A flight control method and system for a biomimetic bird unmanned aerial vehicle

Abstract

The application discloses a kind of bionic bird unmanned aerial vehicle flight control method and system, it is related to unmanned aerial vehicle flight control technical field.The bionic bird unmanned aerial vehicle flight control method, by obtaining the flight state data of bionic bird unmanned aerial vehicle, environmental perception data and flapping motion data;And carry out aerodynamic disturbance stripping processing, generate instantaneous angle of attack;Based on flapping motion data, carry out dynamic aerodynamic compensation processing, generate equivalent resultant angle of attack;Read tail rudder control instruction, and combine flight state data and equivalent resultant angle of attack as joint input;Based on estimation algorithm, carry out joint estimation processing, generate aerodynamic parameter estimation set;It is input to the adaptive control model established in advance and carries out control optimization processing, generates joint control instruction set, the application is handled by joint control instruction set to bionic bird unmanned aerial vehicle, to effectively respond to aerodynamic disturbance in flapping flight, let unmanned aerial vehicle fly more steadily, and improve control precision.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) flight control technology, specifically to a biomimetic bird-inspired UAV flight control method and system. Background Technology

[0002] Bionic flapping-wing UAVs generate lift and thrust by mimicking the flapping motion of birds or insects. Due to their high biomimetic stealth, excellent low-speed / hovering performance and good aerodynamic efficiency, they have shown great application potential in military reconnaissance, environmental monitoring, urban search and rescue and other fields. However, the inherent strong nonlinearity, high coupling and significant unsteady aerodynamic effects of flapping-wing flight make its flight control problem far more complex than that of traditional fixed-wing and rotary-wing UAVs.

[0003] Existing technologies, such as the control method, system, and VTOL ring-wing UAV disclosed in patent application CN121411478A, involve the following steps: gyroscope calibration of the VTOL ring-wing UAV using a three-axis gyroscope based on a flight detection module, followed by takeoff in multi-rotor mode; after the VTOL ring-wing UAV climbs vertically to a preset altitude in multi-rotor mode, a mode switch is performed to change it from multi-rotor to fixed-wing mode; the VTOL ring-wing UAV cruises in fixed-wing mode and receives navigation monitoring information returned by the flight detection module; attitude adjustment is performed on the VTOL ring-wing UAV based on first angle monitoring information and motor speed information; and the motor control logic of the VTOL ring-wing UAV is redistributed to complete the control of the VTOL ring-wing UAV. This application can reduce the difficulty of operating UAVs for pilots and improve the flexibility of UAV flight.

[0004] Based on the above findings, the limitations of existing technologies include at least the following problems: Existing technologies do not construct dedicated control logic for the unsteady aerodynamic characteristics of flapping-wing flight of biomimetic bird drones, resulting in the inability to accurately isolate aerodynamic disturbances to generate an instantaneous angle of attack that fits the flapping-wing condition. The calculated angle of attack deviates significantly from the actual aerodynamic flow state of the drone. Furthermore, existing technologies do not consider the dynamic aerodynamic excitation effect brought about by flapping-wing motion and do not perform dynamic aerodynamic compensation processing on the instantaneous angle of attack, making it difficult to generate an equivalent composite angle of attack that adapts to the unsteady characteristics of flapping-wing motion. The angle of attack characterization cannot reflect the real-time impact of flapping-wing motion on the aerodynamic environment. Additionally, the control law is not optimized for the time-varying aerodynamic characteristics of flapping-wing motion, resulting in insufficient dynamic adaptation of the control model, making it difficult to cope with aerodynamic disturbances and parameter fluctuations during flapping-wing process, thus easily leading to insufficient control accuracy. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a flight control method and system for a biomimetic bird-inspired unmanned aerial vehicle (UAV), which solves the problems of insufficient accuracy caused by existing technologies failing to adapt to the unsteady aerodynamic characteristics of the flapping wings of biomimetic birds, deviations in angle of attack calculations, and inadequate model adaptation.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a flight control method for a biomimetic bird-like unmanned aerial vehicle (UAV), comprising the following steps: acquiring flight status data, environmental perception data, and flapping wing motion data of the biomimetic bird-like UAV; performing aerodynamic disturbance stripping processing on the flight status data and environmental perception data to generate the instantaneous angle of attack of the biomimetic bird-like UAV; and performing dynamic aerodynamic compensation processing on the instantaneous angle of attack based on the flapping wing motion data to generate the equivalent composite angle of attack of the biomimetic bird-like UAV; reading the tail fin control commands of the biomimetic bird-like UAV and combining the flight status data and the equivalent composite angle of attack as joint inputs; performing joint estimation processing on the joint inputs based on an estimation algorithm to generate an aerodynamic parameter estimation set for the biomimetic bird-like UAV, including an estimated value of the equivalent lift line slope and an estimated value of the pitch damping derivative; inputting the estimated value of the equivalent lift line slope and the estimated value of the pitch damping derivative into a pre-established adaptive control model for control optimization processing to generate a joint control command set for the biomimetic bird-like UAV; and performing flight control processing on the biomimetic bird-like UAV based on the joint control command set.

[0007] Further, the flight status data includes pitch angle, pitch rate, airspeed, and vertical linear acceleration, while the environmental perception data includes ambient wind speed, ambient wind direction angle, turbulence intensity, and air density. The specific steps for generating the instantaneous angle of attack of the bionic bird UAV are as follows: Based on the ambient wind speed, ambient wind direction angle, and airspeed, a fusion correction process is performed to generate the effective incoming flow velocity and direction for the bionic bird UAV; based on the effective incoming flow velocity and direction, and combined with the pitch angle, a decoupling calibration process is performed to generate the initial geometric angle of attack of the bionic bird UAV; based on the air density, vertical linear acceleration, pitch rate, and turbulence intensity, a preset unsteady aerodynamic disturbance model is used for identification and fitting to generate the unsteady aerodynamic disturbance correction value for the bionic bird UAV; based on the initial geometric angle of attack and the unsteady aerodynamic disturbance correction value, a disturbance stripping correction process is performed to generate the instantaneous angle of attack of the bionic bird UAV.

[0008] Further, the specific steps for generating the unsteady aerodynamic disturbance correction values ​​for the bionic bird UAV are as follows: The air density, vertical linear acceleration, pitch angular velocity, and turbulence intensity are used to construct the disturbance input vector of the bionic bird UAV; the disturbance input vector is input into a preset unsteady aerodynamic disturbance model, and unsteady disturbance feature extraction processing is performed to generate the aerodynamic disturbance feature vector of the bionic bird UAV; the aerodynamic disturbance feature vector is subjected to feature optimization fitting processing to generate the unsteady aerodynamic disturbance correction values ​​for the bionic bird UAV.

[0009] Further, the specific steps for generating the equivalent composite angle of attack of the bionic bird UAV are as follows: preprocessing the flapping wing motion data; performing flapping wing aerodynamic excitation feature analysis processing based on the preprocessed flapping wing motion data to generate the flapping wing dynamic aerodynamic increment of the bionic bird UAV; and performing compensation fusion processing based on the instantaneous angle of attack and the flapping wing dynamic aerodynamic increment to generate the equivalent composite angle of attack of the bionic bird UAV.

[0010] Furthermore, the specific steps for generating the flapping wing dynamic aerodynamic increment of the bionic bird UAV are as follows: Based on the preprocessed flapping wing motion data, construct the aerodynamic excitation input vector of the bionic bird UAV; input the aerodynamic excitation input vector into a preset flapping wing unsteady aerodynamic analytical model, perform aerodynamic excitation nonlinear analytical processing, and generate the angle of attack increment component of the bionic bird UAV; perform calibration and optimization processing on the angle of attack increment component to generate the flapping wing dynamic aerodynamic increment of the bionic bird UAV.

[0011] Further, the specific steps for generating the aerodynamic parameter estimation set of the bionic bird UAV are as follows: Based on the joint input, construct the pitch regression observation matrix of the bionic bird UAV; based on the estimation algorithm, perform adaptive recursive estimation processing on the regression observation matrix to generate the coarse parameter set of the bionic bird UAV; based on the preset constraints, perform verification processing on the coarse parameter set to generate the aerodynamic parameter estimation set of the bionic bird UAV.

[0012] Furthermore, the specific steps for generating the coarse parameter set of the bionic bird UAV are as follows: setting an initial reference parameter set; based on the pitch regression observation matrix and combined with the initial reference parameter set, using an estimation algorithm to perform a single-step recursive update process; and outputting the coarse parameter set of the bionic bird UAV based on the result of the single-step recursive update process.

[0013] Further, the specific steps for generating the joint control command set of the bionic bird UAV are as follows: The estimated equivalent lift line slope and the estimated pitch damping derivative are input as aerodynamic feedback parameters into a pre-established adaptive control model and corrected; the desired pitch attitude command of the bionic bird UAV is read, and combined with the flight state data, real-time control calculations are performed through the corrected adaptive control model to obtain the basic control set of the bionic bird UAV; the basic control set is constrained and regularized to generate the joint control command set of the bionic bird UAV.

[0014] Further, the specific steps for obtaining the basic control set of the bionic bird UAV are as follows: Based on the modified adaptive control model, and combined with the pitch desired attitude command and the flight state data, the attitude deviation is calculated to obtain the pitch channel control deviation of the bionic bird UAV; the pitch channel control deviation is processed by the control law to obtain the basic control set of the bionic bird UAV.

[0015] A flight control system for a biomimetic bird-like unmanned aerial vehicle (UAV) includes: a data acquisition module for acquiring flight status data, environmental perception data, and flapping wing motion data of the UAV; an aerodynamic disturbance removal module for performing aerodynamic disturbance removal processing on the flight status data and environmental perception data to generate the instantaneous angle of attack of the UAV; an aerodynamic compensation module for performing dynamic aerodynamic compensation processing on the instantaneous angle of attack based on the flapping wing motion data to generate the equivalent composite angle of attack of the UAV; and an aerodynamic parameter estimation module for reading the tail fin control commands of the UAV. Combining the flight state data with the equivalent composite angle of attack, a joint estimation algorithm is used to generate an aerodynamic parameter estimation set for the bionic bird UAV, including the estimated value of the equivalent lift line slope and the estimated value of the pitch damping derivative. An adaptive control optimization module is used to input the estimated value of the equivalent lift line slope and the estimated value of the pitch damping derivative into a pre-established adaptive control model for control optimization processing, generating a joint control command set for the bionic bird UAV. A flight command execution module is used to perform flight control processing on the bionic bird UAV based on the joint control command set.

[0016] The present invention has the following beneficial effects: (1) The biomimetic bird UAV flight control method calculates the initial geometric angle of attack by synchronously collecting flight status, environmental perception and flapping motion data; and combines the unsteady aerodynamic disturbance model to remove the disturbances caused by factors such as air density, vertical linear acceleration, pitch angular velocity and turbulence intensity, so as to obtain an instantaneous angle of attack that is closer to the actual flight; then the flapping motion data is preprocessed to analyze the dynamic aerodynamic increment, and the instantaneous angle of attack is fused and compensated with the increment to finally generate an equivalent composite angle of attack, so that the angle of attack can truly reflect the current aerodynamic flow state of the UAV, avoid subsequent control inaccuracies caused by angle of attack deviation, and make the entire control process more in line with the actual flight conditions.

[0017] (2) The biomimetic bird UAV flight control method first constructs a pitch regression observation matrix by using tail fin control commands, flight status data and equivalent composite angle of attack as joint inputs, and then combines an adaptive recursive estimation algorithm to update the initial reference parameter set step by step to obtain a coarse estimate of the equivalent lift line slope and pitch damping derivative. Then, the coarse estimate parameters are verified by preset constraints to obtain a reliable aerodynamic parameter estimation set, thereby keeping up with the dynamic changes of aerodynamic parameters in flapping flight. The constraint verification ensures the rationality of the parameters and avoids the aerodynamic parameters changing with motion and environment in real time during flapping flight, which is difficult to match with traditional fixed parameters. This allows the estimated aerodynamic parameters to truly reflect the current aerodynamic characteristics of the UAV.

[0018] (3) The biomimetic bird UAV flight control method first corrects the model by inputting the estimated equivalent lift line slope and pitch damping derivative into the adaptive control model, so that the model can adapt to the current aerodynamic state, then reads the pitch desired attitude command, calculates the attitude deviation in combination with flight state data, and obtains the basic control set through control law calculation; then constrains and regularizes the basic control set, and finally generates joint control commands that can be directly issued and executed. This can not only accurately eliminate pitch attitude deviation and make the pitch attitude tracking of the UAV more accurate, but also avoid the servo motor and flapping wing mechanism from jamming or being damaged due to command exceeding the limit. At the same time, it can also effectively deal with aerodynamic disturbances in flapping wing flight, make the UAV flight more stable, and improve control accuracy.

[0019] (4) The flight control system of this bionic bird UAV uniformly acquires the flight status, environmental perception and flapping motion data of the bionic bird UAV through the data acquisition module; the aerodynamic disturbance stripping module completes the aerodynamic disturbance stripping processing based on the acquired multi-source data and outputs the instantaneous angle of attack that is close to the actual flight conditions; the aerodynamic compensation module combines the flapping motion data to carry out dynamic aerodynamic compensation for the instantaneous angle of attack and generate an equivalent composite angle of attack that is adapted to the unsteady flight characteristics of flapping wings; the aerodynamic parameter estimation module integrates the tail fin control commands, flight status data and equivalent composite angle of attack to complete the joint estimation of key aerodynamic parameters such as the equivalent lift line slope and pitch damping derivative; the adaptive control optimization module corrects the adaptive model based on the estimated aerodynamic parameters and generates joint control commands that meet the flight requirements; the flight command execution module completes the flight control of the UAV according to the final command, ensuring the execution efficiency of each processing link, thereby improving the smoothness and stability of the entire control system and thus improving the control accuracy.

[0020] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0021] Figure 1 This is a flowchart of a flight control method for a biomimetic bird-inspired unmanned aerial vehicle according to the present invention.

[0022] Figure 2 This is a flowchart illustrating the specific steps involved in generating the instantaneous angle of attack of a biomimetic bird-shaped drone in the flight control method of this invention.

[0023] Figure 3 This is a block diagram of a biomimetic bird-inspired drone flight control system according to the present invention. Detailed Implementation

[0024] Please see Figure 1This invention provides a technical solution: a flight control method for a biomimetic bird-like unmanned aerial vehicle (UAV), comprising the following steps: acquiring flight status data, environmental perception data, and flapping wing motion data of the biomimetic bird-like UAV; performing aerodynamic disturbance stripping processing on the flight status data and environmental perception data to generate the instantaneous angle of attack of the biomimetic bird-like UAV; and performing dynamic aerodynamic compensation processing on the instantaneous angle of attack based on the flapping wing motion data to generate the equivalent composite angle of attack of the biomimetic bird-like UAV; reading the tail fin control command of the biomimetic bird-like UAV (generated and executed at the previous control moment) (if it is the initial moment, then a preset initial command is used), and combining the flight status data and the equivalent composite angle of attack as joint inputs; performing joint estimation processing on the joint inputs based on an estimation algorithm to generate an aerodynamic parameter estimation set for the biomimetic bird-like UAV, including an estimated value of the equivalent lift line slope and an estimated value of the pitch damping derivative; and inputting the estimated value of the equivalent lift line slope and the estimated value of the pitch damping derivative into a pre-established adaptive control model for control optimization processing to generate a joint control command set for the biomimetic bird-like UAV. Based on the joint control command set, flight control processing is performed on the bionic bird UAV (i.e., tail fin pitch angle command and tail fin deflection rate command are sent to the servo controller of the tail fin servo, and flapping frequency command and flapping amplitude command are sent to the flapping mechanism's driver, such as the motor controller, to control the bionic bird UAV's flight in the next moment), and the generated tail fin control commands are updated and stored as historical values ​​for the next control moment, that is: At the end of each control cycle, the system writes the tail control commands calculated and ultimately output to the actuator (servo) for that cycle—namely, the pitch deflection angle command and the control surface yaw rate command—into the non-volatile memory of the flight control computer or a designated register variable. These stored values ​​physically represent the action commands that are about to be executed. At the start of the next control moment, the control system reads these command values ​​from the storage location and uses them as the control surface deflection angle state of the previous control moment for status feedback.

[0025] Specifically, such as Figure 2 As shown, the flight status data includes pitch angle, pitch rate, airspeed, and vertical linear acceleration; the environmental perception data includes ambient wind speed, ambient wind direction angle, turbulence intensity, and air density. The specific steps for generating the instantaneous angle of attack of the bionic bird UAV are as follows: Based on the fusion and correction of ambient wind speed, ambient wind direction angle, and airspeed, the effective incoming flow velocity and direction (relative to the atmosphere) of the biomimetic bird drone are generated, specifically as follows: Establish the transformation relationship between the UAV's ground coordinate system and its body coordinate system, namely: In the software, two right-handed rectangular coordinate systems are mathematically defined. The ground coordinate system is fixed to the earth and is usually defined as follows: the origin is located at the take-off point, the X-axis points to the geographic north, the Y-axis points to the east, and the Z-axis is perpendicular to the ground and pointing downwards. The body coordinate system is fixed to the UAV body and is defined as follows: the origin is located at the UAV's center of mass, the X-axis points along the longitudinal axis of the body towards the nose, the Y-axis points to the right side of the body, and the Z-axis is perpendicular to the XOY plane and points downward (completing the right-hand rule). The transformation relationship between the two can be completely described by a third-order rotation matrix; The rotation matrix is ​​obtained by multiplying the basic rotation matrices of the UAV’s three current Euler angles—yaw, pitch, and roll—in a predetermined order (e.g., first rotate the yaw angle around the Z-axis, then rotate the pitch angle around the new Y-axis, and finally rotate the roll angle around the new X-axis). The attitude reference system (AHRS) or inertial measurement unit (IMU) on the drone measures and provides these three Euler angles in real time; The flight control software includes a pre-defined function for calculating the rotation matrix and its transpose (inverse transformation) based on this general rule; When it is necessary to convert the wind speed vector from the ground system to the machine system, the program calls this function, calculates the rotation matrix using the currently measured attitude angle, and then performs matrix multiplication with the wind speed vector to obtain the wind speed component in the machine coordinate system. Environmental wind speed relative to ambient wind direction Convert to wind speed vector in ground coordinate system Among them, the environmental wind direction angle The angle between the wind direction and the northward direction of the ground coordinate system, and the airspeed. This is the scalar velocity value of the UAV relative to the atmosphere, with its direction aligned with the longitudinal axis of the UAV, combined with the UAV's current heading angle. With pitch angle Convert the flight airspeed into the UAV velocity vector in the ground coordinate system. The component form is represented as: ; The drone's velocity vector With ambient wind speed vector Vector fusion and measurement error calibration are performed to obtain the effective airflow relative velocity vector of the atmosphere with respect to the UAV. Satisfies the vector relationship: ; Effective airflow relative velocity vector Transform from the ground coordinate system to the body coordinate system to obtain the components in the body coordinate system. Based on this, the effective incoming flow velocity is calculated as follows: ; The effective incoming flow direction is defined as the pitch angle between the incoming flow vector and the transverse axis of the aircraft within the plane of symmetry. : ; Based on the effective incoming flow velocity and direction, and combined with the pitch angle for decoupling calibration, the initial geometric angle of attack of the biomimetic bird UAV is generated, specifically as follows: Based on effective incoming flow velocity Perform aerodynamic effectiveness assessment and set a minimum effective aerodynamic velocity threshold. The determination of this threshold combines experimental and engineering judgments: In wind tunnel experiments, the UAV model is placed in a very low and stable airflow, and the wind speed is gradually reduced while the aerodynamic torque efficiency generated by its control surfaces (such as elevators) is observed and measured. When the wind speed drops below a certain critical value, the torque generated by the control surface will be significantly lower than the noise and interference levels that the flight control system can effectively compensate for, and the attitude control will become unstable. It is necessary to evaluate the measurement error characteristics of airspeed sensors (such as pitot tubes) at low speeds. When the airspeed is below a certain value, the signal-to-noise ratio of the measurement signal will deteriorate sharply, and the data will be unreliable. Based on the test data from the two aspects above, the engineers will select a speed value that is more conservative than the critical speed for pneumatic control failure and the low-precision threshold of the sensor, leaving a safety margin, as the final minimum effective aerodynamic speed threshold. when ≥ If the current atmospheric flow is determined to have effective aerodynamic significance, then the pitch angle of the UAV is used. As the reference for aircraft attitude, the effective pitch angle of the incoming flow Using the atmospheric inflow as a reference, decoupling calculations are performed to isolate the coupling effects of attitude and airflow, thus calibrating the initial geometric angle of attack without additional unsteady disturbances. The calibration relationship is as follows: ,when < At this time, the drone is considered to be in a low-speed, windless state, and the initial geometric angle of attack is directly set to 0; Based on air density, vertical linear acceleration, pitch angular velocity and turbulence intensity, the unsteady aerodynamic disturbance correction values ​​of the bionic bird UAV are generated by identification and fitting through a preset unsteady aerodynamic disturbance model. Based on the initial geometric angle of attack and the unsteady aerodynamic disturbance correction value, disturbance stripping correction is performed to generate the instantaneous angle of attack of the biomimetic bird UAV. Specifically, this involves: adjusting the initial geometric angle of attack... Correction values ​​for unsteady aerodynamic disturbances Unit normalization is performed to convert both to the same unit of angle measurement, i.e., both use angle units. The initial geometric angle of attack after unit normalization is... As the baseline parameter, the unsteady aerodynamic disturbance correction value after unit normalization is deducted. The unsteady aerodynamic disturbances are stripped and corrected, and the disturbance deviation components contained in the initial geometric angle of attack are removed to obtain the instantaneous angle of attack. , .

[0026] The specific steps for generating unsteady aerodynamic disturbance correction values ​​for the biomimetic bird UAV are as follows: The disturbance input vector for the biomimetic bird UAV is constructed using air density, vertical linear acceleration, pitch angular velocity, and turbulence intensity. Specifically, the air density... Vertical linear acceleration Pitch angular velocity turbulence intensity The vectors are normalized and concatenated in a preset order (the normalization process maps the values ​​of each parameter to a uniform range) to construct the perturbation input vector. The vector expression is: ; The disturbance input vector is input into a preset unsteady aerodynamic disturbance model, and unsteady disturbance feature extraction processing is performed to generate the aerodynamic disturbance feature vector of the biomimetic bird UAV, specifically as follows: The unsteady aerodynamic disturbance model is a multivariable linear identification model suitable for biomimetic flapping-wing UAVs, pre-trained using wind tunnel tests and simulation data. The core of the model consists of an input layer, a feature extraction layer, and an output layer. The feature extraction layer contains three independent feature extraction operators, corresponding to the extraction of three types of disturbance features: Vertical motion perturbation feature extraction operator: A first-order inertial coupling operator is adopted. (In the formula, This is the coupling gain coefficient. (The inertial time constant is determined by wind tunnel testing), and the disturbance input vector... air density Vertical linear acceleration Perform coupling operations (coupling relationship is) Then, substitute the values ​​into the operator for filtering and feature mapping to obtain the vertical motion disturbance features. The expression is: ; Pitch and rotation disturbance feature extraction operator: Differential feature operator is used (In the formula, (The differential gain, calibrated from simulation data), for the perturbation input vector pitch angular velocity By performing differential operations, the instantaneous disturbance change characteristics of pitch rotation are extracted, thus obtaining the pitch rotation disturbance characteristics. The expression is: ; Atmospheric turbulence feature extraction operator: Low-pass filter fitting operator (In the formula, For filter gain, The filtering time constant (calibrated jointly by wind tunnel test and simulation data) is applied to the disturbance input vector. Turbulence intensity in Low-pass filtering is performed to remove high-frequency noise interference, and steady-state turbulent disturbance characteristics are extracted to obtain atmospheric turbulent disturbance characteristics. The expression is: ; The constructed perturbation input vector After the model is input, the input layer receives the normalized four-dimensional parameters and passes them to the feature extraction layer. The three independent feature extraction operators are used simultaneously to extract and separate three types of disturbance features: vertical motion, pitch and rotation, and atmospheric turbulence. The output layer then integrates these three types of features according to a fixed dimension to output the aerodynamic disturbance feature vector. (This aerodynamic disturbance eigenvector fully characterizes the response characteristics of various unsteady disturbances under the current flight state and environmental conditions): ; All gain coefficients and time constants of the model are calibrated in the following way: the wind tunnel test simulates disturbance conditions under different flight altitudes and turbulence levels, collects corresponding aerodynamic response data, and combines simulation iterative optimization to finally determine the optimal parameter values ​​of each operator; The aerodynamic disturbance feature vector is subjected to feature optimization fitting to generate unsteady aerodynamic disturbance correction values ​​for the biomimetic bird UAV, specifically as follows: Characteristics of vertical motion disturbance Pitch and rotation disturbance characteristics Atmospheric turbulence characteristics To address the cross-coupling effects between different disturbance features, an orthogonal decomposition decoupling algorithm is used to remove interference. A feature coupling correction matrix is ​​constructed, which is a third-order square matrix whose elements are coupling coefficients calibrated through wind tunnel coupling tests, representing the degree of cross-influence between different disturbance features. aerodynamic disturbance eigenvectors Performing matrix multiplication with the inverse of the coupling correction matrix yields independent and uncoupled pure eigenvectors. , ; Introducing a feature weight coefficient matrix pre-determined through multiple sets of wind tunnel calibration tests and digital simulation iterations. The steps to determine this are as follows: In the wind tunnel, a large number of combined perturbation tests covering the expected flight envelope of the UAV were designed and implemented. For example, under different reference wind speeds, turbulence of different intensities and frequencies was superimposed, and the model was made to perform pitch oscillations of different amplitudes and frequencies, thereby creating a rich variety of known unsteady aerodynamic perturbation scenarios. In each scenario, the true aerodynamic perturbation correction value was directly measured or obtained by back-solving the motion response using high-precision force measuring equipment (such as strain gauge) as the standard answer. At the same time, the output values ​​of three perturbation feature extraction operators calculated in real time based on sensor data were recorded. The collected feature values ​​are combined with the corresponding standard answer data pairs and processed using multiple linear regression or other fitting algorithms (such as least squares method) to solve for a set of weight coefficients that minimize the overall error between the model's predicted values ​​and the measured values. The initial set of weighting coefficients was placed into a digital simulation environment containing a high-fidelity UAV dynamics model to test its performance in more complex and variable virtual flight missions. Based on the simulation results, the weighting coefficients were fine-tuned and optimized to ensure that it remained robust under boundary conditions not covered by wind tunnel testing. This matrix is ​​a row vector, expressed as: ,in The dimensionless influence weights for the vertical motion perturbation characteristics are given. The dimensionless influence weight of the pitch rotation disturbance characteristics. The dimensionless influence weights of atmospheric turbulence disturbance characteristics are represented by all weight coefficients, which are non-negative real numbers and satisfy normalization constraints. The weight coefficient matrix With pure feature vectors Perform a weighted dot product operation to obtain the weighted fitting median value. The operational relationship is as follows: ; Introduce pre-calibrated static deviation compensation values The compensation value is determined by statistically analyzing the difference between the model output and the wind tunnel measured aerodynamic disturbance data. The specific steps are as follows: In the wind tunnel test where the weighting coefficients are determined, for each test condition, after calculating the median value of the weighted fit, it is compared with the actual aerodynamic disturbance correction value obtained by precise measurement under that condition, and a difference is calculated. This difference reflects a fixed, systematic bias in the model under that specific condition, in addition to the main feature influence represented by the weighting coefficients. Collect these deviation data under all operating conditions, and then perform statistical analysis; The average value of these deviations is usually calculated, or their stable offset is observed under the main flight conditions. The average value or stable offset obtained by statistics is the systematic static deviation of the model. It is stored as a static deviation compensation value, also as a fixed constant, along with the weighting coefficient; During real-time flight, each time the disturbance correction value is calculated, after completing the weighted summation, the fixed compensation value is directly added to offset the inherent bias of the model at the system level, thereby improving the overall estimation accuracy. To eliminate systematic numerical bias in the above steps, the weighted fitting intermediate value is used. Compensation value for static deviation Algebraic superposition calibration is performed to obtain accurate correction values ​​for unsteady aerodynamic disturbances. ,Right now: .

[0027] In this implementation plan, the ground coordinate system and the airframe coordinate system are first defined in a standardized manner. A rotation matrix is ​​constructed based on Euler angles to achieve vector transformation. The ambient wind speed, wind direction, and airspeed are then fused into a unified vector calculation to accurately obtain the effective incoming flow velocity and direction of the UAV. This avoids the calculation deviation caused by mixing data from different coordinate systems. At the same time, a minimum effective aerodynamic velocity threshold is set by combining wind tunnel tests and sensor measurement characteristics to distinguish between effective aerodynamics and low-speed, windless states. This makes the calibration of the initial geometric angle of attack more consistent with actual engineering scenarios. Then, by constructing a disturbance input vector, the vertical motion, pitch rotation, and atmospheric turbulence disturbance features are extracted using a preset model. After decoupling, weighted fitting, and static deviation compensation, various unsteady aerodynamic interferences are completely removed. The final instantaneous angle of attack eliminates the disturbance deviation caused by environmental and airframe motion, which can truly reflect the current aerodynamic incoming flow state of the UAV. The data accuracy is higher and it is more suitable for the complex working conditions of flapping-wing flight.

[0028] Specifically, the flapping motion data includes flapping frequency (the number of times the UAV wing completes up-and-down flapping cycles per unit time), flapping amplitude (the maximum deflection angle during the flapping process), and flapping phase angle (the instantaneous position of the wing within a single flapping cycle). The specific steps for generating the equivalent composite angle of attack of the biomimetic bird UAV are as follows: The flapping wing motion data is preprocessed as follows: a first-order real-time lag filtering algorithm is used to perform real-time noise reduction on the raw flapping wing motion data. The filtering formula is as follows: ; In the formula, For the current moment The raw flapping motion data (frequency / amplitude / phase angle) collected. This is the real-time filtered data at the current moment. The filtering result from the previous time step, the filter coefficients A value of 0.9 is used to filter out high-frequency noise from the sensor and instantaneous interference from mechanical vibration of the fuselage while preserving real-time motion characteristics. Based on the preprocessed flapping wing motion data, flapping wing aerodynamic excitation feature analysis is performed to generate flapping wing dynamic aerodynamic increments of the biomimetic bird UAV. Based on the instantaneous angle of attack and flapping wing dynamic aerodynamic increments, a compensation and fusion process is performed to generate the equivalent composite angle of attack for the biomimetic bird UAV. Specifically, it refers to the instantaneous angle of attack. Using the reference angle, the dynamic aerodynamic increment of the flapping wing For non-steady correction quantities, real-time compensation and fusion calculations are performed, i.e.: ,in, For dynamic compensation coefficients; The steps are as follows: Read the design flapping frequency range of the bionic bird UAV stored in the database (determined during the design phase), and design the wind tunnel calibration conditions using the controlled variable method: Set the flapping frequency to be within the design flapping frequency range, and make it a gradient variable (no less than 5 sets of gradients). Fix parameters such as incoming flow velocity and flapping amplitude, and repeat the test 3 times for each set of conditions; measure the instantaneous angle of attack using a five-component balance in the wind tunnel. Equivalent composite angle of attack The corresponding relationship is used to simultaneously record the dynamic aerodynamic increments of the flapping wings. With the goal of minimizing the fitting error, the piecewise linear mapping relationship between the flapping frequency and the compensation coefficient was obtained using the least squares method; the flapping frequency Substituting the above-defined piecewise linear mapping relationship, the current dynamic compensation coefficient can be obtained through single-step direct calculation. .

[0029] The specific steps for generating the flapping wing dynamic aerodynamic increments of the biomimetic bird UAV are as follows: Based on the preprocessed flapping wing motion data, the aerodynamic excitation input vector of the biomimetic bird UAV is constructed, specifically as follows: flapping frequency (Unit: Hz), flapping amplitude (Unit: rad) Real-time flapping wing phase angle (Unit: rad), while also introducing the effective incoming flow velocity. (Unit: m / s) As the associated parameters of the aerodynamic excitation, the aerodynamic excitation input vector is constructed according to a preset order, and the vector expression is: ; The aerodynamic excitation input vector is input into a preset flapping-wing unsteady aerodynamic analytical model, and aerodynamic excitation nonlinear analytical processing is performed to generate the angle of attack increment component of the biomimetic bird UAV, which is as follows: The flapping wing unsteady aerodynamic analytical model is a lightweight unsteady aerodynamic model adapted for real-time flight control. This model is based on the classic Theodorsen unsteady lift theory, whose core is the unsteady circulation function, used to describe the unsteady aerodynamic effects generated by airflow vortex shedding and added mass during the periodic motion of the wing surface. Its standard circulation function expression is a real-time solvable form. This model directly adopts the real-time algebraic approximation of this function (avoiding complex variable function calculations and meeting the requirements of airborne real-time computation), namely: ; In the formula, To reduce the frequency in real time, The chord length of the wing surface of the biomimetic bird drone (an inherent structural parameter of the drone, which is pre-fixed in the flight control system and cannot be fitted in real time). , For the real and imaginary parts of the Theodorsen function, this model, to achieve real-time computation, uses a wind tunnel-verified real-time algebraic fitting formula to replace the complex function solution. The fitting formula is pre-calibrated and requires no iterative computation, specifically: , ; aerodynamic excitation input vector After inputting the model, Calculate the instantaneous angular velocity of the wing surface based on the real-time flapping frequency. With real-time flapping phase angle Combining the simple harmonic motion characteristics of flapping wing motion, the instantaneous rotational angular velocity of the wing surface is calculated in real time. It represents the angular velocity of the wing surface at the current moment and is the core input of the wing surface motion in Theodorsen's theory. The calculation formula is: This formula is directly derived from the kinematic relationship of flapping wing simple harmonic motion; Real-time calculation of unsteady aerodynamic excitation intensity of flapping wing, combined with flapping wing amplitude Effective incoming flow velocity With the above instantaneous rotational angular velocity of the wing surface It also introduces pre-calibrated inherent structural parameters of the biomimetic bird drone: wing chord length. (Unit: m, constant parameters determined during the UAV design phase and embedded in the flight control system), comprehensively calculate the real-time unsteady aerodynamic excitation intensity of the flapping wings. It integrates the wing surface motion amplitude, motion rate, and relative airflow velocity, directly corresponding to the wing surface motion excitation term in Theodorsen's theory. It characterizes the equivalent aerodynamic circulation excitation generated by the wing surface during real-time flapping. The calculation formula is: ; The core unsteady aerodynamic term in the computational model is based on real-time frequency reduction. Call the pre-pre-defined Theodorsen real part In addition to the wing area S and the total mass m of the UAV, calculate the unsteady circulation-induced aerodynamic terms. This term corresponds to the angle of attack increment generated by the circulation lift produced by the periodic motion of the wing surface. The calculation formula is as follows: ; Analysis of the basic angle of attack increment components : ; Corresponding to the Theodorsen unsteady circulation angle of attack increment, The calibration coefficient for the circulation effect; Corresponding to the periodic aerodynamic lag effect of flapping wing phase These are the phase effect calibration coefficients; Corresponding to the incremental mass angle of attack at high flapping frequencies, This is the calibration coefficient for the additional mass effect.

[0030] The calibration process is as follows: Multiple sets of gradient test conditions are designed using the controlled variable method. Based on the design specifications of the bionic bird UAV, steady-state setpoints for flapping frequency, flapping amplitude, and wind tunnel inflow velocity are set. Each parameter takes no less than 5 sets of gradient values, which are combined to form no less than 125 sets of independent test conditions, covering the entire flight envelope and the entire flapping working range of the UAV, to ensure the comprehensiveness of the calibration results. For each set of steady-state test conditions, the UAV prototype was kept fixed in the wind tunnel test section, and the wing surface was driven to flap periodically according to set parameters. The actual angle of attack increment corresponding to the unsteady aerodynamic forces on the wing surface was measured and recorded in real time using a five-component aerodynamic balance built into the wind tunnel. Simultaneously, the model input parameters corresponding to each set of conditions were collected, including the flapping frequency setpoint, flapping amplitude setpoint, flapping phase angle, and wind tunnel inflow velocity, and the corresponding parameters were calculated in real time according to the model formulas mentioned above. All collected data retains the original physical quantities without any normalization. Each set of working conditions is repeated 3 times. After removing abnormal deviation data, the average value is taken as the valid sample to ensure data reliability. The least squares method was used to solve for the coefficients, with the optimization objective being to minimize the sum of squared errors between the experimentally measured angle-of-attack increment and the model-calculated base angle-of-attack increment. A system of linear equations was constructed and solved directly using matrix inversion to obtain the coefficient values ​​that satisfy the optimal fit. After the solution was completed, it was verified whether the dimensions of each coefficient met the preset constraints. The coefficient is dimensionless. The dimension is rad. Dimensions are ; The angle-of-attack increment component is calibrated and optimized to generate the flapping-wing dynamic aerodynamic increment of the biomimetic bird UAV. Specifically, this involves reading the maximum design flight speed of the UAV stored in the database. And combined with effective incoming flow velocity Generate operating condition calibration factors , ; Use operating condition calibration factor For the basic angle of attack increment component Real-time dynamic corrections are performed to obtain the dynamic aerodynamic increment of the flapping wing. , .

[0031] In this implementation scheme, the original flapping wing motion data is processed by first-order real-time hysteresis filtering. While fully preserving the core motion characteristics of flapping wing frequency, amplitude, and phase angle, it effectively filters out high-frequency noise from sensors and instantaneous interference from fuselage mechanical vibration, ensuring the stability and reliability of the input data. Then, the flapping wing motion parameters and effective incoming flow velocity are integrated to construct an aerodynamic excitation input vector. The aerodynamic excitation analysis is completed by relying on a lightweight unsteady aerodynamic analytical model adapted for airborne computing. The model adopts a simplified fitting form to replace complex calculations, balancing computational accuracy and flight control real-time performance. At the same time, the model coefficients are calibrated through multiple sets of wind tunnel gradient tests covering the entire flight envelope to ensure that the analytical results conform to the actual aerodynamic laws. Subsequently, compensation fusion is carried out by combining the instantaneous angle of attack. The dynamic compensation coefficient of the wind tunnel fitting and the operating condition calibration factor are used for double correction. The final equivalent composite angle of attack not only retains the core characteristics of the basic aerodynamic incoming flow, but also fully integrates the dynamic aerodynamic changes generated by flapping wing motion, which can more realistically reflect the real-time aerodynamic state of the UAV.

[0032] Specifically, the tail control commands are tail pitch deflection angle and tail deflection rate. The specific steps for generating the aerodynamic parameter estimation set of the biomimetic bird UAV are as follows: Based on the joint input, a pitch regression observation matrix for the biomimetic bird UAV is constructed. Specifically, this involves extracting observations from the joint input that are strongly correlated with the aerodynamic characteristics of the pitch channel, namely, the equivalent composite angle of attack. Tail fin pitch deflection Tail fin deflection rate Pitch angular velocity The above observations are combined in an orderly manner to generate an elevation regression observation matrix. ,Right now: ; Based on the estimation algorithm, the regression observation matrix is ​​subjected to adaptive recursive estimation processing to generate a coarse parameter set for the bionic bird UAV. Based on preset constraints, the coarsely estimated parameter set is validated to generate an aerodynamic parameter estimation set for the biomimetic bird UAV, specifically as follows: The constraint is: the physically feasible range for the rough estimate of the slope of the equivalent lift line is set as follows: The physical feasible range for a rough estimate of the pitch damping derivative is: Then the mathematical expression for the parameter constraint is: ; The coarse estimates of the equivalent lift slope and pitch damping derivative from the coarse parameter set are substituted into the constraints mentioned above for judgment. If the coarse estimates exceed the corresponding interval, they are corrected to the interval boundary values; if all constraints are satisfied, they are directly retained. The corrected output parameter results are the final aerodynamic parameter estimation set of the biomimetic bird UAV, i.e., the estimated equivalent lift slope values. Estimated pitch damping derivative .

[0033] It should be noted that the equivalent lift line slope range values , The method for obtaining the equivalent lift line slope is as follows: First, based on the airfoil parameters (such as airfoil type and relative thickness), wing geometry parameters (aspect ratio and tip-to-root ratio) and the motion characteristics of the flapping wing mechanism of the biomimetic bird UAV, and referring to the known lift line slope range of similar flapping wing / fixed-wing aircraft in the aviation field, the basic range of the equivalent lift line slope is initially determined; among which... The minimum lift line slope that allows the airfoil to generate effective lift is taken as the benchmark value. The baseline value of the maximum lift line slope before the airfoil experiences aerodynamic stall is taken.

[0034] Pitch Damping Derivative Interval Values , Methods for obtaining the data: Aerodynamic design benchmark calibration: Based on the aerodynamic layout parameters of the biomimetic bird UAV, such as fuselage layout, tail size, and the relative position of flapping wings and fuselage, and referring to the known physical range of pitch damping derivatives of similar aircraft, the range is initially determined; since the physical essence of the pitch damping derivative is the damping effect that suppresses the pitch angular velocity, its value is naturally negative. Take the maximum negative reference value (maximum absolute value) with the strongest damping effect. Take the smallest negative reference value with the weakest damping effect (the absolute value is the smallest, close to 0).

[0035] The specific steps for generating the coarse parameter set of the biomimetic bird UAV are as follows: Set the initial reference parameter set, which is specifically: the initial reference parameter set is a priori vector of aerodynamic parameters calibrated based on wind tunnel tests, and is only enabled at the initial control moment. Define the initial reference parameter vector as follows: Its mathematical expression is: In the formula, The equivalent lift line inclination is the a priori reference value. The prior reference value for the pitch damping derivative is set, and the initial estimation error covariance matrix is ​​also set. , is a symmetric positive definite constant matrix, representing the confidence level of the initial prior parameters. If it is not the initial time, it is composed of the estimated value of the equivalent lift line slope and the estimated value of the pitch damping derivative at the previous time. Based on the pitch regression observation matrix and combined with the initial reference parameter set, an estimation algorithm is used for single-step recursive update processing. Specifically, this estimation algorithm can be an adaptive recursive least squares algorithm, that is: the aerodynamic parameter vector to be estimated is defined as... And calculate the recursive gain: ; In the formula, Let be the recursive gain matrix at the current time. The error covariance matrix is ​​estimated for the previous time step. Forgetting factor; Single-step update of aerodynamic parameters: ; In the formula, Output the observation values ​​for the pitch channel at the current moment. The parameter estimation results from the previous time step (initial time t=0, ); Error covariance matrix update: ; In the formula, The identity matrix is ​​used at the initial time, with the initial reference parameter vector. With the initial covariance matrix Based on this, perform the above single-step recursive calculation once; for non-initial control times, use the recursive result from the previous time step. , Based on the current elevation regression observation matrix, Complete single-step update; Based on the single-step recursive update process, a coarse parameter set of the bionic bird UAV is output, namely: extracting the parameter estimation vector at the current time. The corresponding elements in the rough estimate parameter set are expressed and output in the following mathematical form: ;in, For a rough estimate of the slope of the equivalent lift line, A rough estimate of the pitch damping derivative is given.

[0036] In this implementation scheme, firstly, observations strongly correlated with the aerodynamic characteristics of the pitch channel, such as the equivalent combined angle of attack and tail rudder deflection, are selected to construct a pitch regression observation matrix. This effectively avoids interference from irrelevant data and ensures the relevance of the observation input. The initial reference parameters adopt prior values ​​calibrated by wind tunnel tests. For non-initial moments, the estimated results from the previous moment are used as the reference, and a single-step update is performed through an adaptive recursive algorithm. Combined with the forgetting factor and the error covariance matrix, this not only keeps up with the dynamic changes of aerodynamic parameters during flapping-wing flight in real time but also ensures the rationality and stability of the update process. At the same time, the coarsely estimated parameters are verified by a preset physical feasible interval. The interval values ​​are determined by combining the UAV's own airfoil, layout parameters, and references to similar aircraft, which can eliminate outliers that exceed the physical range. The final estimated values ​​of the equivalent lift line slope and pitch damping derivative not only conform to the actual aerodynamic laws but also accurately match the current flight state, avoiding control failure due to parameter deviations.

[0037] Specifically, the steps for generating the joint control command set of the biomimetic bird UAV are as follows: The estimated values ​​of the equivalent lift line slope and pitch damping derivative are input as aerodynamic feedback parameters into the pre-established adaptive control model, and then corrected. Specifically: The adaptive control model employs a structure combining a linear quadratic regulator (LQR) and a dynamic inversion, defining the augmented control state vector as follows: , For pitch angle tracking error, The integral term of this deviation is used to eliminate steady-state error. For the externally input pitch angle reference command (provided by the pitch desired attitude command generated and executed in the previous control moment), construct the augmented system state space equations: ; in, , ; The aerodynamic derivative coefficients in the matrix are defined as follows: , , , , , in, , The known constant aerodynamic derivatives calibrated by wind tunnel tests are obtained by fixing a biomimetic bird UAV model in a wind tunnel and measuring the changes in lift and pitch moment generated by the UAV when the tail fin deflection or the model angle of attack changes, while controlling other variables such as wind speed and model attitude. By performing linear regression analysis on the measurement data, the derivatives of lift and pitch moment with respect to the fin deflection angle or angle of attack are calculated, thus obtaining the specific values ​​of the aforementioned constant aerodynamic derivatives. The time constant is the time required for the servo surface deflection angle to reach 63.2% of the steady-state value after inputting a step voltage signal to the servo motor. The typical value of the time constant is usually between 0.01 seconds and 0.1 seconds, and the specific value is determined according to the selected servo motor model. For pitch rotational inertia, by updating and To correct the matrix In , and ; The pitch moment of inertia is obtained in the following way: During the detailed design phase of the UAV, based on the final three-dimensional digital model, the mass attribute analysis function of the computer-aided design software is used to automatically calculate the moment of inertia about the Y-axis (pitch axis) of the airframe coordinate system after setting the material density or actual mass of each component. After the drone is manufactured, physical experimental methods can be used to verify or measure it in order to obtain more accurate values. A common method is the compound pendulum method. The drone is suspended by a blade or bearing to form a compound pendulum, allowing it to swing freely in the pitch plane with a small amplitude. Its swing period is measured with a high-precision timer. Based on the physical formula between the compound pendulum period and the moment of inertia, the moment of inertia of the UAV about the suspension axis can be calculated, and then the pitch moment of inertia about its own center of mass can be obtained by converting it through the parallel axis theorem. By comparing theoretical calculations with experimental measurements, a final accurate value is determined and used as the inherent parameter of the UAV. Based on the state-space equation of this augmented system By solving and The problem of linear quadratic regulators (LQR) with given parameters is solved online by calculating the optimal state feedback gain matrix. In this context, the state weight matrix and input weight matrix of the LQR problem are preset as follows: and Therefore, based on real-time aerodynamic parameter estimation, the adaptive control model (i.e., feedback gain) is completed. Correction of ) Pre-setting the weight matrix: First, based on flight quality requirements, the relative magnitudes of these weights are initially set. For example, if closer pitch tracking is desired, the weights corresponding to the pitch angle error and its integral term are increased; if smoother control actions and energy saving are desired, the control weights are increased. These initially set weights will undergo extensive testing on a hardware-in-the-loop simulation platform. The simulation platform connects a real flight control computer and simulated servo and sensor models. Under various simulated flight scenarios and fault modes, the control performance is observed, and the weights are repeatedly adjusted based on the simulation results until the optimal performance that meets all design specifications is obtained. The final set of weight matrices and coefficients is used as the optimal configuration parameters. Online calculation of the gain matrix: In each control cycle, after the system state matrix is ​​updated using the latest aerodynamic parameter estimates, the flight control program calls a built-in, optimized Riccati equation solver to ensure real-time performance. The solver takes the currently updated system matrix, control matrix, and fixed weight matrix as inputs, and calculates the optimal state feedback gain matrix at the current moment through a fixed numerical calculation process (e.g., iterative solution or explicit method based on matrix operations). This process is a self-contained numerical calculation that does not depend on additional online learning or iterative convergence, and its output is directly used to construct the control law for the current period. It reads the pitch desired attitude command (generated and executed in the previous control moment) of the bionic bird UAV, combines it with flight status data, and performs real-time control calculation through the modified adaptive control model to obtain the basic control set of the bionic bird UAV. The basic control set is constrained and normalized to generate a joint control instruction set for the biomimetic bird UAV, specifically as follows: The basic control set includes: the desired deflection rate command of the basic tail control surface calculated by the adaptive control model. and flapping wing frequency compensation amount And flapping amplitude compensation amount ; Tail fin control surface command generation and amplitude limiting: Rate limiting is applied to the base expected deflection rate command: , The maximum allowable deflection rate of the servo can be obtained from the technical specifications of the tail servo product. The servo manufacturer will clearly indicate the maximum no-load angular velocity of the servo model under rated operating voltage, usually in degrees per second. When designing the flight control system, the servo model that meets the overall control performance requirements will be selected, and its nominal maximum angular velocity value will be converted to a safe limit value and written into the software after unit conversion (such as to radians per second). In order to leave a margin and protect the servo, 90% to 95% of the nominal value will be used as the actual limit value used in the software. Based on the tail fin pitch deflection and Calculate the expected tail fin pitch deflection at the next moment: ; Desired tail fin pitch deflection Position limiting: , and The physical deflection limit of the control surface is determined through motion simulation of a 3D model during the mechanical design phase. This allows for precise analysis of the maximum angle before interference occurs between the control surface and the tail stabilizer, fuselage, or other components during deflection. This angle is the maximum permissible deflection angle mechanically. After the UAV is physically assembled, the limit positions of the control surface's left and right swings need to be measured manually or automatically to verify the theoretical value and confirm the absence of interference. Finally, the measured maximum left and right deflection angles (usually negative for left and positive for right) are used as the hard limit of the control surface deflection angle command and written into the flight control software. Any calculated command that exceeds this limit will be forcibly truncated to the boundary value to prevent mechanical damage. Flapping Parameter Command Synthesis and Limiting: Reading Flapping Frequency flapping amplitude ; Calculate flapping frequency command flapping amplitude command ; Limit the flapping wing command: , ,in, , , , The values ​​are, in order, the minimum and maximum flapping frequencies, the minimum flapping amplitude, and the maximum flapping amplitude stored in the database. Taking the minimum and maximum flapping frequencies as an example: You can consult the manufacturer's rated parameter table for the flapping wing drive motor (brushless motor or servo) to obtain the minimum safe speed at which the motor can stably output torque and the maximum safe speed to avoid overheating and burning out; then measure the actual reduction ratio of the flapping wing transmission mechanism (gear, connecting rod or crank-slider) (you can directly determine this by counting the number of gear teeth and looking at the transmission ratio of the connecting rod), and directly convert the motor's speed range into the basic frequency range of the flapping wing mechanism to obtain the minimum and maximum values ​​of the flapping wing frequency; Generate and output a unified control instruction set: .

[0038] The specific steps to obtain the basic control set of the bionic bird drone are as follows: Based on the modified adaptive control model, and combined with the desired pitch attitude command and flight state data, the attitude deviation is calculated to obtain the pitch channel control deviation of the bionic bird UAV, which is as follows: Extract pitch angle from flight status data and pitch angular velocity ; Calculate pitch tracking error and update its points items. (Using numerical integration methods); Read the tail fin pitch angle (Execution feedback from the previous control cycle); The pitch channel control deviation constitutes the augmented state vector used for control calculation. ; The control law is calculated and processed to obtain the basic control set of the bionic bird UAV, which is as follows: Tail rudder system command calculation: using the updated optimal state feedback gain matrix Calculate the basic tail fin rudder system commands (i.e., the desired deflection rate). ; Flapping-wing control compensation calculation: Flapping-wing control is used to provide additional lift compensation and attitude adjustment. The compensation is calculated based on the state vector. ; ; in, , , , , , The preset flapping wing control gain coefficient was determined through wind tunnel testing and simulation parameter tuning, namely: Based on the design flight envelope of the bionic bird drone, select no less than 5 typical flight speeds and no less than 5 typical flapping frequencies to form no less than 25 benchmark flight conditions. Under each benchmark flight condition, the drone is made to reach a stable trim state through wind tunnel testing. In the trim state of each reference flight condition, a preset excitation signal is applied to the pitch channel. The excitation signal includes, but is not limited to, pitch angle step signal, frequency sweep signal or pseudo-random sequence. The following data are collected and recorded simultaneously during the excitation application process: pitch angle, pitch rate, equivalent composite angle of attack, pitch tracking error, and corresponding flapping frequency command and flapping amplitude command in the time series. The optimization cost function is defined as the weighted sum of squared integrals of the pitch tracking error under all test conditions. Minimizing this cost function is the optimization objective. Using the flapping wing control gain coefficients as optimization variables, a numerical optimization algorithm (such as gradient descent, genetic algorithm, or sequential quadratic programming) is employed for iterative solution to obtain a set of gain coefficient values ​​that optimize the system's pitch tracking performance. , , , , , ; Finally, the basic control set is obtained: .

[0039] In this implementation scheme, the adaptive control model combines LQR and dynamic inverse structure, using previously estimated aerodynamic parameters to correct the model matrix in real time. Simultaneously, the pitch inertia is confirmed through both 3D design and physical experiments, and the weight matrix is ​​optimized through simulation parameter tuning to ensure the model always adapts to the flight state. When solving the basic control set, steady-state error is eliminated through attitude deviation integral terms. Combining flight state data and desired attitude, accurate tail servo and flapping wing compensation commands are obtained. In the subsequent constraint regularization stage, servo amplitude limiting references manufacturer parameters and mechanically measured values, while flapping wing amplitude limiting combines motor characteristics and transmission mechanism parameters, effectively avoiding damage caused by commands exceeding the actuator's capabilities. The final generated joint control command set balances the accuracy of pitch attitude control with the safety of command execution, and can adapt to the dynamic changes of flapping wing flight, making the UAV's pitch attitude control smoother and command execution more reliable, fully meeting actual flight requirements.

[0040] Please see Figure 3 This invention provides a technical solution: a flight control system for a bionic bird-like unmanned aerial vehicle (UAV), comprising: a data acquisition module for acquiring flight status data, environmental perception data, and flapping wing motion data of the UAV; an aerodynamic disturbance removal module for performing aerodynamic disturbance removal processing on the flight status data and environmental perception data to generate the instantaneous angle of attack of the UAV; an aerodynamic compensation module for performing dynamic aerodynamic compensation processing on the instantaneous angle of attack based on the flapping wing motion data to generate the equivalent composite angle of attack of the UAV; and an aerodynamic parameter estimation module for reading the parameters of the UAV. The system generates a set of aerodynamic parameters for the bionic bird UAV by combining tail control commands with flight status data and equivalent composite angle of attack through an estimation algorithm. These parameters include estimates of the equivalent lift line slope and pitch damping derivative. An adaptive control optimization module inputs these estimates into a pre-established adaptive control model for control optimization, generating a set of joint control commands for the bionic bird UAV. A flight command execution module performs flight control processing on the bionic bird UAV based on the joint control command set.

[0041] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0042] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A flight control method for a biomimetic bird-inspired unmanned aerial vehicle, characterized in that, Includes the following steps: Acquire flight status data, environmental perception data, and flapping wing motion data of the bionic bird drone; Aerodynamic disturbance stripping is performed on the flight status data and environmental perception data to generate the instantaneous angle of attack of the bionic bird UAV. Based on flapping wing motion data, dynamic aerodynamic compensation processing is performed on the instantaneous angle of attack to generate the equivalent composite angle of attack of the bionic bird UAV; Read the tail fin control commands of the bionic bird UAV and combine the flight status data with the equivalent composite angle of attack as joint inputs; Based on the estimation algorithm, the joint input is subjected to joint estimation processing to generate an aerodynamic parameter estimation set for the biomimetic bird UAV, including the estimated value of the equivalent lift line slope and the estimated value of the pitch damping derivative. The estimated equivalent lift line slope and pitch damping derivative are input into a pre-established adaptive control model for control optimization, generating a joint control command set for the bionic bird UAV. Based on the aforementioned joint control instruction set, flight control processing is performed on the bionic bird drone.

2. The biomimetic bird unmanned aerial vehicle flight control method according to claim 1, characterized in that, The flight status data includes pitch angle, pitch rate, airspeed, and vertical linear acceleration; the environmental perception data includes ambient wind speed, ambient wind direction angle, turbulence intensity, and air density. The specific steps for generating the instantaneous angle of attack of the bionic bird UAV are as follows: Based on the environmental wind speed, the environmental wind direction angle, and the airspeed, a fusion correction process is performed to generate the effective incoming flow velocity and direction of the bionic bird drone. Based on the effective incoming flow velocity and the incoming flow direction, and combined with the pitch angle, a decoupling calibration process is performed to generate the initial geometric angle of attack of the bionic bird UAV. Based on the air density, vertical linear acceleration, pitch angular velocity, and turbulence intensity, the unsteady aerodynamic disturbance correction values ​​of the bionic bird UAV are generated by identification and fitting through a preset unsteady aerodynamic disturbance model. Based on the initial geometric angle of attack and the unsteady aerodynamic disturbance correction value, disturbance stripping correction processing is performed to generate the instantaneous angle of attack of the bionic bird UAV.

3. The biomimetic bird unmanned aerial vehicle flight control method according to claim 2, characterized in that, The specific steps for generating unsteady aerodynamic disturbance correction values ​​for the biomimetic bird drone are as follows: The perturbation input vector of the biomimetic bird UAV is constructed using the air density, the vertical linear acceleration, the pitch angular velocity, and the turbulence intensity. The disturbance input vector is input into a preset unsteady aerodynamic disturbance model, and unsteady disturbance feature extraction processing is performed to generate the aerodynamic disturbance feature vector of the bionic bird UAV. The aerodynamic disturbance feature vector is subjected to feature optimization fitting to generate unsteady aerodynamic disturbance correction values ​​for the bionic bird UAV.

4. The biomimetic bird unmanned aerial vehicle flight control method according to claim 1, characterized in that, The specific steps for generating the equivalent composite angle of attack of the bionic bird drone are as follows: The flapping wing motion data is preprocessed; Based on the preprocessed flapping motion data, flapping aerodynamic excitation feature analysis is performed to generate flapping dynamic aerodynamic increments of the bionic bird UAV. Based on the instantaneous angle of attack and the flapping wing dynamic aerodynamic increment, a compensation and fusion process is performed to generate the equivalent composite angle of attack of the bionic bird UAV.

5. The biomimetic bird unmanned aerial vehicle flight control method according to claim 4, characterized in that, The specific steps for generating the flapping wing dynamic aerodynamic increments of the biomimetic bird drone are as follows: Based on the preprocessed flapping motion data, an aerodynamic excitation input vector for the bionic bird UAV is constructed. The aerodynamic excitation input vector is input into a preset flapping-wing unsteady aerodynamic analytical model, and aerodynamic excitation nonlinear analytical processing is performed to generate the angle of attack increment component of the bionic bird UAV. The angle of attack increment component is calibrated and optimized to generate the flapping wing dynamic aerodynamic increment of the bionic bird UAV.

6. The biomimetic bird unmanned aerial vehicle flight control method according to claim 1, characterized in that, The specific steps for generating the aerodynamic parameter estimation set of the biomimetic bird drone are as follows: Based on the joint input, a pitch regression observation matrix for the bionic bird UAV is constructed; Based on the estimation algorithm, the regression observation matrix is ​​subjected to adaptive recursive estimation processing to generate a coarse parameter set for the bionic bird UAV. Based on preset constraints, the coarse parameter set is verified to generate an aerodynamic parameter estimation set for the bionic bird UAV.

7. The biomimetic bird unmanned aerial vehicle flight control method according to claim 6, characterized in that, The specific steps for generating a coarse parameter set for the biomimetic bird drone are as follows: Set the initial baseline parameter set; Based on the pitch regression observation matrix and the initial reference parameter set, an estimation algorithm is used for single-step recursive update processing. Based on the single-step recursive update process results, a coarse set of parameters for the bionic bird drone is output.

8. The biomimetic bird unmanned aerial vehicle flight control method according to claim 1, characterized in that, The specific steps for generating the joint control instruction set for the bionic bird drone are as follows: The estimated equivalent lift line slope and the estimated pitch damping derivative are used as aerodynamic feedback parameters and input to the pre-established adaptive control model for correction. It reads the pitch and attitude command of the bionic bird UAV, combines it with the flight state data, and performs real-time control calculation through the modified adaptive control model to obtain the basic control set of the bionic bird UAV. The basic control set is constrained and normalized to generate a joint control instruction set for the bionic bird UAV.

9. The biomimetic bird unmanned aerial vehicle flight control method according to claim 8, characterized in that, The specific steps to obtain the basic control set of the bionic bird drone are as follows: Based on the modified adaptive control model, and combined with the pitch desired attitude command and the flight state data, the attitude deviation is calculated to obtain the pitch channel control deviation of the bionic bird UAV. The control law is calculated and processed to obtain the basic control set of the bionic bird UAV by solving the control deviation of the pitch channel.

10. A flight control system for a biomimetic bird-inspired unmanned aerial vehicle (UAV), employing the flight control method for a biomimetic bird-inspired UAV as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire flight status data, environmental perception data, and flapping wing motion data of the bionic bird drone. The aerodynamic disturbance removal module is used to perform aerodynamic disturbance removal processing on the flight status data and environmental perception data to generate the instantaneous angle of attack of the bionic bird UAV. The aerodynamic compensation module is used to perform dynamic aerodynamic compensation processing on the instantaneous angle of attack based on flapping wing motion data, and generate the equivalent composite angle of attack of the bionic bird UAV. The aerodynamic parameter estimation module is used to read the tail fin control commands of the bionic bird UAV, and combine the flight state data with the equivalent composite angle of attack, and perform joint estimation processing through the estimation algorithm to generate an aerodynamic parameter estimation set for the bionic bird UAV, including the estimated value of the equivalent lift line slope and the estimated value of the pitch damping derivative. The adaptive control optimization module is used to input the estimated value of the equivalent lift line slope and the estimated value of the pitch damping derivative into the pre-established adaptive control model for control optimization processing, and generate a joint control command set for the bionic bird UAV. The flight command execution module is used to perform flight control processing on the bionic bird drone based on the joint control command set.

Images