Pan-tilt preview feedforward and body maneuver decoupling cooperative control method for power transmission inspection

Abstract

The application discloses a gimbal pre-look feedforward and body maneuver decoupling cooperative control method for power transmission inspection, comprising the following steps: acquiring a body maneuver instruction sequence to be executed by a UAV, and synchronously collecting multi-source state parameters of the body and the gimbal; dynamically estimating the body response delay based on the multi-source state parameters containing the load, the voltage and the wind field, and determining the starting timing of the gimbal pre-look compensation in combination with the instruction characteristics; simultaneously, analyzing the track geometric feature variation of the maneuver instruction sequence, establishing a mapping relationship from the track geometric feature to the gimbal dynamics demand, and generating a gimbal feedforward control instruction; and based on the starting timing, injecting the gimbal feedforward control instruction into the gimbal control loop to drive the gimbal to execute the pre-look compensation action. Through the multi-physical field delay estimation and the track geometric feedforward, the gimbal shaking problem caused by the pre-look timing deviation due to the working condition change and the curvature mutation is solved, and precise image stabilization under large maneuver is realized.

Description

Technical Field

[0001] This invention relates to the field of precision control technology for unmanned aerial vehicles (UAVs), and in particular to a decoupled and coordinated control method for gimbal aiming feedforward and airframe maneuvering for power transmission inspection. Background Technology

[0002] In the scenario of refined power transmission line inspection, multi-rotor UAVs need to perform rapid maneuvers such as circling towers and changing lanes along high-voltage lines. Maintaining absolute stability of the onboard gimbal's line of sight is a crucial technical prerequisite for acquiring high-quality inspection images and ensuring accurate defect identification. Under highly dynamic flight conditions, overcoming the disturbances caused by the aircraft's motion to the gimbal and achieving motion decoupling between the gimbal's line of sight and the aircraft's maneuvers are key technical aspects for improving the efficiency and data quality of automated inspection operations.

[0003] Existing UAV gimbal stabilization control technologies mostly employ a feedback control architecture based on a high-frequency inertial measurement unit (IMU), with some solutions supplemented by feedforward compensation based on the real-time angular velocity of the aircraft. These solutions typically rely on sensors detecting attitude disturbances or changes in angular velocity before the gimbal controller generates a counter-motion command to compensate; or they simply superimpose a feedforward quantity based on the current motion state while the aircraft is performing an action. This type of method primarily relies on passive response or immediate compensation based on the aircraft's current motion state, focusing on suppressing disturbances that have already occurred.

[0004] However, existing technologies suffer from spatiotemporal response mismatches under complex and variable inspection conditions, primarily manifested in the dynamic nonlinearity of the aircraft's response delay and the lack of perception of trajectory geometry. Specifically, existing pre-aiming or feedforward control systems often assume a fixed constant for the aircraft's response delay, neglecting the impact of battery voltage drops, load mass changes, and environmental wind resistance on the aircraft's dynamic response speed. This results in pre-aiming compensation actions often being ahead or behind, failing to achieve precise timing synchronization. Existing control strategies lack analysis of higher-order trajectory geometry features, focusing only on the current angular velocity or angular error, ignoring the instantaneous demand on the gimbal motor's angular acceleration Jerk due to sudden changes in trajectory curvature (such as at the moment of entering a curve). This lack of dynamic feedforward information makes the gimbal prone to mechanical vibration or tracking overshoot when responding to drastic curvature changes, such as the aircraft transitioning from a straight line to a curve, making it difficult to ensure a smooth transition of the line of sight. Summary of the Invention

[0005] The purpose of this invention is to provide a decoupled and coordinated control method for gimbal aiming feedforward and body maneuvering for power transmission inspection, in order to solve the above-mentioned problems in the prior art.

[0006] Technical solution: A decoupled and coordinated control method for gimbal aiming feedforward and airframe maneuvering for power transmission line inspection, comprising:

[0007] Acquire the sequence of maneuver commands to be executed by the UAV, and simultaneously collect the current multi-source status parameters of the UAV's body and gimbal;

[0008] Based on the dynamic estimation of the aircraft response delay using multi-source state parameters, and combined with the command characteristics of the aircraft maneuver command sequence, the activation timing of the gimbal aiming compensation is determined.

[0009] The changes in the trajectory geometry of the aircraft's maneuver command sequence are analyzed, a mapping relationship between trajectory geometry and gimbal dynamic requirements is established, and gimbal feedforward control commands are generated.

[0010] Based on the startup timing, the gimbal feedforward control command is injected into the gimbal control loop to drive the gimbal to perform aiming compensation action.

[0011] Beneficial effects: This invention solves the gimbal jitter problem caused by pre-aiming timing deviation and curvature abrupt change due to changes in operating conditions by using multi-physics delay estimation and track geometry feedforward, thus achieving accurate image stabilization under high maneuverability. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating the overall architecture and method of collaborative control between the UAV body and gimbal for power transmission inspection in this application embodiment.

[0013] Figure 2 This is a flowchart illustrating the steps for determining the start-up timing of gimbal aiming compensation in an embodiment of this application.

[0014] Figure 3 This is a flowchart illustrating the steps involved in establishing a mapping model from the rate of curvature change to the gimbal angular acceleration in this embodiment of the application.

[0015] Figure 4 This is a flowchart illustrating the steps involved in generating gimbal feedforward control commands in an embodiment of this application.

[0016] Figure 5 This is a flowchart illustrating the steps of injecting gimbal feedforward control commands into the gimbal control loop based on the startup timing in an embodiment of this application. Detailed Implementation

[0017] Example 1

[0018] This describes the overall architecture and methodology of collaborative control between the UAV airframe and gimbal for power transmission line inspection, such as... Figure 1 As shown. Specifically, when a UAV performs high-maneuver flight, such as rapid tower inspection, the inconsistency between the aircraft's response delay and the gimbal's response delay causes the gimbal to be unable to compensate for the aircraft's movement in time, resulting in image jitter or target loss. This embodiment provides a decoupled control strategy based on pre-aiming feedforward, enabling the gimbal to respond in advance before the aircraft's actual movement.

[0019] To implement the above methods, a corresponding UAV hardware control system needs to be constructed. This system typically includes: a flight controller (FAST), responsible for flight path planning and attitude control; a gimbal controller, responsible for driving and stabilizing the gimbal motors; and various sensors, including an inertial measurement unit (IMU) mounted on the airframe, an IMU and encoder mounted on the gimbal, a battery management system (BMS), an airspeed meter or anemometer, and a payload identification interface. These hardware modules are connected via an onboard bus (such as a CAN bus or serial port) to achieve high-speed data exchange and time synchronization.

[0020] Step 101: Obtain the sequence of body maneuver commands to be executed by the UAV, and simultaneously collect the current body and gimbal multi-source status parameters of the UAV.

[0021] Specifically, the sequence of maneuvering commands does not refer to the current real-time state of the UAV, but rather to the control commands generated by the trajectory planning layer of the flight control system that will be executed within a certain period of time in the future. These commands are typically stored in the flight control system's command buffer, containing the target position (x, y, z) and target heading angle ψ for several future control cycles (e.g., within the next 500 milliseconds). cmd Target pitch angle θ cmd And the expected execution time t corresponding to each instruction point cmd By pre-reading these types of instructions, the system gains the ability to predict the future.

[0022] Simultaneously, the system collects current multi-source state parameters in real time. These parameters are fundamental to assessing the current physical state of the UAV. The airframe and gimbal status includes the current attitude angles (yaw, pitch, roll) and angular velocity measured by the IMU. In addition to kinematic parameters, the system also focuses on collecting non-kinematic parameters that affect the dynamic response, such as battery voltage, ambient wind speed and direction, and the mass of the payload. To ensure data validity, the system performs timestamp alignment on data from different sources. For example, using the flight control master clock as a reference, lower-frequency anemometer data and higher-frequency IMU data are interpolated to the same time point to eliminate timing errors caused by asynchronous sampling.

[0023] Step 102: Based on the dynamic estimation of the aircraft response delay using multi-source state parameters, and combined with the command characteristics of the aircraft maneuver command sequence, determine the start timing of the gimbal aiming compensation.

[0024] In this embodiment, the aircraft response delay refers to the time difference between the flight controller issuing a maneuver command and the aircraft actually producing observable motion, such as a change in angular velocity. Traditional control systems typically assume this delay to be a fixed value, but in reality, it is affected by various physical factors. For example, when fully loaded, the aircraft's inertia is large, resulting in a slower response; when the battery is low, the motor torque output weakens, also slowing down the response.

[0025] Accordingly, this embodiment estimates the latency value in real time by establishing a dynamic model. After obtaining an accurate latency estimate, the system calculates the startup timing by combining the characteristics of the instruction to be executed (such as the magnitude of the movement). This timing defines how long before the maneuvering instruction should begin to move. By utilizing the gimbal's fast response speed or by adding a time lead, the gimbal generates a counter-current compensating movement the instant the aircraft begins to shake, visually canceling out the aircraft's disturbance.

[0026] Step 103: Analyze the changes in the trajectory geometry of the aircraft's maneuver command sequence, establish a mapping relationship from trajectory geometry to gimbal dynamic requirements, and generate gimbal feedforward control commands.

[0027] Specifically, traditional gimbal control is passive, relying on sensors to detect movement before initiating a reverse movement, resulting in a lag. In this embodiment, trajectory planning data is analyzed, and changes in trajectory geometric characteristics mainly refer to the degree of curvature of the trajectory and the rate of change, such as curvature and the rate of change of curvature.

[0028] Furthermore, the system utilizes physical mapping relationships to transform the geometric characteristics of the spatial domain into the dynamic requirements of the gimbal motors. For example, when a drone is about to enter a sharp turn, its trajectory curvature changes. Based on this rate of change, the system derives the angular acceleration that the drone will generate and calculates the reverse angular acceleration (Jerk) required by the gimbal to maintain line-of-sight stability. The resulting gimbal feedforward control commands are a set of timestamped, continuous control signals, generated not based on error feedback but on future predictions.

[0029] Step 104: Based on the startup timing, inject the gimbal feedforward control command into the gimbal control loop to drive the gimbal to perform the aiming compensation action.

[0030] In this embodiment, the system combines the calculated timing and waveform. Specifically, the generated gimbal feedforward control command is shifted forward on the time axis by an amount equal to the calculated start-up timing advance. The gimbal motor will then begin executing the pre-calculated compensation action at a specific moment before the actual maneuvering action of the machine.

[0031] This control strategy, which involves issuing a command but not moving the aircraft, prioritizes gimbal movement, thus decoupling aircraft maneuverability from gimbal stability. Even during high-dynamic flight, the gimbal can eliminate image jitter caused by mechanical response lag through advance aiming, ensuring clear and stable inspection footage.

[0032] Example 2

[0033] This embodiment describes a specific implementation method for dynamic estimation of the body response delay. A white-box model that considers the coupling effects of multiple physical fields such as load, voltage, and wind field is constructed to replace the traditional fixed delay parameter or black-box estimation, thereby improving the accuracy of delay estimation under complex operating conditions.

[0034] Step 201: The multi-source state parameters include at least the load mass factor for characterizing the rotational inertia characteristics of the machine body, the dynamic voltage factor for characterizing the motor drive torque response characteristics, and the environmental wind field factor for characterizing the aerodynamic damping characteristics.

[0035] In this embodiment, the physical causes of the aircraft's response delay are analyzed in depth. The delay is not solely determined by the computation time of the control algorithm, but rather depends more on the dynamic characteristics of the physical system. Specifically, the payload mass factor corresponds to the mass m of the mission payload currently carried by the UAV. load The load mass directly changes the moment of inertia I of the machine body about each axis. According to Newton's second law (M=I*α), with the same driving torque M, the larger the moment of inertia I, the smaller the angular acceleration α, and the longer the time (i.e., delay) required to reach the target angular velocity. The power voltage factor corresponds to the current battery voltage U. bat The output torque of a brushless DC motor is proportional to the current, while the current response speed is limited by the terminal voltage. When the battery voltage drops (e.g., in a low-charge state), the motor's dynamic response to sudden commands weakens, leading to increased latency. The environmental wind field factor corresponds to the external wind speed v. wind and wind direction θ wind Aerodynamics can generate additional damping or thrust torque on the aircraft, altering its response characteristics.

[0036] It should be noted that during the multi-source state parameter acquisition process in this embodiment, the system also acquired the current aircraft angular velocity and current flight speed. In the subsequent dynamic estimation model for response delay, these two motion state factors were not included in the calculation of the correction coefficient. This is because power transmission line inspection operations typically employ a low-speed, stable flight mode, with typical flight speeds of 5-15 meters per second, and aircraft angular velocities usually less than 30 degrees per second. Within this operating range, the impact of aerodynamic damping changes caused by flight speed variations and gyroscopic effects caused by angular velocity on the aircraft response delay is relatively small.

[0037] Specifically, comparative tests on multiple UAV models show that, under the aforementioned typical operating conditions, the delay variation caused by the motion state factor is typically less than 2% to 3% of the baseline delay. This magnitude is far smaller than the influence of the payload mass factor, dynamic voltage factor, and environmental wind field factor. Therefore, to simplify engineering implementation and reduce computational complexity, this embodiment ignores the correction effect of the motion state factor in the delay estimation model. In special scenarios such as high-speed flight or high angular velocity maneuvers, the model can be extended as needed to introduce motion state correction coefficients.

[0038] Step 202: Obtain the pre-calibrated and stored baseline delay, which is the inherent response time of the UAV under standard operating conditions.

[0039] Optionally, the system needs a reference baseline. The reference baseline delay τ base It was obtained through offline experimental calibration. The standard operating condition is typically defined as: the payload m of the UAV carrying its nominal mass. nominal For example, 1.0 kg, the battery is at the nominal voltage U nominal For example, 22.2V, and the environment is windless. wind <0.5 m / s. Under this ideal condition, the body response time measured by the step response test is τ. base For example, the τ of a certain type of inspection drone base The calibration value is 20 milliseconds.

[0040] Step 203: Calculate the inertia correction coefficient based on the load mass factor, calculate the voltage correction coefficient based on the dynamic voltage factor, and calculate the wind disturbance correction coefficient based on the environmental wind field factor.

[0041] Specifically, the wind disturbance correction coefficient is calculated based on the environmental wind field factor, including: obtaining the wind direction angle of the environmental wind and the current heading angle of the aircraft, and calculating the angle difference between the two; calculating the crosswind velocity component based on the angle difference, and determining whether the wind direction has a damping effect or a boosting effect on the aircraft's maneuvering direction; and determining the magnitude and positive or negative direction of the wind disturbance correction coefficient according to the magnitude of the crosswind velocity component and the type of damping or boosting effect.

[0042] Accordingly, the system calculates the correction coefficients for each individual factor based on the real-time collected state parameters. For the inertia correction coefficient k... inertia The following linear model is used to calculate: k inertia =1+α m *(m load -m nominal ) / m nominal ; where α m This is the inertia sensitivity coefficient, and its specific value is determined through offline calibration. The formula shows that the heavier the load, the larger the delay correction coefficient.

[0043] For the voltage correction factor kvoltage The following model is used to calculate: k voltage =1+α U *(U nominal -U bat ) / U nominal ; where α U This is the voltage sensitivity coefficient, and its specific value is determined through offline calibration. The formula shows that the lower the voltage (the greater the loss relative to the nominal voltage), the larger the delay correction coefficient.

[0044] For the wind disturbance correction factor k wind The calculation process involves specific geometric logic. It involves obtaining the wind direction angle θ of the ambient wind. wind and the aircraft's current heading angle ψ body Calculate the angle difference between the two. Calculate the crosswind velocity component v based on the angle difference. cross :v cross =v wind *sin(θ wind -ψ body The system determines whether the wind direction has a damping or boosting effect on the aircraft's maneuvering direction. This is based on the target yaw direction Δψ in the maneuvering command. cmd Determine the direction of the crosswind. If the wind direction hinders the aircraft's rotation (e.g., the aircraft turns to the left and the wind blows from the left), it is a damping effect; otherwise, it is a boosting effect. Finally, determine k. wind Size: k wind =1+α w *S dir *(|v cross | / v ref ); where v ref For reference wind speed (e.g., 5 m / s), α w S represents the wind sensitivity coefficient. dir The sign term takes a positive value when it represents a damping effect, increasing the delay; and takes a negative value when it represents a boosting effect, decreasing the delay.

[0045] Step 204: Use the inertia correction factor, voltage correction factor and wind disturbance correction factor to jointly correct the reference delay to obtain the body response delay; the joint correction of the reference delay specifically refers to the calculation according to the multi-factor coupled product model, the body response delay is equal to the product of the reference delay, inertia correction factor, voltage correction factor and wind disturbance correction factor.

[0046] As an optional implementation, a product model is used to couple the above coefficients to obtain the estimated body response delay τ under the current operating conditions. delay τ delay =τ base *k inertia *k voltage *kwind ;

[0047] The product model is used here instead of the addition model because the effect of each physical factor on the delay is usually a scaling relationship rather than a simple summation of absolute values.

[0048] Numerical calculation example: Suppose that in a certain inspection mission, the UAV parameters are as follows: baseline delay τ base =20ms. Nominal load m nominal =1.0 kg, actual mounted load m load =1.3 kg. Inertia sensitivity coefficient α m =0.4. Nominal voltage U nominal =22.2V, current voltage U bat =20.0V. Voltage sensitivity coefficient α U =0.3. Reference wind speed v ref =5m / s, current crosswind v cross =3m / s, and the wind direction is opposite to the direction of maneuvering (damping effect, S) dir =1). Wind sensitivity coefficient α w =0.1.

[0049] The calculation process is as follows: k inertia =1 + 0.4 * (1.3 - 1.0) / 1.0 = 1.12k voltage =1+0.3*(22.2-20.0) / 22.2≈1.03k wind =1+0.1*1*(3 / 5)=1.06τ delay =20*1.12*1.03*1.06≈24.46ms.

[0050] Using this model, the system calculated the actual latency under the current operating conditions to be approximately 24.46 milliseconds, which is about 22% higher than the baseline value, providing a precise basis for subsequent timing scheduling.

[0051] Step 205: Determine the range of values ​​for the sensitivity coefficients among the correction coefficients.

[0052] Specifically, the physical basis of the inertia sensitivity coefficient lies in the principles of rigid body rotational dynamics. According to Newton's and Euler's equations, under the condition of constant driving torque, an increase in rotational inertia will lead to a decrease in angular acceleration, thus prolonging the time required to reach the target angular velocity. Changes in load mass affect the body's rotational inertia through the parallel axis theorem, thereby affecting the response delay. Considering the nonlinear characteristics of the motor drive system and factors such as structural elastic deformation, there is a non-strict linear relationship between the actual rate of change of delay and the rate of change of inertia. The specific value of the inertia sensitivity coefficient is determined through offline calibration experiments.

[0053] The physical basis of the voltage sensitivity coefficient lies in the characteristics of brushless DC motors. When the battery voltage drops, the initial torque of the motor decreases, leading to an increase in response delay. However, the current loop controller in the motor drive system can partially compensate for the effect of insufficient voltage by increasing the pulse width modulation duty cycle. The specific value of the voltage sensitivity coefficient is determined through offline calibration tests.

[0054] The physical basis of the wind sensitivity coefficient lies in aerodynamic moment analysis. The aerodynamic yaw moment generated by crosswinds acting on the aircraft is related to the crosswind dynamic pressure and the lateral projected area of ​​the aircraft. This aerodynamic moment is relatively small compared to the motor yaw control moment, and its impact on response delay is limited. The specific value of the wind sensitivity coefficient is determined through offline calibration tests based on the differences in the aerodynamic shape of the aircraft model.

[0055] The above typical value range serves as a reference interval for offline calibration tests. If the parameter values ​​obtained from calibration exceed the typical range, the test conditions and data quality need to be checked. The final parameter values ​​are based on the results of the offline calibration tests.

[0056] Example 3

[0057] Describe the specific implementation details regarding the startup timing of determining gimbal aiming compensation, such as... Figure 2 As shown.

[0058] Step 301: Extract the target angle increment to be executed from the sequence of machine maneuver commands, and calculate the motion amplitude characteristics accordingly.

[0059] Specifically, the system parses the target angle change for the current maneuver from the pre-read command sequence. This typically includes the target yaw increment Δψ. cmd The target increment Δθ of the pitch axis cmd Computer motion amplitude characteristics A cmd This feature is used to calculate the total intensity of the maneuver. In one implementation, it can be represented using the Euclidean norm: A cmd =sqrt(Δψ cmd 2 +Δθ cmd 2 The larger the value of this characteristic, the more intense the action that the body is about to perform.

[0060] Step 302: Calculate the additional lead based on the maneuver amplitude characteristics. The additional lead is positively correlated with the maneuver amplitude characteristics to compensate for the gimbal's response lag under large maneuvers.

[0061] Accordingly, to address the challenges posed by large maneuvers, the system introduces an additional lead time τ. amplitudeThis lead time is not to compensate for physical communication delays, but rather to compensate for the dynamic lag of the gimbal servo system under large deviations. Physical laws show that when the gimbal motor tracks a large-angle step signal, the setting time increases with the amplitude. Therefore, the larger the maneuver amplitude, the earlier the aiming action needs to begin. The formula for calculating the additional lead time can be expressed as: τ amplitude =β*A cmd / ω gimbal_max ; where ω gimbal_max β represents the maximum angular velocity capability of the gimbal motor, and β is the amplitude sensitivity coefficient, a dimensionless constant, which is taken as 0.8 in this embodiment. This formula reflects τ amplitude With A cmd A positive correlation.

[0062] Step 303: The machine response delay and the additional advance amount are superimposed to obtain the specific start-up timing advance amount.

[0063] Furthermore, the start-up timing advance τ advance It is a combined result of physical latency and amplitude compensation. This takes into account the inherent slight latency τ of the gimbal itself from receiving commands to taking action. gimbal And a safety margin τ needs to be reserved to prevent reverse jitter caused by premature action. margin The overall calculation formula is: τ advance =τ delay +τ amplitude +τ gimbal -τ margin Using this formula, the system calculates the precise length of the time window.

[0064] Step 304: Establish a triggering determination mechanism based on the start-up timing advance: monitor the expected execution time of each instruction in the aircraft's maneuvering instruction sequence in real time, and calculate the time difference between the expected execution time and the current time; when the time difference is less than or equal to the start-up timing advance, generate a pre-aiming start-up trigger signal to trigger the injection of gimbal feedforward control instructions.

[0065] Preferably, the system enters a real-time countdown monitoring state. For each frame of maneuver command in the queue, the system reads its timestamp (expected execution time t). execute ), and the current system time t current Compare and calculate the time difference Δt: Δt = t execute -t current The system logic continuously checks the condition: Δt≤τ advanceOnce this condition is met, it indicates that the time before the aircraft's movement has been shortened to within the preset lead time range. At this point, the system immediately generates a pre-aiming trigger signal. This signal acts like a starting pistol, instructing subsequent control modules to begin injecting the calculated feedforward waveform into the gimbal controller, achieving a pre-aiming effect where the gimbal moves before the aircraft itself.

[0066] Example 4

[0067] This paper describes the specific implementation process of analyzing the changes in the geometric characteristics of the flight path and generating gimbal feedforward control commands in the overall method. Specifically, traditional gimbal control is based on attitude error feedback adjustment. This lag adjustment is prone to causing transient vibration or overshoot of the gimbal when the UAV enters sections of road with sudden curvature changes such as curves.

[0068] Step 401, analyze the changes in the geometric characteristics of the aircraft maneuver command sequence, including: calculating the instantaneous curvature of the track based on the position information in the aircraft maneuver command sequence; calculating the derivative of the instantaneous curvature of the track with respect to time to obtain the rate of change of curvature, which reflects the speed of change of the curvature degree of the track.

[0069] In this embodiment, the system reconstructs the geometric properties of the track using discrete waypoints. The instantaneous curvature κ of the track is a physical quantity describing the degree of curvature of the curve. For consecutive position points P in the pre-fetch command queue... i-1 ,P i ,P i+1 It can be calculated using the method of adjacent three points or differential geometry formulas. The specific calculation formula can be expressed as: κ=|v vec ×a vec | / |v vec | 3 ; where v vec a represents the instantaneous velocity vector of the organism. vec Let represent the instantaneous acceleration vector of the aircraft, × represent the cross product of the vectors, and |...| represent the magnitude of the vectors. Using this formula, the system can calculate the curvature value at each point on the trajectory.

[0070] Furthermore, the system calculates the rate of change of curvature, dκ / dt. This physical quantity reflects how quickly a trajectory transitions from a straight line into a curve, or from a curve back to a straight line. Mathematically, it corresponds to the first derivative of curvature with respect to time. The system uses the central difference method to process the curvature sequence: dκ i / dt=(κ i +1-κ i -1) / (t i +1-t i -1), this indicator can detect impending violent maneuvers. For example, when a drone suddenly cuts from straight flight into a small-radius hover, the curvature will step, and the rate of change of curvature will show a large peak.

[0071] Optionally, the system identifies abrupt changes in track curvature and extracts abrupt change characteristic parameters. Specifically, after obtaining the curvature change rate sequence, the system further identifies abrupt curvature changes. Abrupt curvature changes refer to track locations where the curvature changes drastically, corresponding to the transition region where the UAV moves from a straight line into a curve or returns to a straight line from a curve.

[0072] Optionally, an adaptive curvature change rate threshold is established. Considering the varying sensitivity to curvature changes at different flight speeds, an adaptive threshold calculation method related to flight speed is adopted. The formula for calculating the curvature change rate threshold is:

[0073] Th κ =κ ref / T transition ;

[0074] Among them, Th κ κ is the threshold for the rate of change of curvature, expressed in meters per second; ref For reference curvature, the unit is per meter; T transition This is the desired curvature transition time, expressed in seconds. The reference curvature is taken as the curvature value of a circular arc with a radius of ten meters, i.e., 0.1 per meter. The desired transition time is set according to the gimbal's response capability and the stability requirements of the inspected images, typically ranging from 0.3 to 0.5 seconds. This threshold is only identified as abrupt changes requiring feedforward compensation when the curvature change is large, thus avoiding over-response to normal, slow turns.

[0075] Optionally, abrupt change point determination is performed. The data points in the curvature change rate sequence are traversed, and when the absolute value of the curvature change rate at a point exceeds a threshold, that point is determined to be a curvature abrupt change point. The determination condition is expressed as follows:

[0076] |dκ i / dt|>Th κ ;

[0077] Among them, dκ i di represents the curvature change at point i, in meters; dt represents the sampling time interval, in seconds; dk represents the curvature change at point i. i / dt represents the rate of change of curvature at point i; the vertical bar indicates the absolute value operation; Th κ Let be the threshold for the rate of change of curvature. When the above inequality holds, the i-th point is determined to be a point of sudden change in curvature.

[0078] Optionally, curvature abrupt change feature parameters are extracted. For each identified curvature abrupt change point, the following feature parameters are recorded: the time t of the abrupt change point. transition , indicating the expected time of the mutation; the pre-mutation curvature κ before , representing the steady-state curvature value before entering the transition region; curvature κ after the abrupt change. after, representing the steady-state curvature value after leaving the transition region; the peak rate of curvature change represents the degree of most drastic curvature change during the transition. The above set of characteristic parameters is used in subsequent steps to generate the feedforward curve.

[0079] Step 402: Establish the mapping relationship from trajectory geometry to gimbal dynamics requirements. Specifically, this refers to establishing a mapping model from the rate of curvature change to the gimbal angular jerk, such as... Figure 3 As shown: Based on the product relationship between the rate of curvature change and the UAV flight speed obtained from the body maneuver command sequence, the body angular acceleration requirement when performing curve maneuvers is derived; based on the body angular acceleration requirement, the gimbal angular acceleration feedforward amount used to offset the influence of body maneuvers is calculated, and the gimbal angular acceleration feedforward amount is related to the time second derivative of the instantaneous curvature of the trajectory.

[0080] In this embodiment, a cross-layer mapping from the spatial geometry domain to the motor dynamics domain is achieved. According to kinematic principles, the yaw rate ω of the aircraft flying along a curve... psi With curvature κ and flight speed v flight There is a constraint relationship: ω psi =κ*v flight Differentiating this equation, we get dω / dt = dκ / dt·v + κ·dv / dt; simplifying, we obtain the required angular acceleration α of the body. body : α body ≈(dκ / dt)*v flight This formula shows that the greater the rate of change of curvature, or the faster the flight speed, the greater the angular acceleration generated by the aircraft.

[0081] Specifically, to maintain the stability of the gimbal's line of sight relative to the ground, the gimbal generates a compensating motion of equal magnitude and opposite direction to the aircraft's motion. Therefore, the gimbal's angular acceleration α... gimbal It should be equal to -α body Further differentiating with respect to time, J = -( (d²κ) / (dt²v) + (dκ / dt) * (dv / dt) ); after processing, the angular accelerometer feedforward J of the gimbal motor is obtained. gimbal (i.e., Jerk): J gimbal =d(α gimbal ) / dt=-(d 2 κ / dt 2 )*v flight ; where d 2 κ / dt 2 It is the second derivative of curvature with respect to time. This physical quantity directly corresponds to the rate of change of motor torque. Through this mapping model, the system accurately calculates how rapidly the gimbal motor torque should change in order to counteract cornering disturbances.

[0082] As an example of numerical calculation, assume a drone enters a circular turn with a radius of 20 meters at a speed of 10 meters per second, with a transition time of 0.4 seconds. The curvature before entering the turn is 0, and the curvature after entering the turn is 0.05 (1 / 20). Then, the peak value of the rate of change of curvature is approximately: dκ / dt≈(0.05-0) / 0.4=0.125; the required angular acceleration is: α body ≈0.125*10=1.25rad / s 2 The required compensation angular acceleration from the gimbal is -1.25 rad / s². 2 The corresponding peak value of the Jerk feedforward is further calculated based on the second derivative.

[0083] Step 403: Generate gimbal feedforward control commands, such as... Figure 4 As shown, it includes: using a Sigmoid curve smoothing filter to continuously process the gimbal angular acceleration feedforward quantity to filter out high-frequency abrupt changes; and using the smoothed gimbal angular acceleration feedforward quantity as a gimbal feedforward control command for the gimbal control loop to execute.

[0084] In this embodiment, the cutoff frequency of the Sigmoid curve smoothing filter is set based on: obtaining the mechanical resonant frequency of the gimbal, which is obtained through factory frequency sweep calibration or online identification of the pulse response during flight; setting the cutoff frequency to a safe value lower than the mechanical resonant frequency to suppress the excitation of the gimbal's mechanical resonance by the feedforward control command.

[0085] Correspondingly, the calculated Jerk signal may contain numerical noise or step abrupt changes, which can easily excite vibrations in the mechanical structure when applied to the motor. Therefore, a Sigmoid curve smoothing filter is introduced into the system. This filter uses the derivative form of the Sigmoid function as the transition curve, and the specific smoothing function σ(t) can be expressed as: σ(t) = 1 / (1+exp(-k smooth *(tt center ) / T transition )); where k smooth The steepness of the control curve, t center This is the transition center moment.

[0086] Furthermore, the cutoff frequency f of the filter cutoff The settings are based on clear physical principles. The system acquires the mechanical resonant frequency f of the gimbal. resonance (For example, if measured offline by frequency sweep, it is 40 Hz), then the cutoff frequency is set to a safe value below this resonant frequency (for example, set to 0.3 times the resonant frequency, i.e., 12 Hz). cutoff =k safety *f resonanceThrough this design, the generated gimbal feedforward control commands contain both the low-frequency energy required to counteract the movement of the machine body and filter out the high-frequency components that may cause the gimbal to self-excite oscillation, thus achieving smooth and powerful control.

[0087] Example 5

[0088] This document describes the specific implementation of the injection gimbal control loop and collaborative control. This embodiment focuses on addressing the coexistence problem between the feedforward signal and the original feedback control system, as well as system safety issues under extreme conditions.

[0089] Step 501: Inject the gimbal feedforward control command into the gimbal control loop based on the startup timing, such as... Figure 5 As shown, this includes: shifting the time axis of the gimbal feedforward control command so that the effective range of the gimbal feedforward control command starts the timing sequence earlier than the execution time of the aircraft maneuver command sequence; if there are multiple time-overlapping aircraft maneuver commands, the amplitude of the shifted gimbal feedforward control commands corresponding to each command is superimposed to generate a comprehensive gimbal feedforward control command.

[0090] In this embodiment, the system performs timing alignment. Assume a startup timing advance of τ. advance The timeframe is 50 milliseconds, and the execution time of the airframe maneuver command is T0. The system shifts the entire time axis of the Jerk feedforward curve forward by 50 milliseconds, making its effective range start from T0-50ms. Furthermore, in actual flight, the UAV may simultaneously perform yaw and pitch maneuvers, or execute two closely spaced maneuvers consecutively. In this case, the system employs an amplitude superposition strategy. For example, for the yaw axis, the system superimposes the feedforward component J from the heading maneuver command... psi1 With the feedforward component J from the next maneuver command psi2 Perform algebraic summation at the corresponding time points: J total (t)=J psi1 (t)+J psi2 (t); Through this superposition, the system can generate integrated feedforward commands that adapt to complex combined maneuvers.

[0091] Step 502, the injection of the gimbal control loop also includes a weighted fusion process of feedforward and feedback: obtaining the feedback control quantity output in real time by the gimbal closed-loop controller; using dynamic weighting coefficients to weight and sum the integrated gimbal feedforward control command and feedback control quantity to form the final gimbal collaborative control output that drives the gimbal motor; wherein, within the preview window triggered by the startup timing, the dynamic weighting coefficients are adjusted to increase the proportion of feedforward control.

[0092] As an optional implementation, the system employs a dynamic weighted fusion algorithm to synthesize the final motor control quantity u. gimbal The specific fusion formula is: ugimbal =w ff (t)*u feedforward +(1-w ff (t))*u feedback ; where u feedforward The torque or speed command is obtained by integrating the feedforward Jerk command. feedback This is the feedback command calculated by the original PID controller or robust controller based on sensor error. Dynamic weighting coefficient w ff (t) is a scalar that varies with time. Within a specific time window after the system detects the pre-aiming trigger signal (i.e., the period between when the maneuver is about to occur and when it is actually occurring), the system will... ff When adjusted to a larger value (e.g., 0.8), control is handed over to the feedforward channel for rapid response. During stable flight, w... ff Gradually decay to a smaller value (e.g., 0.2 or 0), at which point the control authority returns to the feedback channel, using integral action to eliminate steady-state error.

[0093] Step 503: Establish an execution monitoring and closed-loop verification mechanism: monitor the actual attitude change of the gimbal in real time and calculate the tracking error between it and the expected attitude change corresponding to the gimbal feedforward control command; when the tracking error exceeds the preset safety threshold, determine that the feedforward compensation has failed, forcibly adjust the dynamic weight coefficient to zero, switch the system to pure feedback control mode, and record the abnormal state.

[0094] Specifically, to ensure the safety of engineering applications, this embodiment introduces a failure protection mechanism. Feedforward control has a fast response, but based on open-loop model prediction, the feedforward signal may be misleading when model parameters (such as load mass) are estimated incorrectly. The system calculates the tracking error in real time. track Error track =|Angle actual -Angle expected |;When an Error is detected track If the temperature continues to exceed a preset safety threshold (e.g., 5 degrees), the system immediately determines that the feedforward has failed. At this point, the system triggers circuit breaker logic, forcibly reducing the dynamic weighting coefficient w. ff Setting it to 0 indicates that the system instantly cuts off the feedforward path and reverts to the traditional pure feedback control mode. This type of mechanism ensures that the gimbal will not malfunction or move erratically under any algorithmic anomaly, thus guaranteeing the safety of the equipment.

[0095] Example 6

[0096] The source of the key parameters (load mass, delay correction coefficient, resonant frequency) in the above embodiments is described. The acquisition of these parameters is the basis for the effective operation of the control method.

[0097] Step 601, the payload mass factor can be obtained in any of the following ways or a combination thereof: query the pre-built payload configuration database and match the corresponding mass parameters according to the identified payload type code; or, establish the UAV lift balance equation and estimate the current payload mass online based on the difference between the total thrust of the motor and the unloaded weight of the airframe in hovering or level flight state.

[0098] Optionally, the system provides two payload mass acquisition modes for different operational scenarios. The first is a pre-configured mode, suitable for standardized operations. The system pre-builds a database storing mass data for various camera models, such as a specific camera model (e.g., a full-frame camera weighing 0.66 kg). When the drone powers on, it obtains the type code by reading the electronic tag of the payload device (such as RFID, Radio Frequency Identification, or USB handshake information) and directly looks up the precise mass in the table. The second is an online estimation mode, as an alternative, suitable for carrying non-standard or unknown payloads. The system uses the lift balance principle for estimation. In hovering mode, the total motor thrust F... thrust The total gravitational force that needs to be balanced is: F thrust =(m body +m load )*g; where m body Let F be the known empty weight of the machine, and g be the acceleration due to gravity. thrust The load mass can be estimated in real time using the motor current. The formula for estimating the load mass can be obtained from the deformation: m load =(F thrust / g)-m body .

[0099] When both the pre-configuration mode and the online estimation mode are available, the system adopts a fusion verification strategy to determine the final load quality value used.

[0100] Specifically, the quality value obtained from the pre-stored configuration mode is used as the baseline value m. preset The quality value obtained from the online estimation model is used as the verification value m. online Calculate the absolute value of the deviation between the two:

[0101] Δm=|m online -m preset |;

[0102] Where, Δ m The load mass deviation obtained for the two modes is expressed in kilograms; m online The load mass is estimated online, in kilograms; m preset The pre-configured load mass is in kilograms; the vertical line symbol indicates absolute value calculation.

[0103] The deviation is compared with a preset threshold. The threshold is calculated using the following formula:

[0104] Th m =k tol ×m preset ;

[0105] Among them, Th m This is the mass deviation threshold, in kilograms; k tol is the tolerance coefficient, dimensionless, with a value of 0.1; m preset For the pre-configured load mass. If Δ m Less than Th m If the load configuration is determined to be correct, the system uses the pre-stored configuration value as the final load mass because the pre-stored value usually has higher accuracy. If Δ m Greater than or equal to Th m The system prompts the operator through the ground station interface to confirm whether the payload configuration is correct, and decides whether to use the pre-stored configuration value or the online estimated value according to the operator's instructions.

[0106] This fusion strategy leverages the high precision of pre-stored configurations while simultaneously performing cross-validation through online estimation, enabling timely detection of load configuration errors or abnormal load changes.

[0107] Step 602: The reference delay, inertia correction factor, voltage correction factor, and wind disturbance correction factor are obtained through pre-executed offline calibration steps: Under standard load, full voltage, and windless conditions, the response time of the machine to a step command is measured to obtain the reference delay; under single variable conditions of changing load mass, changing power supply voltage, and changing crosswind speed, the rate of change of response time relative to the reference delay is measured, and the corresponding inertia correction factor, voltage correction factor, and wind disturbance correction factor are obtained through fitting calculation.

[0108] In this embodiment, the baseline delay and various sensitivity coefficients are obtained through a pre-executed offline calibration step, specifically including:

[0109] An airframe response delay calibration test platform was constructed. This platform comprises the following components: a ground-mounted UAV test bracket that allows the airframe to rotate freely around the yaw axis but restricts movement along the pitch and roll axes; a high-precision angular velocity measuring device, employing a fiber optic gyroscope or laser gyroscope with a sampling rate of at least 1 kHz as the reference sensor, mounted at the airframe's center of gravity; a flight control command recording device for recording the timing of yaw maneuver commands with microsecond-level accuracy; an adjustable counterweight device for simulating different load mass conditions; an adjustable DC power supply for simulating different battery voltage conditions; and an industrial fan array for generating a simulated wind field with controllable direction and speed.

[0110] Perform the baseline delay calibration procedure. The standard operating condition is defined as: load mass equal to nominal load mass, battery voltage equal to nominal voltage, and ambient wind speed less than 0.5 meters per second. Specific steps include:

[0111] The first step is to fix the drone to the test bracket, install the nominal load, connect the nominal voltage power supply, and ensure that the ambient wind speed meets the requirements.

[0112] The second step is to send a step yaw command to the flight controller, with the command amplitude set to 30 degrees, and simultaneously record the time t when the command is issued. cmd ;

[0113] The third step is to continuously collect the yaw rate response curve of the aircraft using a reference angular velocity sensor;

[0114] The fourth step is to set the angular velocity response detection threshold to 5% of the steady-state angular velocity peak value, and then identify the moment t when the angular velocity response curve first exceeds the detection threshold. response ;

[0115] Fifth, calculate the response delay for a single trial using the following formula:

[0116] τ single =t response -t cmd ;

[0117] Where, τ single The response delay measured in a single test is expressed in seconds; t response The moment when the body's angular velocity first exceeds the detection threshold, measured in seconds; t cmd The time when the yaw command was issued is in seconds.

[0118] Step 6: Repeat the above steps at least fifty times to calculate the mean and standard deviation of the response delay samples. Take the mean as the baseline delay. If the standard deviation exceeds 10% of the mean, recalibration is required.

[0119] Optionally, perform the inertia sensitivity coefficient calibration procedure. Maintain voltage and wind speed conditions as standard operating conditions, and conduct multiple calibration tests by varying only the load mass. Set the load mass test sequence to include at least five test points, covering a range from 0.5 times the nominal load to 1.5 times the nominal load. Perform no fewer than thirty step response tests for each load mass, obtaining the average response delay under that load condition. Calculate the normalized load change and normalized delay change for each test point. Perform a least-squares linear fit with the normalized load change as the independent variable and the normalized delay change as the dependent variable; the slope obtained is the inertia sensitivity coefficient. Calculate the goodness of fit; if the goodness of fit is below 0.9, the test data quality needs to be checked.

[0120] Optionally, a voltage sensitivity coefficient calibration procedure can be performed. Maintaining standard load mass and wind speed conditions, multiple calibration tests are conducted by varying only the supply voltage. The voltage test sequence should include at least five test points, covering a range from 0.7 times to 1.0 times the nominal voltage. The calibration method is similar to the inertia sensitivity coefficient calibration procedure, obtaining the voltage sensitivity coefficient through least-squares linear fitting.

[0121] Optionally, a wind sensitivity coefficient calibration procedure is performed. Maintaining standard load mass and voltage conditions, calibration tests are conducted using simulated wind fields of different directions and speeds generated by an industrial fan array. A crosswind speed test sequence is set, containing at least five test points covering the range from zero to the reference wind speed, with the wind direction set perpendicular to the aircraft's initial heading to form a pure crosswind condition. For each crosswind speed, both tailwind and headwind turning scenarios are tested, and the delay change caused by wind disturbance is calculated. The wind sensitivity coefficient is then obtained through fitting.

[0122] Optionally, the calibration results are verified and stored. Ten or more different combinations of load, voltage, and wind speed are randomly selected, and the actual response delay is measured and compared with the model's predicted value. If the root mean square value of the prediction error is less than 15% of the actual delay, the calibration result is considered valid. The verified parameter values ​​are stored in the parameter configuration area of ​​the flight control non-volatile memory.

[0123] This embodiment describes the source of each coefficient and requires a standardized set of offline experiments. The baseline delay τ is measured. base Under standard operating conditions (e.g., 1kg payload, fully charged, no wind), a 30-degree step yaw command was sent to the drone, and the time difference from the command issuance to the drone's angular velocity reaching 5% of its peak value was recorded using a high-frequency motion capture system. Next, the inertia sensitivity coefficient α was measured. m Keeping all other conditions unchanged, only increasing the load to 1.5 kg, the new delay τ was measured. new Calculate the rate of change of delay (τ) new -τ base ) / τ base The ratio of the load change rate (1.5-1.0) / 1.0 is α. m Similarly, the voltage sensitivity coefficient α can be calibrated by changing the voltage and wind speed. U Wind sensitivity coefficient α w .

[0124] In some alternative implementations, a similar dual-mode strategy is also employed to obtain the mechanical resonant frequency of the gimbal. Method one is factory sweep calibration. On the gimbal's factory test bench, a sinusoidal sweep signal with a frequency gradually varying from 5Hz to 50Hz is input, and the peak vibration response is measured using an accelerometer to lock the resonant frequency. Method two is online pulse identification. During flight, using the free decay oscillation data of the gimbal after being disturbed, the current resonant frequency is estimated in real time through zero-crossing detection or Fast Fourier Transform (FFT). This frequency value is directly used to update the filter's cutoff frequency to adapt to characteristic drift caused by mechanical aging.

[0125] Example 7

[0126] The problem addressed in this embodiment is how to deploy the control algorithm logic from the above embodiments into actual physical devices, enabling them to execute methods.

[0127] This embodiment provides an onboard computing device for unmanned aerial vehicles (UAVs). The device physically includes at least one processor, a memory, and a communication interface. These components communicate via a bus system. The processor can be a general-purpose central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or a field-programmable gate array (FPGA). In practical applications, a high-performance embedded computing platform, such as the NVIDIA Jetson series or a high-performance flight control computer based on the ARM architecture, is preferred to meet the computing power requirements for real-time calculation of trajectory curvature and multiphysics models. The memory stores program instructions and temporary data generated during operation (such as synchronization status datasets and pre-fetch command queues). The communication interface establishes connections with external devices such as the flight control system, gimbal controller, BMS, and sensors to receive data (such as maneuver commands and sensor status) and send commands (such as gimbal feedforward control commands).

[0128] This embodiment also provides a computer-readable storage medium (such as flash memory, SD card, hard disk) on which a computer program is stored. When the computer program is executed by a processor, it can realize a decoupled and coordinated control method for gimbal aiming and feedforward and body maneuvering for power transmission inspection.

[0129] This invention calculates the specific impact of changes in moment of inertia, motor torque attenuation, and aerodynamic damping / boost effects on the aircraft's response speed by real-time acquisition of load mass, battery voltage, and environmental wind field parameters, and dynamically calculates the start-up sequence in conjunction with maneuver amplitude characteristics.

[0130] Furthermore, the aiming actions are made ahead of or behind due to changes in operating conditions (such as low battery, heavy load, crosswind), ensuring that the gimbal compensation actions and the actual maneuvers of the aircraft are synchronized on the time axis. For the timing deviation problem caused by the dynamic nonlinearity of the aircraft response delay, a delay estimation model based on multi-physics coupling is adopted, rather than a fixed delay assumption.

[0131] Correspondingly, by calculating the track curvature and its derivative with respect to time, such as the rate of change of curvature, the system can detect abrupt changes in the degree of track curvature in advance and directly map these geometric features into Jerk feedforward commands for the gimbal motors' angular jerk. Combined with S-shaped smoothing filtering, this allows the gimbal to generate a counter-torque that conforms to dynamic laws at the moment the aircraft transitions from a straight path to a curve, solving the mechanical vibration and tracking overshoot problems caused by passively responding to abrupt curvature changes in existing technologies.

[0132] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A gimbal pre-look feedforward and airframe maneuver decoupled cooperative control method for power transmission inspection, characterized in that, include: Acquire the sequence of maneuver commands to be executed by the UAV, and simultaneously collect the current multi-source status parameters of the UAV's body and gimbal; Based on the dynamic estimation of the aircraft response delay using multi-source state parameters, and combined with the command characteristics of the aircraft maneuver command sequence, the activation timing of the gimbal aiming compensation is determined. The changes in the trajectory geometry of the aircraft's maneuver command sequence are analyzed, a mapping relationship between trajectory geometry and gimbal dynamic requirements is established, and gimbal feedforward control commands are generated. Based on the startup timing, the gimbal feedforward control command is injected into the gimbal control loop to drive the gimbal to perform aiming compensation action. Determine the activation sequence of gimbal aiming compensation, including: Extract the target angle increment to be executed from the sequence of machine maneuver commands, and calculate the motion amplitude characteristics accordingly; The additional lead is calculated based on the maneuver amplitude characteristics. The additional lead is positively correlated with the maneuver amplitude characteristics and compensates for the gimbal's response lag under large maneuvers. The specific startup timing advance is obtained by adding the machine response delay and the additional advance. The method also includes establishing a trigger determination mechanism based on startup timing advance: Real-time monitoring of the expected execution time of each instruction in the aircraft's maneuvering instruction sequence, and calculation of the time difference between the expected execution time and the current time; When the time difference is less than or equal to the start-up timing advance, a pre-aiming start-up trigger signal is generated, triggering the injection of the PTZ feedforward control command; Analyze the changes in the trajectory geometry of the aircraft's maneuver command sequence, including: Calculate the instantaneous curvature of the trajectory based on the position information in the sequence of aircraft maneuver commands; Calculate the derivative of the instantaneous curvature of the track with respect to time to obtain the rate of change of curvature, which reflects how fast the curvature of the track changes. Establishing a mapping relationship from flight path geometry to gimbal dynamics requirements, specifically referring to establishing a mapping model from the rate of curvature change to the gimbal angular jerk: Based on the product relationship between the rate of curvature change and the UAV flight speed obtained from the airframe maneuver command sequence, the airframe angular acceleration requirement when performing cornering maneuvers is derived. Based on the requirements of the airframe angular acceleration, the gimbal angular acceleration feedforward is calculated to counteract the effects of airframe maneuvering. The gimbal angular acceleration feedforward is related to the time second derivative of the instantaneous curvature of the trajectory. Generate gimbal feedforward control commands, including: The Sigmoid curve smoothing filter is used to continuously process the feedforward of the gimbal angular acceleration to filter out high-frequency abrupt components. The smoothed angular acceleration feedforward is used as the gimbal feedforward control command for the gimbal control loop to execute.

2. The method of claim 1, wherein, The multi-source state parameters include at least the load mass factor for characterizing the rotational inertia characteristics of the machine body, the dynamic voltage factor for characterizing the motor driving torque response characteristics, and the environmental wind field factor for characterizing the aerodynamic damping characteristics.

3. The method of claim 2, wherein, Dynamic estimation of organism response delay based on multi-source state parameters includes: Obtain the pre-calibrated and stored baseline delay, which is the inherent response time of the UAV under standard operating conditions; The inertia correction factor is calculated based on the load mass factor, the voltage correction factor is calculated based on the dynamic voltage factor, and the wind disturbance correction factor is calculated based on the environmental wind field factor. The reference delay is jointly corrected by inertia correction factor, voltage correction factor and wind disturbance correction factor to obtain the body response delay.

4. The method according to claim 3, characterized in that, The joint correction of the reference delay specifically refers to the calculation according to the multi-factor coupled product model, where the body response delay is equal to the product of the reference delay, the inertia correction coefficient, the voltage correction coefficient, and the wind disturbance correction coefficient.

5. The method according to claim 1, characterized in that, Based on the startup timing, the gimbal feedforward control command is injected into the gimbal control loop, including: The time axis of the gimbal feedforward control command is shifted so that the effective range of the gimbal feedforward control command is started earlier than the execution time of the aircraft maneuver command sequence. If there are multiple overlapping machine maneuver commands, the amplitude of the translational gimbal feedforward control commands corresponding to each command is superimposed to generate a comprehensive gimbal feedforward control command.

Images