinematic analysis is performed on the aircraft attitude motion data and the real-time force data of the mooring points to obtain the aircraft's attitude motion model. The motion model is described by the Newton-Euler equations as:

[0050] ;

[0051] ;

[0052] where, is the UAV's mass; is the inertia matrix; is the inertia damping term; is the gravity term; is the wind field acting force; is the wind field acting moment. Based on the wind field distribution data, the aerodynamic moment model generated by the wind field on the UAV is calculated. The aerodynamic moment calculation formula is:

[0053] ;

[0054] where, is the pressure distribution of the wind field acting on the UAV's surface; is the differential area. By combining the mooring rope force model, the aircraft attitude motion model and the wind load interference model, a complete system integrated dynamics model is constructed. This model describes the comprehensive behavior of the UAV under the action of mooring force, wind load force and aircraft dynamics. To improve the accuracy of the model, parameter identification technology is used to optimize the key parameters in the integrated dynamics model to obtain the system dynamics model. By comparing the actual data measured through experiments with the model output data, the model parameters, such as the mooring rope stiffness, damping coefficient, UAV inertia matrix, etc., are adjusted using the least squares method or the recursive least squares method, so that the model can accurately reflect the actual operating conditions.

[0055] In one example, state estimation is performed based on a system dynamics model to obtain system wind load disturbance data and desired attitude data, including:

[0056] The system dynamics model is input into a distributed control network for node partitioning to obtain multiple subsystem control units, and event trigger thresholds are set for the state data of the multiple subsystem control units to obtain trigger condition criteria;

[0057] The system dynamics model is input into a reference model for response characteristic calculation to obtain desired dynamic response data, and error calculation is performed on the actual states of the multiple subsystem control units according to the desired dynamic response data to obtain state tracking error data;

[0058] Adaptive law update is performed on the state tracking error data to obtain system wind load disturbance data;

[0059] The state tracking error data and the trigger condition criteria are input into an event judgment module for comparison to obtain a state update trigger command;

[0060] The windward attitude of the aircraft is optimized according to the state update trigger command and the system wind load disturbance data to obtain optimal attitude angle data, and wind resistance characteristic analysis is performed on the optimal attitude angle data to obtain desired attitude data.

[0061] In this example, the system dynamics model is input into a distributed control network, and the system nodes are decomposed through a network partitioning algorithm (such as a graph theory method or a clustering algorithm), and the overall model is divided into multiple subsystem control units. These subsystem control units respectively correspond to different control modules of the unmanned aerial vehicle, such as a pitch control unit, a roll control unit, and a yaw control unit, or are divided according to the mechanical characteristics of each mooring point. Assume that the system dynamics model is represented in matrix form as:

[0062] ;

[0063] where represents the local dynamic characteristics of subsystem , represents the coupling relationship between subsystem and subsystem . Through the node partitioning algorithm, the complex global model is decomposed into multiple relatively independent subsystem control units, which is convenient for parallel processing and local optimization. After the subsystem partitioning is completed, event trigger thresholds are set for the state data of each subsystem to construct trigger condition criteria. Assume that the state vector of subsystem is , and its trigger threshold is expressed as:

[0064] ;

[0065] where is the reference state, is the threshold parameter of the subsystem, which is used to determine whether the system needs to trigger a state update. Each subsystem judges whether it needs to recalculate the control input by comparing the deviation between the current state and the reference state. The system dynamics model is input into the reference model for response characteristic calculation. The reference model generates expected dynamic response data by simulating the dynamic behavior of the UAV under ideal conditions . These data include the expected attitude angles (such as the pitch angle , roll angle , yaw angle ) and their change rates. The expected dynamic response is calculated by model predictive control or optimal control methods, and is specifically expressed as:

[0066] ;

[0067] where is the control input, is the system state, is the reference state. Based on the expected dynamic response data, the error calculation is performed on the actual states of multiple subsystems to obtain the state tracking error data. The state tracking error is defined as:

[0068] ;

[0069] where is the error vector of subsystem . The error data is input into the adaptive law for update. The role of the adaptive law is to dynamically adjust the control gain of the system to compensate for the errors caused by external disturbances or system uncertainties. The update formula of the adaptive law is expressed as:

[0070] ;

[0071] where is the control gain matrix of subsystem , is the learning rate parameter. After the adaptive law update is completed, the state tracking error data and the trigger condition criterion are input into the event judgment module for comparison. If the error exceeds the set trigger threshold, that is , the module will generate a state update trigger instruction. This instruction is used to activate the control algorithm to recalculate the control input of the subsystem, so as to respond to the change of the system state in real time. According to the state update trigger instruction and the system wind load disturbance data, the upwind attitude of the UAV is optimized. The goal of upwind attitude optimization is to minimize the influence of wind load on the UAV by adjusting the attitude angle. The optimization problem is expressed as:

[0072] ;

[0073] where is the vector function of the wind load force. By solving this optimization problem, the optimal attitude angle data is obtained. The wind resistance characteristics of the optimal attitude angle data are analyzed to calculate the expected aerodynamic moment and wind resistance moment of the UAV at this attitude. The wind resistance characteristics analysis is achieved through the following formula:

[0074] ;

[0075] where is the wind resistance moment, is the displacement vector of the acting point, is the wind force acting on the surface of the UAV. The results of the wind resistance characteristics analysis are combined with the attitude optimization data to generate the expected attitude data, ensuring that the UAV can maintain stable operation in a complex wind field.

[0076] In an example, based on the system wind load disturbance data and the expected attitude data, the optimal control solution is obtained for multiple mooring points, and the tension distribution control data of each mooring point is obtained, including:

[0077] Input the system wind load disturbance data and the expected attitude data into the Hamilton-Jacobi-Bellman equation to obtain the time interval division data;

[0078] Set the sub-interval control objectives for the time interval division data to obtain the local control constraint conditions;

[0079] Calculate the position of each mooring point relative to the wind direction according to the local control constraint conditions to obtain the mooring point position distribution data;

[0080] Input the mooring point position distribution data into the tension calculation module for force analysis to obtain the ideal tension data of the mooring point;

[0081] Solve the Hamilton-Jacobi-Bellman equation for the ideal tension data of the mooring point to obtain the optimal tension distribution scheme, and input the optimal tension distribution scheme into the position actuator for tension adjustment to obtain the mooring point tension control amount;

[0082] Perform over-tension protection on the mooring point tension control amount to obtain the tension protection threshold data, and limit the mooring point tension control amount according to the tension protection threshold data to obtain the tension distribution control data of each mooring point.

[0083] In this example, the system wind load disturbance data and the desired attitude data are input into the HJB equation. The HJB equation finds an optimal control strategy from the current state to the desired state through dynamic programming, and its form is:

[0084] ;

[0085] where is the cost function, is the state variable, is the control input, is the cost term, is the state transition function. By solving the HJB equation, the time interval partition data is obtained, which is used to divide the global dynamic process into several time sub-intervals, and the control objectives are more clearly defined within each sub-interval. After the time interval partition is completed, the control objectives are set for each time sub-interval to obtain the local control constraint conditions. For example, for a time sub-interval, the goal is to maintain a certain attitude of the UAV (such as the stability of the pitch angle ), while satisfying the balance constraint of the tether point tension. The local control constraint conditions are expressed as:

[0086] ;

[0087] ;

[0088] where is the tension of the th tether point, is the tension direction vector, is the position vector of the tether point relative to the UAV's centroid, is the wind load force, is the wind load moment. According to the local control constraint conditions, the position distribution data of each tether point relative to the wind direction is calculated. This calculation process is based on geometric relationships and mechanical equilibrium. For example, if the wind load mainly comes from the -axis direction, the position distribution of the tether points needs to be adjusted to counteract the wind load moment so that the UAV's attitude remains stable. Let the coordinates of the tether points be , and solve through the optimization problem:

[0089] ;

[0090] ;

[0091] where is the total tether force, is a set of position constraints. The mooring point position distribution data is input into the tension calculation module, and force analysis is performed based on the mechanical properties of the mooring ropes to obtain the ideal tension data. These ideal tension data are obtained by solving the following equations:

[0092] ;

[0093] where is the component force of the th mooring point. The ideal tension data is optimized and solved again using the HJB equation to generate an optimal tension distribution scheme. This process needs to consider the uniformity and efficiency of the tension distribution based on the global cost function. For example:

[0094] ;

[0095] where is the weight parameter, is the reference position. The HJB equation provides a step-by-step optimization strategy at this stage to ensure that the tension distribution scheme can balance the wind load disturbance and the UAV attitude requirements. The optimal tension distribution scheme is input into the position actuator, and adjustment is achieved by adjusting the mooring rope length and tension to generate the mooring point tension control quantity. To ensure the safety of the tension control, over-limit protection is performed on the tension control quantity. Set the tension protection threshold data , and limit the tension outside the range:

[0096] ;

[0097] Through the limiting operation, ensure that the mooring ropes will neither break (exceed ), nor be too loose (below ), thus maintaining the stability of the UAV. Through the above calculations, the tension distribution control data of each mooring point is obtained.

[0098] In an example, the system wind load disturbance data and the tension distribution control data are input into a discrete-time neural network controller for attitude compensation calculation to obtain the thrust vector compensation control data, including:

[0099] The system wind load disturbance data and the tension distribution control data are input into the input layer of the discrete-time neural network controller. The input layer includes 12 neuron nodes, corresponding to three-dimensional wind load data and the tension data of four mooring points respectively;

[0100] The data in the input layer is processed by the first hidden layer. The first hidden layer contains 24 neuron nodes, and the ReLU activation function is used for non-linear mapping to obtain the first layer feature mapping data;

[0101] Input the first layer of feature mapping data into the wind load compensation control layer for processing. The wind load compensation control layer contains 36 neuron nodes, and uses the Tanh activation function to calculate the required thrust compensation amount to obtain wind load compensation data;

[0102] Perform second hidden layer processing on the wind load compensation data. The second hidden layer contains 24 neuron nodes, and uses the ReLU activation function for feature extraction to obtain the second layer of feature mapping data;

[0103] Input the second layer of feature mapping data into the attitude tracking control layer for processing. The attitude tracking control layer contains 18 neuron nodes, and uses the Sigmoid activation function to calculate the desired attitude angle to obtain attitude compensation data;

[0104] Perform third hidden layer processing on the attitude compensation data. The third hidden layer contains 12 neuron nodes, and uses the ReLU activation function for dimensionality reduction mapping to obtain the third layer of feature mapping data;

[0105] Input the third layer of feature mapping data into the thrust allocation layer for processing. The thrust allocation layer contains 8 neuron nodes, and uses the linear activation function to calculate the thrust components of each rotor to obtain thrust allocation data;

[0106] Perform output layer processing on the thrust allocation data. The output layer contains 4 neuron nodes, and uses the Tanh activation function for normalization processing to obtain thrust vector compensation control data.

[0107] In this example, input the system wind load disturbance data and the tension distribution control data into the input layer of the discrete-time neural network. The input layer contains 12 neuron nodes. Among them, the three-dimensional wind load data is represented as , reflecting the components of the wind load in the three axes of the UAV. The tension data of the four mooring points are respectively , representing the real-time tension magnitude of each mooring point. The input layer passes these input values to the next layer through the simple linear operation of the neurons. Pass the data of the input layer to the first hidden layer for processing. The first hidden layer contains 24 neuron nodes, and each node performs a non-linear mapping through the ReLU activation function. The form of the ReLU function is:

[0108] ;

[0109] where is the input data. This function enhances the network's ability to express non-linear features by truncating negative values to zero. In this layer, the interaction between the wind load data and the tension data is extracted as the first layer of feature mapping data, denoted as , these data contain the preliminary characteristics of the dynamic impact of wind load disturbances and tension distribution on the UAV system. The first-layer feature mapping data is input into the wind load compensation control layer, which contains 36 neuron nodes and is processed by the Tanh activation function. The Tanh activation function is defined as:

[0110] ;

[0111] Its output range is between [-1, 1], which is suitable for smooth transition non-linear mapping. The task of the wind load compensation control layer is to calculate the thrust compensation amount required for the UAV according to the feature mapping data, denoted as . These thrust compensation data are used to offset the disturbances of the wind load on the UAV's attitude, ensuring that the UAV's attitude can be maintained stable. The wind load compensation data is passed to the second hidden layer, which contains 24 neuron nodes and uses the ReLU activation function for feature extraction. Similar to the first hidden layer, the second hidden layer further mines the potential information in the wind load compensation data through non-linear mapping to generate the second-layer feature mapping data, denoted as . The second-layer feature mapping data is input into the attitude tracking control layer for processing, which contains 18 neuron nodes and uses the Sigmoid activation function to calculate the desired attitude angles. The Sigmoid function is defined as:

[0112] ;

[0113] Its output range is [0, 1], which is suitable for generating attitude compensation data. These data represent the desired pitch angle, roll angle, and yaw angle, denoted as . The attitude compensation data provides the attitude adjustment target for thrust allocation. The attitude compensation data is passed to the third hidden layer for dimensionality reduction mapping. The third hidden layer contains 12 neuron nodes and uses the ReLU activation function. Through non-linear mapping, the high-dimensional features are compressed into a data dimension suitable for thrust allocation while retaining key information. The third-layer feature mapping data is denoted as . The third-layer feature mapping data is input into the thrust allocation layer for processing. The thrust allocation layer contains 8 neuron nodes and uses the linear activation function to calculate the thrust components of each rotor. The output of the thrust allocation layer is denoted as , where represents the thrust of the th rotor. The thrust allocation relationship is represented in the following matrix form:

[0114] ;

[0115] where is the thrust distribution matrix, which describes the linear relationship between the feature mapping data and the thrust components. The thrust distribution data is passed to the output layer for normalization processing. The output layer contains 4 neuron nodes and uses the Tanh activation function. The output of the Tanh function normalizes the thrust distribution data to the range of [-1, 1], ensuring that the data adapts to the motor drive control requirements to generate the final thrust vector compensation control data .

[0116] In one example, the attitude of the tethered UAV is adaptively and robustly controlled according to the thrust vector compensation control data to obtain the rotational speed control commands for each motor of the quadrotor, including:

[0117] Input the thrust vector compensation control data into the nominal controller for decomposition calculation to obtain the desired angle data of the roll angle, pitch angle, and yaw angle;

[0118] Calculate the attitude error of the desired angle data to obtain the three-axis attitude angle error and angular velocity error data, and input the three-axis attitude angle error and angular velocity error data into the adaptive robust controller for tracking control calculation to obtain the attitude angular acceleration control quantity;

[0119] Perform dynamic inverse mapping on the attitude angular acceleration control quantity to obtain the total thrust control quantity of the quadrotor and the three-axis torque control quantity;

[0120] Solve the thrust mapping matrix according to the total thrust control quantity of the quadrotor and the three-axis torque control quantity to obtain the thrust distribution data of the four rotor motors;

[0121] Input the thrust distribution data of the four rotor motors into the characteristic curve for rotational speed mapping to obtain the equivalent rotational speed data of the four motors, and perform dead zone compensation and saturation limiting on the equivalent rotational speed data of the four motors to obtain the safe rotational speed range data;

[0122] Input the safe rotational speed range data into the ESC drive module for PWM signal conversion to obtain the rotational speed control commands for each motor of the quadrotor.

[0123] In this example, the thrust vector compensation control data is input into the nominal controller for decomposition calculation. The nominal controller, based on the dynamic characteristics of the UAV, solves the relationship between the input thrust vector and the attitude angle to obtain the desired roll angle , pitch angle and yaw angle . This decomposition is achieved through the following relationships:

[0124] ;

[0125] ;

[0126] ;

[0127] where is the total thrust, offset is the reference value of the yaw angle, is the yaw sensitivity coefficient. Compare the desired angle data with the current attitude angle data, and calculate the three-axis attitude angle error and angular velocity error. Let the current attitude angle be , the angular velocity be , and the error calculation formula is:

[0128] ;

[0129] ;

[0130] These error data reflect the deviation between the current attitude state and the target, and are the core basis for the control system to adjust. Input the attitude angle error and angular velocity error into the adaptive robust controller. The adaptive robust controller compensates for external disturbances and system uncertainties by adjusting the control parameters in real time, thereby generating the attitude angular acceleration control quantity . The adaptation law is implemented through the following formula:

[0131] ;

[0132] ;

[0133] ;

[0134] where is the proportional and differential gain, is the adaptive compensation term for dealing with nonlinearity and disturbances. After obtaining the attitude angular acceleration control quantity, perform dynamic inverse mapping to convert it into the total thrust control quantity of the quadrotor and the three-axis torque control quantity . The dynamic relationship of the quadrotor is as follows:

[0135] ;

[0136] ;

[0137] where is the arm length of the rotor relative to the centroid of mass, is the torque coefficient. According to the above thrust and torque relationships, construct the thrust mapping matrix:

[0138] ;

[0139] where is the thrust distribution matrix, which describes the mapping relationship between the rotor thrust and the overall thrust and moment. The thrust distribution data is input into the rotor characteristic curve for rotational speed mapping. The characteristic curve correlates the thrust with the rotor rotational speed and is transformed through the following formula:

[0140] ;

[0141] where is the rotational speed of the th rotor, and is the lift coefficient. To ensure the reliability and efficiency of the motor operation, dead zone compensation and saturation limiting processing are performed on the rotational speed data. Dead zone compensation is used to eliminate the non-responsive interval of the motor at small inputs, while limiting prevents the rotational speed from exceeding the motor design range. The formula is:

[0142] ;

[0143] where and are the minimum and maximum rotational speed thresholds respectively. The processed safe rotational speed range data is input into the ESC drive module, and control commands for the motor are generated through PWM signal conversion to drive the quadrotor system to achieve the desired attitude and motion control.

[0144] In an example, safety threshold judgments are made on the body attitude data, the force data at the mooring point, and the environmental wind field data. When any of the monitored data exceeds the preset threshold, a working mode switching command is output and safety landing control is executed, including:

[0145] Set safety thresholds for the body attitude data to obtain attitude threshold range data;

[0146] Set safety thresholds for the force data at the mooring point to obtain mooring tension threshold range data;

[0147] Set safety thresholds for the environmental wind field data to obtain wind field threshold range data;

[0148] Input the attitude threshold range data, the mooring tension threshold range data, and the wind field threshold range data into the monitoring module for multi-threshold judgment to obtain multi-threshold judgment result data;

[0149] Switch the control mode according to the multi-threshold judgment result data to obtain emergency control mode data;

[0150] Input the emergency control mode data into the attitude protection module for calculation to obtain attitude self-stabilization data, which includes the desired attitude angles and angular velocities;

[0151] Perform landing trajectory planning on the attitude self-stabilization data to obtain the descent speed control data, and perform motor control allocation on the descent speed control data to obtain the safe landing control commands for each motor of the quadrotor.

[0152] In this example, set safety thresholds for the body attitude data. By analyzing the stability and mission requirements of the UAV, determine the limit conditions for the attitude angle range and its change rate. The safety thresholds for the pitch angle ( ), roll angle ( ), and yaw angle ( ) of the attitude are expressed as:

[0153] ;

[0154] The limit conditions for the angular velocity are expressed as:

[0155] ;

[0156] These data constitute the attitude threshold range, which is used to determine whether the UAV exceeds the stable working range. At the same time, set safety thresholds for the force data of the mooring points. By analyzing the mechanical properties and ultimate bearing capacity of the mooring ropes, set the allowable range of the tension of each mooring point. Assume that the total number of mooring points is , and the tension of the th mooring point is , then its safety threshold is:

[0157] ;

[0158] Among them, is the minimum tension to prevent the rope from loosening, and is the maximum tension to avoid rope breakage. Set safety thresholds for the environmental wind field data. By analyzing the variation ranges of wind speed, wind direction, and turbulence intensity, determine the wind resistance ability of the UAV. Assume that the wind speed is , the wind direction angle is , and the turbulence intensity is , then its safety threshold is expressed as:

[0159] ;

[0160] These wind field threshold ranges are used to evaluate whether the UAV can work normally under the current wind field conditions. Input the attitude threshold range, mooring tension threshold range, and wind field threshold range data into the monitoring module. The monitoring module collects the state data of the UAV in real time and performs multi-threshold judgment. When the multi-threshold judgment result data indicates that the UAV exceeds the safe range, trigger the emergency control mode and input the mode data into the attitude protection module for calculation. The attitude protection module generates the desired attitude angle and angular velocity . The self-stabilization algorithm is based on the dynamic model of the UAV, and calculates the safe attitude target by minimizing the attitude error and disturbance influence:

[0161] ;

[0162] Where is the target attitude, is the current attitude, is the control input. Landing trajectory planning is performed on the attitude self-stabilization data. By planning the descent path and speed control strategy, it is ensured that the UAV can gradually reduce its altitude in a safe manner. Let the current altitude be , the target landing altitude be , and the planned descent speed be , then the trajectory planning is expressed as:

[0163] ;

[0164] ;

[0165] Where is the initial altitude, is the descent speed control gain. The planned descent speed control data is input into the motor control distribution module to calculate the thrust requirements of the quadrotor motors. Thrust distribution distributes the overall thrust and torque requirements to each motor according to the dynamic characteristics of the UAV:

[0166] ;

[0167] Where is the thrust distribution matrix, is the total thrust, are the roll, pitch, and yaw torques respectively. Speed mapping is performed on the calculated thrust distribution data to convert the thrust into motor speed through the rotor characteristic curve:

[0168] ;

[0169] Where is the lift coefficient. To ensure speed safety, dead zone compensation and saturation limiting are performed on the speed:

[0170] ;

[0171] Where and are the minimum and maximum speeds allowed by the motor. Finally, the safe landing control instructions for each quadrotor motor are obtained.

[0172] Referring to Figure 2 , this embodiment provides a control device for a tethered UAV against wind loads, including:

[0173] A modeling module 1, configured to collect and model the force data, body attitude data, and environmental wind field data of multiple mooring points of a tethered drone to obtain a system dynamics model;

[0174] A state estimation module 2, configured to perform state estimation based on the system dynamics model to obtain system wind load disturbance data and desired attitude data;

[0175] A solution module 3, configured to perform optimal control solution on multiple mooring points according to the system wind load disturbance data and the desired attitude data to obtain the tension distribution control data of each mooring point;

[0176] A calculation module 4, configured to input the system wind load disturbance data and the tension distribution control data into a discrete-time neural network controller for attitude compensation calculation to obtain thrust vector compensation control data;

[0177] A control module 5, configured to perform adaptive robust control on the attitude of the tethered drone according to the thrust vector compensation control data to obtain the rotation speed control instructions of each motor of the quadrotor;

[0178] A judgment module 6, configured to perform safety threshold judgment on the body attitude data, the force data of the mooring point, and the environmental wind field data. When any monitored data exceeds a preset threshold, output a working mode switching instruction and execute a safe landing control.

[0179] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to that described in the above method embodiment, and details are not described herein again.

[0180] Refer to Figure 3 , in the embodiment of the present invention, a computer device is further provided. The computer device may be a server, and its internal structure may be as Figure 3 shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0181] Those skilled in the art can understand, Figure 3The structure shown is only a block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0182] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0183] Those of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, storage, database or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0184] It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, device, article or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, device, article or method including that element.

[0185] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be similarly included in the patent protection scope of the present invention.

Claims

1. A method for controlling wind load of a tethered UAV, characterized in that: The following steps are involved: The force data of multiple tethering points of the tethered UAV, the body posture data and the environmental wind field data are collected and modeled to obtain the system dynamics model; Perform state estimation based on the system dynamics model to obtain system wind load disturbance data and expected attitude data; According to the system wind load disturbance data and the desired attitude data, the optimal control of the multiple mooring points is solved to obtain the tension distribution control data of each mooring point; specifically, the method includes: inputting the system wind load disturbance data and the desired attitude data into the Hamilton-Jacobi-Bellman equation to obtain time interval division data; setting sub-interval control targets for the time interval division data to obtain local control constraints; calculating the position of each mooring point relative to the wind direction according to the local control constraints to obtain mooring point position distribution data; inputting the mooring point position distribution data into the tension calculation module for force analysis to obtain ideal tension data of the mooring point; solving the Hamilton-Jacobi-Bellman equation for the ideal tension data of the mooring point to obtain an optimal tension distribution scheme, and inputting the optimal tension distribution scheme into the position actuator for tension adjustment to obtain the tension control amount of the mooring point; performing tension over-limit protection on the tension control amount of the mooring point to obtain tension protection threshold data, and limiting the tension control amount of the mooring point according to the tension protection threshold data to obtain the tension distribution control data of each mooring point; Inputting the system wind load disturbance data and the tension distribution control data into a discrete time neural network controller to perform attitude compensation calculation to obtain thrust vector compensation control data; Adaptively and robustly controlling the attitude of the tethered UAV according to the thrust vector compensation control data to obtain a speed control instruction for each motor of the quadrotor; A safety threshold is determined for the aircraft body posture data, the force data of the mooring point, and the environmental wind field data. When any of the monitoring data exceeds a preset threshold, a working mode switching instruction is output and safe landing control is executed.

2. The method for controlling wind load of a tethered UAV according to claim 1, characterized in that: The force data of multiple tethering points of the tethered UAV, the body posture data and the environmental wind field data are collected and modeled to obtain a system dynamics model, including: Sampling the force signals of the torque sensors of the multiple mooring points to obtain real-time force data of the mooring points, and fusing the data of the three-axis acceleration sensor and angular velocity sensor of the moored drone to obtain the body posture motion data; Performing spatial interpolation on the data collected by the wind speed and direction sensor array in the operating area of ​​the tethered UAV to obtain environmental wind field distribution data; Based on the real-time force data of the mooring point and the environmental wind field distribution data, mooring mechanics calculation is performed to obtain a mooring rope force model, and kinematic analysis is performed on the body posture motion data and the real-time force data of the mooring point to obtain a body posture motion model; Performing aerodynamic moment calculation on the environmental wind field distribution data to obtain a wind load interference model; The mooring rope force model, the body posture motion model and the wind load interference model are combined to obtain a system comprehensive dynamics model, and parameter identification is performed on the system comprehensive dynamics model to obtain a system dynamics model.

3. The method for controlling wind load of a tethered UAV according to claim 2, characterized in that: The state estimation is performed based on the system dynamics model to obtain system wind load disturbance data and expected attitude data, including: Inputting the system dynamics model into a distributed control network for node division to obtain a plurality of subsystem control units, and setting event trigger thresholds for state data of the plurality of subsystem control units to obtain a trigger condition criterion; Inputting the system dynamics model into a reference model to calculate response characteristics to obtain expected dynamic response data, and performing error calculation on actual states of the plurality of subsystem control units according to the expected dynamic response data to obtain state tracking error data; Adaptively updating the state tracking error data to obtain system wind load disturbance data; The state tracking error data and the trigger condition criterion are input into an event judgment module for comparison to obtain a state update trigger instruction; The windward posture of the aircraft body is optimized according to the state update trigger instruction and the system wind load disturbance data to obtain optimal posture angle data, and the wind resistance characteristics of the optimal posture angle data are analyzed to obtain expected posture data.

4. The method for controlling wind load of a tethered UAV according to claim 1, characterized in that: The step of inputting the system wind load disturbance data and the tension distribution control data into a discrete time neural network controller for posture compensation calculation to obtain thrust vector compensation control data comprises: Inputting the system wind load disturbance data and the tension distribution control data into the input layer of the discrete time neural network controller, wherein the input layer includes 12 neuron nodes corresponding to the three-dimensional wind load data and the tension data of the four mooring points respectively; The data of the input layer is processed by the first hidden layer, wherein the first hidden layer includes 24 neuron nodes and uses the ReLU activation function for nonlinear mapping to obtain the first layer feature mapping data; The first layer of feature mapping data is input into the wind load compensation control layer for processing. The wind load compensation control layer includes 36 neuron nodes and uses the Tanh activation function to calculate the required thrust compensation amount to obtain wind load compensation data. Performing second hidden layer processing on the wind load compensation data, the second hidden layer includes 24 neuron nodes, and adopts ReLU activation function to perform feature extraction to obtain second layer feature mapping data; The second layer feature map data is input into the posture tracking control layer for processing, the posture tracking control layer includes 18 neuron nodes, and the Sigmoid activation function is used to calculate the expected posture angle to obtain the posture compensation data; Performing third hidden layer processing on the posture compensation data, the third hidden layer includes 12 neuron nodes, and adopts ReLU activation function to perform dimensionality reduction mapping to obtain third layer feature mapping data; Inputting the third layer feature map data into the thrust distribution layer for processing, the thrust distribution layer comprises 8 neuron nodes, and uses a linear activation function to calculate the thrust component of each rotor to obtain thrust distribution data; The thrust distribution data is processed by an output layer, wherein the output layer includes four neuron nodes and is normalized by using a Tanh activation function to obtain thrust vector compensation control data.

5. The method for controlling wind load of a tethered UAV according to claim 4, characterized in that: The method of performing adaptive robust control on the attitude of the tethered drone according to the thrust vector compensation control data to obtain a speed control instruction for each motor of the quadrotor includes: Inputting the thrust vector compensation control data into a nominal controller for decomposition calculation to obtain desired angle data of roll angle, pitch angle and yaw angle; Performing attitude error calculation on the expected angle data to obtain three-axis attitude angle error and angular velocity error data, and inputting the three-axis attitude angle error and angular velocity error data into an adaptive robust controller for tracking control calculation to obtain an attitude angular acceleration control amount; Performing dynamic inverse mapping on the attitude angular acceleration control quantity to obtain the overall thrust control quantity and three-axis torque control quantity of the quadrotor; Solving the thrust mapping matrix according to the overall thrust control amount of the quadrotor and the three-axis torque control amount to obtain thrust distribution data of the four rotor motors; Inputting the thrust distribution data of the four rotor motors into the characteristic curve for speed mapping to obtain equivalent speed data of the four motors, and performing dead zone compensation and saturation limiting on the equivalent speed data of the four motors to obtain safe speed range data; The safe speed range data is input into the electric speed controller driving module for PWM signal conversion to obtain the speed control instructions of each motor of the quadrotor.

6. The method for controlling wind load of a tethered UAV according to claim 5, characterized in that: The safety threshold judgment is performed on the body posture data, the force data of the mooring point and the environmental wind field data, and when any monitoring data exceeds a preset threshold, a working mode switching instruction is output and a safe landing control is executed, including: Setting a safety threshold for the body posture data to obtain posture threshold range data; Setting a safety threshold for the force data of the mooring point to obtain mooring tension threshold range data; Setting a safety threshold for the environmental wind field data to obtain wind field threshold range data; Inputting the attitude threshold range data, the mooring tension threshold range data and the wind field threshold range data into a monitoring module for multi-threshold judgment to obtain multi-threshold judgment result data; Switching the control mode according to the multi-threshold judgment result data to obtain emergency control mode data; Inputting the emergency control mode data into the attitude protection module for calculation to obtain attitude self-stabilization data, wherein the attitude self-stabilization data includes a desired attitude angle and an angular velocity; The attitude self-stabilization data is used for landing trajectory planning to obtain descent speed control data, and the descent speed control data is used for motor control allocation to obtain safe landing control instructions for each motor of the quadrotor.

7. A control device for wind load resistance of a tethered UAV, characterized in that: Steps for implementing the method for controlling the wind load of a tethered drone according to any one of claims 1 to 6, wherein the control device for the wind load of a tethered drone comprises: A modeling module is used to collect and model the force data of multiple tethering points of the tethered UAV, the body posture data, and the environmental wind field data to obtain a system dynamics model; A state estimation module, used to perform state estimation based on the system dynamics model to obtain system wind load disturbance data and expected attitude data; A solution module, used for performing optimal control solution on the multiple mooring points according to the system wind load disturbance data and the desired attitude data, and obtaining tension distribution control data of each mooring point; A calculation module, used for inputting the system wind load disturbance data and the tension distribution control data into a discrete time neural network controller to perform attitude compensation calculation to obtain thrust vector compensation control data; A control module, used for performing adaptive robust control on the attitude of the tethered UAV according to the thrust vector compensation control data, and obtaining a speed control instruction for each motor of the quadrotor; The judgment module is used to perform safety threshold judgment on the body posture data, the force data of the mooring point and the environmental wind field data. When any monitoring data exceeds a preset threshold, a working mode switching instruction is output and a safe landing control is executed.

8. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method for controlling the wind load of a tethered drone as described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for controlling the wind load of a tethered drone according to any one of claims 1 to 6 are implemented.

Images