A wind-resistant robust flight control method and system for quadrotor unmanned aerial vehicles

Abstract

This invention proposes a robust wind-disruption-resistant flight control method and system for quadrotor unmanned aerial vehicles (UAVs), belonging to the field of UAV control technology. The method includes: acquiring real-time state data of the UAV, including position, velocity, altitude, three-axis attitude angles, and three-axis angular velocities; obtaining a desired inspection trajectory command and generating a desired attitude angle command and a total lift command based on the real-time state data; using a neural network to estimate and compensate for wind disturbances in the desired attitude angle command and the total lift command; calculating the control torque increment based on the desired attitude angle command and the real-time state data using an incremental backstepping method, and obtaining the control torque command by combining it with the control torque at the previous moment; distributing the control torque command and the total lift command to the rotational speeds of the four motors of the quadrotor UAV, with the motors executing the speed commands to drive the UAV to complete the wind-disruption-resistant flight mission. This invention can actively adapt to wind field changes and ensure that the UAV can still fly accurately, stably, and safely under disturbances.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) control technology, and particularly relates to a wind-resistant robust flight control method and system for quadrotor UAVs. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Due to its flexibility and efficiency, drone technology is widely used in the daily inspection of photovoltaic power plants to replace traditional manual inspections and automatically identify panel faults such as hot spots and cracks. However, photovoltaic power plants are usually located in open areas with complex wind conditions, including stable crosswinds, unpredictable gusts, and canyon wind effects generated between photovoltaic arrays.

[0004] These wind disturbances can severely disrupt the flight stability of drones, causing them to deviate from the preset precision inspection route and resulting in positioning errors. More seriously, violent shaking can cause images captured by the onboard visible light or infrared cameras to become blurry and shaky, severely affecting the accuracy of fault diagnosis, and may even lead to collisions between the drone and expensive photovoltaic panels, resulting in significant cost losses.

[0005] Currently, most photovoltaic inspection drones employ traditional PID or linear quadratic (LQR) control methods. These methods heavily rely on the precise mathematical model of the drone during design and assume a calm flight environment. However, wind disturbances are difficult to model accurately and are time-varying external disturbances. This leads to traditional controllers based on fixed parameters exhibiting insufficient robustness and poor anti-interference capabilities in practice, making it difficult to ensure flight safety while maintaining inspection accuracy. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, this invention proposes a robust wind-disruption-resistant flight control method and system for quadrotor UAVs, used for automatic inspection of photovoltaic power plants. By combining adaptive neural network wind disturbance compensation with incremental finite-time backstepping control, the robust wind-disruption-resistant hierarchical control method overcomes the shortcomings of existing technologies, such as strong dependence on accurate models and weak wind disturbance resistance. This enables the UAV to quickly suppress disturbances, accurately track the preset inspection trajectory, and strictly ensure a safe distance between itself and the photovoltaic panels to prevent collisions in complex wind field environments.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0008] In a first aspect, the present invention discloses a wind-resistant robust flight control method for a quadcopter unmanned aerial vehicle, comprising:

[0009] Collect real-time status data of the UAV, including position, speed, altitude, three-axis attitude angles and three-axis angular velocities;

[0010] The desired inspection trajectory command is obtained and combined with the real-time status data to generate the desired attitude angle command and total lift command. A neural network is used to estimate wind disturbance and compensate for the desired attitude angle command and total lift command.

[0011] Based on the desired attitude angle command and real-time status data, the incremental backstepping method is used to calculate the control torque increment, and the control torque command is obtained by combining the control torque at the previous moment.

[0012] The control torque command and the total lift command are distributed to the rotational speeds of the four motors of the quadcopter drone, and the motors execute the rotational speed commands to drive the drone to complete the anti-wind disturbance flight mission.

[0013] Secondly, this invention discloses a wind-resistant and robust flight control system for a quadcopter unmanned aerial vehicle (UAV), comprising:

[0014] The data acquisition module is configured to collect real-time status data of the UAV, including position, speed, altitude, three-axis attitude angles and three-axis angular velocities.

[0015] The outer loop control module is configured to: acquire the desired inspection trajectory command and generate the desired attitude angle command and total lift command by combining the real-time status data; and use a neural network to estimate wind disturbance and compensate for the desired attitude angle command and total lift command.

[0016] The inner loop control module is configured to: calculate the control torque increment based on the desired attitude angle command and real-time status data using an incremental backstepping method, and obtain the control torque command by combining the control torque at the previous moment;

[0017] The control distribution module is configured to distribute the control torque command and the total lift command to the rotational speeds of the four motors of the quadcopter UAV, and the motors execute the rotational speed commands to drive the UAV to complete the anti-wind disturbance flight mission.

[0018] Thirdly, the present invention discloses an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when run by the processor, complete the steps of the above-mentioned wind-resistant robust flight control method for quadcopter UAVs.

[0019] Fourthly, the present invention discloses a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the above-described wind-resistant robust flight control method for quadcopter UAVs.

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

[0021] This invention possesses excellent wind disturbance resistance performance. Through a composite anti-disturbance mechanism of SPNN adaptive compensation and incremental control, it forms a multi-level suppression capability against wind disturbance, which can effectively cope with stable crosswinds, sudden gusts and canyon wind effects in photovoltaic power plants, and significantly improve the clarity and stability of inspection images.

[0022] This invention possesses strong robustness and adaptability. The incremental control method significantly reduces the dependence on precise mathematical models, and the SPNN compensation mechanism can adaptively adjust wind disturbance estimation, making the system highly robust to changes in UAV parameters and external wind disturbances, thus reducing debugging and maintenance costs.

[0023] This invention provides absolute safety assurance. By strictly constraining the position tracking error through a barrier Lyapunov function and combining it with finite-time convergence characteristics, it ensures that the drone and the photovoltaic panel always maintain a safe distance, fundamentally preventing collision accidents.

[0024] This invention features rapid dynamic response performance. Finite-time control ensures that the system state converges within a finite time, while incremental control provides rapid response capability, significantly reducing the duration of deviations caused by sudden wind disturbances and improving inspection efficiency.

[0025] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0026] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0027] Figure 1 This is a flowchart of the wind-resistant robust flight control method for a quadcopter UAV as described in Embodiment 1 of the present invention.

[0028] Figure 2 This is a structural diagram of the inner and outer loop controllers of the inspection drone described in Embodiment 1 of the present invention.

[0029] Figure 3 This is a diagram of the SPNN neural network structure described in Embodiment 1 of the present invention. Detailed Implementation

[0030] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0031] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0032] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0033] Example 1

[0034] In one or more embodiments, a wind-resistant robust flight control method for a quadcopter unmanned aerial vehicle (UAV) is disclosed, such as... Figures 1-2 As shown, it includes the following steps:

[0035] Step S1: Collect real-time status data of the inspection drone itself through airborne sensors (GPS, IMU, barometer, etc.), including three-dimensional position, speed, altitude, three-axis attitude angles (roll, pitch, yaw) and three-axis angular velocity.

[0036] Step S2: The outer loop finite-time adaptive position controller acquires the desired inspection trajectory command and generates the desired attitude angle command (including the desired roll angle) by combining the real-time status data fed back by the sensors. Desired pitch angle Desired yaw angle ) and total lift command The controller employs a Sigma-Pi neural network (SPNN) to estimate wind disturbance and compensate for the desired attitude angle command and total lift command. A barrier Lyapunov function (BLF) is integrated into the controller to ensure that the position tracking error never exceeds a preset safety constraint. Specifically:

[0037] In the outer loop position controller, a Sigma-Pi neural network (SPNN) is used for wind disturbance estimation and compensation, such as... Figure 3 As shown, the SPNN compensator consists of an input layer, a product layer, and a summation layer connected in sequence.

[0038] Input layer: Receives the system state vector The system has 15 dimensions, including 3D position error, 3D velocity error, 3D desired position, 3D desired velocity, and 3D desired acceleration. The system state vector Z is calculated using real-time data and preset desired inspection trajectory commands, considering position error, velocity error, etc., and ultimately integrated into the system state vector Z. This is a 3D position error vector, representing the deviation between the actual position and the desired position. This is a 3D velocity error vector, representing the deviation between the actual velocity and the expected velocity. Given a 3D desired position vector, the position that the drone should reach in the preset inspection trajectory; The desired 3D velocity vector represents the speed that the drone should maintain during the pre-defined inspection trajectory. It is a 3-dimensional desired acceleration vector, which is the acceleration that the UAV should have in the preset inspection trajectory.

[0039] Productive layer: Divides the input vector into 3 sub-networks (5 dimensions per group), higher-order coupling is achieved through Kronecker product, with a number of neurons. ,in, , , These are the location-dependent wind disturbance component estimation neurons, velocity-dependent wind disturbance component estimation neurons, and acceleration-dependent wind disturbance component estimation neurons, respectively. The total number of neurons is 125.

[0040] The basis functions of the Sigma-Pi neural network are constructed using higher-order coupling achieved through the Kronecker product, as follows:

[0041]

[0042] Here, kron represents the Kronecker product operation, applied to the basis functions of the three subnetworks. , , Perform nested Kronecker product operations to generate a basis function vector containing all possible combinations of the input variables' products. .

[0043] Summation layer: The outputs of the product layer are weighted and summed to output the wind disturbance estimate. Weight vector The dimension is consistent with the number of neurons in the product layer.

[0044] The summation layer can be expressed as:

[0045]

[0046]

[0047] The first equation is the output equation of the SPNN. For the neural network weight vector, Given a basis function vector, their inner product is the estimate of the wind disturbance. The second equation is the adaptive update law for the weights. >0 represents the learning rate matrix, which determines the speed of weight updates. The larger the learning rate, the faster the weights converge, but it may cause system oscillations. A value greater than 0 represents a small positive number, used to prevent weight drift during the learning process and improve system stability. For velocity error. The adaptive law is designed so that the weights can be adjusted according to the velocity error. Adjustments are made by increasing the weight update magnitude when the estimation error is large, thereby gradually reducing the estimation error.

[0048] The basis function vectors use Gaussian radial basis functions:

[0049]

[0050] in, As the center of the basis functions, For width parameter, This represents the number of basis functions (i.e., the number of neurons in the product layer).

[0051] This invention proposes an adaptive wind disturbance compensation mechanism based on a Sigma-Pi Neural Network (SPNN) for photovoltaic power plants operating in complex wind-disturbed environments. Compared to traditional RBF neural networks, SPNN constructs basis functions through Kronecker products, enabling it to better capture the high-order nonlinear coupling between wind disturbance and system state. The SPNN weights are updated online using an improved adaptive law. This mechanism can quickly and accurately estimate and compensate for time-varying wind disturbances, significantly improving the system's anti-interference capability and tracking accuracy in complex wind field environments.

[0052] The outer-loop finite-time adaptive position controller receives preset inspection trajectory commands (desired position, velocity, and acceleration) and actual state information from sensor feedback. By calculating position and velocity errors, it generates the desired attitude angle command. Based on the desired trajectory and actual state, the outer-loop controller outputs the total lift and the desired attitude angle. The core formulas are as follows:

[0053]

[0054]

[0055]

[0056] in, For total lift command, For the quality of drones, These are the actual roll angle and the actual pitch angle. The combined control variables in the x, y, and z directions are defined as follows:

[0057]

[0058]

[0059]

[0060] in, , The position error is in the x, y, and z directions, which is the difference between the expected position and the actual position. The velocity error is in the x, y, and z directions, which is the difference between the expected velocity and the actual velocity. Let x, y, and z be the desired accelerations. The position and velocity proportional gain are used to adjust the error feedback intensity. , Let be the parameters that converge in finite time, and This is used to accelerate error convergence. The wind disturbance estimate is calculated and compensated by an SPNN neural network. g represents the acceleration due to gravity. Trajectory tracking is achieved through position / velocity error feedback and a finite-time convergence term. Combined with wind disturbance estimation to compensate for external disturbances, the final output is the total lift. The desired attitude angle drives the inner-loop attitude control.

[0061] Wind disturbance is a major factor affecting the stability of photovoltaic power plant inspection drones. It exhibits nonlinearity, time-varying characteristics, and uncertainty, making it difficult to accurately describe using traditional models. Therefore, this invention employs a Sigma-Pi neural network (SPNN) in the outer-loop position controller for wind disturbance estimation and compensation. Compared to traditional RBF neural networks, this invention demonstrates stronger nonlinear approximation capabilities and faster convergence speed.

[0062] Furthermore, to strictly ensure the safe distance between the drone and the photovoltaic panel, positional error constraints are defined:

[0063]

[0064] in, To preset a safety threshold, this embodiment uses 1 / 3 of the photovoltaic panel spacing to construct a position error barrier Lyapunov function (BLF):

[0065]

[0066] in, For positional error, As a safety threshold, Speed ​​error, Let Lyapunov be the barrier function for position error, used to limit the position error to a preset safety threshold, where i = x, y, z correspond to the three motion axes of the UAV. The first term is the barrier function term, which is applied when the position error... Approaching the safety threshold When the position error is within the bounds, the first term tends to infinity to prevent the error from exceeding the bounds; the second term is the square of the velocity error, used to ensure the dynamic performance of the system. The entire BLF function is positive definite, and its value remains finite only when the position error does not exceed the bounds.

[0067] right Taking the derivative and combining it with the position controller described above, we can determine the following:

[0068]

[0069] Among them, the summation term Yes The upper bound estimate. , , These are the position proportional gain, velocity proportional gain, and finite-time convergence parameters in three directions, respectively. These are parameters that converge in finite time. This represents the absolute value of the positional error in the x, y, and z directions. Let be the absolute value of the velocity error in the x, y, and z directions. The formula indicates that... It is negative definite, which means It will monotonically decrease over time and eventually converge to a stable value. Due to the initial time... It is finite and it always decreases, therefore the position error Never reach the safety threshold A safe distance was strictly maintained between the drone and the photovoltaic panel, effectively avoiding the risk of collision.

[0070] Step S3: The inner loop incremental finite-time attitude controller receives the desired attitude command calculated by the outer loop and the real-time status data fed back by the sensor. Based on the desired attitude angle command and the real-time status data, the incremental backstepping method is used to calculate the control torque increment. Combined with the control torque at the previous moment, the control torque command is obtained, thereby achieving fast and robust attitude stabilization.

[0071] The inner-loop attitude controller employs an incremental backstepping control method, directly utilizing angular acceleration information measured by the IMU, significantly reducing the reliance on precise mathematical models. Specifically:

[0072] The inner-loop attitude control employs a finite-time incremental backstepping control method, directly utilizing the actual angular acceleration measured by sensors. To design a controller, the angular velocity is numerically differentiated to obtain the angular acceleration.

[0073] Define attitude angle error ,in, As desired, This is the actual posture.

[0074] Traditional backstepping control is based on model calculation of angular acceleration:

[0075]

[0076] In the formula, For traditional angular acceleration, Here is the rotational inertia matrix. To control the torque, For the Coriolis force and centrifugal force terms, This is a gravitational coupling term. Traditional methods require precise knowledge of... and These model terms, whose calculations depend on accurate attitude and angular velocity information, are prone to errors when subjected to wind disturbances or parameter changes, leading to a deterioration in control performance.

[0077] The control torque is expressed as:

[0078]

[0079] in, The control torque at the previous moment, This represents the increment of the control torque at the current moment. By focusing on the control torque at the current moment, the controller only needs to consider the difference between the current moment and the previous moment, without having to calculate the absolute control quantity, thus reducing the dependence on the model.

[0080] The increment of the control torque is expressed as:

[0081]

[0082] In the formula, All are positive definite gain matrices, used to adjust the weights of attitude error, attitude angular velocity error, and finite-time convergence term, respectively; It is a finite-time exponent, and , used to ensure the fast convergence characteristics of the system; sign() is the sign function; This refers to the attitude angle error; This is the actual angular acceleration. This refers to the attitude angular velocity error. Here is the rotational inertia matrix; This is the change in angular acceleration; This represents the change in attitude angle error. This represents the change in attitude angular velocity error.

[0083] in, It is used to compensate for changes in angular acceleration and directly reflects changes in the system's dynamics; and Used to suppress changes in attitude error and angular velocity error; This is used to accelerate the convergence of errors and ensure that stability is achieved within a finite time.

[0084] in This represents the increment between the current value and the previous value, specifically defined as:

[0085]

[0086] In the formula, The control cycle is typically equal to the sensor's sampling period. By calculating this series of increments, the inner-loop controller can promptly sense the rate of change in the system state, adjust the control output, and respond to wind disturbances.

[0087] This invention abandons the traditional model-based backstepping control method and proposes an incremental finite-time control framework. The core of this method lies in directly utilizing the angular acceleration information measured in real time by the IMU to achieve attitude stabilization through incremental control calculations, significantly reducing the dependence on complex attitude dynamics models. The incremental design allows the controller to know only the system's moment of inertia matrix, while possessing inherent robustness to changes in other model parameters. Combined with a finite-time convergence term, it ensures that attitude errors converge rapidly within a finite time, significantly improving the system's response speed and disturbance rejection capability.

[0088] Furthermore, the incremental backstepping control is discretized to obtain the final control torque;

[0089] Practical UAV control systems are digital control systems, requiring the implementation of control algorithms in the discrete time domain. Let the sampling period be... Therefore, the incremental control law needs to be discretized. In the discrete time domain, the increment of the attitude angular velocity error can be approximated by the difference of the attitude error (i.e., Therefore, the discretized incremental control law is:

[0090]

[0091] In the formula, The torque increment is controlled at time k. This represents the current discrete time, where each term represents a value taken at a discrete time point. The sampling period is [value]. The final control torque output is obtained by adding the control torque from the previous moment to the control increment at the current moment. Simultaneously, to prevent motor actuator saturation (i.e., the output torque exceeding the motor's maximum capacity), the control quantity needs to be limited. The final control quantity is:

[0092]

[0093] In the formula, Let k be the torque control quantity. The maximum torque that the motor can provide is used as the threshold. When the calculated control torque exceeds this value, it is truncated. This is to protect the motor and prevent system instability.

[0094] Preferably, in order to ensure that the attitude error converges to zero within a finite time, the controller parameters need to be designed appropriately, mainly including finite time exponential... and gain matrix .Pick ,when When the value approaches 1, the system's convergence characteristics approach asymptotic stability, and the control output is relatively smooth; when... As the value approaches 0.5, the convergence speed increases. (Gain matrix) Must meet:

[0095]

[0096] In the formula, The moment of inertia is in the direction of roll. This represents the maximum amplitude of the wind disturbance moment in the roll direction. It ensures that the gain is large enough and suppresses disturbances, driving error convergence.

[0097] Step S4: Distribute the control torque command and total lift command to the rotational speed of the four motors of the quadcopter drone. The motors execute the rotational speed command to drive the drone to complete the anti-wind disturbance flight mission.

[0098] Based on the above process, this invention creates a hierarchical robust control architecture that combines outer-loop SPNN compensation with inner-loop incremental control. The outer loop is responsible for position tracking and wind disturbance compensation, while the inner loop is responsible for attitude stabilization and rapid response. The two work together to form a composite disturbance rejection mechanism. A barrier Lyapunov function strictly ensures that the position tracking error never exceeds a preset safety constraint, fundamentally preventing collisions with photovoltaic panels. Stability proofs based on Lyapunov theory ensure the finite-time convergence and global stability of the entire closed-loop system, providing theoretical support for safe inspection. This enables proactive adaptation to wind field changes, ensuring that the UAV can still fly accurately, stably, and safely under disturbances.

[0099] This invention has broad application prospects. It is not only applicable to photovoltaic power plant inspection, but can also be extended to other industrial drone application scenarios with extremely high requirements for flight safety and stability, such as power line inspection, bridge inspection, and wind turbine inspection.

[0100] Example 2

[0101] In one or more embodiments, a wind-resistant robust flight control system for a quadcopter unmanned aerial vehicle is disclosed, specifically including:

[0102] The data acquisition module is configured to collect real-time status data of the UAV, including position, speed, altitude, three-axis attitude angles and three-axis angular velocities.

[0103] The outer loop control module is configured to: acquire the desired inspection trajectory command and generate the desired attitude angle command and total lift command by combining the real-time status data; and use a neural network to estimate wind disturbance and compensate for the desired attitude angle command and total lift command.

[0104] The inner loop control module is configured to: calculate the control torque increment based on the desired attitude angle command and real-time status data using an incremental backstepping method, and obtain the control torque command by combining the control torque at the previous moment;

[0105] The control distribution module is configured to distribute the control torque command and the total lift command to the rotational speeds of the four motors of the quadcopter UAV, and the motors execute the rotational speed commands to drive the UAV to complete the anti-wind disturbance flight mission.

[0106] Example 3

[0107] This embodiment provides an electronic device, including a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the computer instructions are executed by the processor, they complete the steps of the above-described method for robust flight control of a quadcopter UAV against wind disturbance.

[0108] Example 4

[0109] This embodiment provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the above-described wind-resistant robust flight control method for quadcopter UAVs.

[0110] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0111] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0112] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0113] The descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0114] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A wind disturbance robust flight control method for a quadrotor unmanned aerial vehicle, characterized in that, The method comprises the following steps: Collecting real-time state data of the unmanned aerial vehicle, wherein the real-time state data comprises position, speed, height, three-axis attitude angle and three-axis angular velocity; Obtaining a desired inspection trajectory instruction and combining the real-time state data to generate a desired attitude angle instruction and a total lift instruction, and using a neural network to estimate wind disturbance and compensate the desired attitude angle instruction and the total lift instruction; Based on the desired attitude angle instruction and the real-time state data, an incremental backstepping method is used to calculate a control torque increment, and a control torque instruction is obtained by combining the control torque at the previous moment; The control torque instruction and the total lift instruction are distributed to the rotating speeds of the four motors of the quad-rotor unmanned aerial vehicle, and the motors execute the rotating speed instructions to drive the unmanned aerial vehicle to complete the wind disturbance resistant flight task; The desired inspection trajectory instruction and the real-time state data are combined to generate a desired attitude angle instruction and a total lift instruction, specifically: wherein, is a total lift command, is a UAV mass, is an actual roll angle, an actual pitch angle, is an x, y, z direction integrated control amount, is a wind disturbance estimate, is a desired pitch angle, is a desired yaw angle; The incremental backstepping method for calculating the control torque increment comprises: wherein are positive definite gain matrices; is a finite time exponent; sign( ) is a sign function; is an attitude angle error; is an actual angular acceleration, is an attitude angular velocity error; is an inertia matrix; is an angular acceleration variation; is an attitude angle error variation; is an attitude angular velocity error variation.

2. The wind disturbance robust flight control method for a quad-rotor UAV of claim 1, wherein, The neural network uses a Sigma-Pi neural network, which comprises an input layer, a product layer and a summation layer connected in sequence; The input layer receives a system state vector; The product layer receives the system state vector and is divided into three groups of sub-networks, and high-order coupling is realized through Kronecker product; The summation layer performs weighted summation on the output of the product layer to output the wind disturbance estimation value. 3.The wind disturbance robust flight control method for a quad-rotor UAV of claim 1, wherein, The desired attitude angle instruction and the total lift instruction are defined as position error constraints: wherein, is a predetermined safety threshold; A position error barrier Lyapunov function is constructed: wherein, is a position error, is a safety threshold, is a velocity error, is a barrier Lyapunov function for the position error, i = x, y, z corresponding to the three axes of motion of the UAV.

4. The wind disturbance robust flight control method for a quad-rotor UAV of claim 1, wherein, The control torque instruction is obtained by combining the control torque at the previous moment: wherein, is the control torque of the previous time instant, is the control torque increment of the current time instant, is the control torque of the current time instant.

5. The wind disturbance robust flight control method for a quadrotor UAV of claim 4, wherein, The control torque instruction is discretized, and the discretized incremental control law is: wherein, is the control torque increment at time k, denotes the current discrete time, denotes the increment of the current time value and the previous time value, is the moment of inertia matrix, is the angular acceleration variation; is the attitude angle error variation; is the attitude angular velocity error variation, is the angular velocity error, is the attitude angle error, and sign() is the sign function, are positive definite gain matrices, is the finite time exponent, is the sampling period; the final control torque output is obtained by adding the control torque at the previous time and the control increment at the current time, and the control amount is limited in amplitude, and the final control amount is: In the formula, is the moment control amount at time k, is the maximum moment that the motor can provide.

6. A wind disturbance robust flight control system for a quadcopter drone, the system comprising: The method comprises the following steps: A data acquisition module configured to collect real-time state data of the unmanned aerial vehicle, wherein the real-time state data comprises position, speed, height, three-axis attitude angle and three-axis angular velocity; An outer loop control module configured to obtain a desired inspection trajectory instruction and combine the real-time state data to generate a desired attitude angle instruction and a total lift instruction, and use a neural network to estimate wind disturbance and compensate the desired attitude angle instruction and the total lift instruction; An inner loop control module configured to calculate a control torque increment based on the desired attitude angle instruction and the real-time state data using an incremental backstepping method, and obtain a control torque instruction by combining the control torque at the previous moment; A control distribution module configured to distribute the control torque instruction and the total lift instruction to the rotating speeds of the four motors of the quad-rotor unmanned aerial vehicle, and the motors execute the rotating speed instructions to drive the unmanned aerial vehicle to complete the wind disturbance resistant flight task; The desired inspection trajectory instruction and the real-time state data are combined to generate a desired attitude angle instruction and a total lift instruction, specifically: wherein, is a total lift command, is a UAV mass, is an actual roll angle, an actual pitch angle, is an x, y, z direction integrated control amount, is a wind disturbance estimate, is a desired pitch angle, is a desired yaw angle; The incremental backstepping method for calculating the control torque increment comprises: wherein are positive definite gain matrices; is a finite time exponent; sign( ) is a sign function; is an attitude angle error; is an actual angular acceleration, is an attitude angular velocity error; is an inertia matrix; is an angular acceleration variation; is an attitude angle error variation; is an attitude angular velocity error variation.

7. An electronic device, comprising: The computer instructions stored in the memory and running on the processor complete the quad-rotor unmanned aerial vehicle wind disturbance resistant robust flight control method of any one of claims 1-5 when the processor runs the computer instructions.

8. A computer-readable storage medium, characterized in that, The computer instructions are used to store the computer instructions, and the computer instructions are executed by the processor to complete the quad-rotor unmanned aerial vehicle wind disturbance resistant robust flight control method of any one of claims 1-5.

Images