Unmanned aerial vehicle offshore landing time delay compensation method based on segmented time delay identification and padé approximation

Abstract

The application discloses a kind of unmanned aerial vehicle offshore landing time delay compensation method based on segmented time delay identification and Padé approximation, comprising: the control channel of unmanned aerial vehicle is segmented time delay identification, to determine the total time delay and smooth time delay estimation;Variable order Padé compensator is constructed;The Padé order of Padé compensator is automatically selected based on control loop operating bandwidth, smooth time delay estimation;Performing sea state adaptive rule, performing stage adaptive rule, and ensuring the continuous output of Padé compensator by state mapping at switching moment;The expected pose of deck is predicted, so that the time window of deck forward prediction is equal to the time delay window of Padé compensator;The prediction result is coupled to the reference input end of Padé compensator, and the control signal after feedforward coupling is executed by unmanned aerial vehicle actuator, to realize the dynamic landing of unmanned aerial vehicle for dynamic platform.The application significantly improves the landing accuracy and reliability of offshore dynamic platform landing task under the condition of long time delay, strong sea state and gust disturbance.

Description

Technical Field

[0001] This invention belongs to the field of autonomous landing control technology for unmanned aerial vehicles (UAVs), specifically relating to a method for compensating for UAV landing time delay at sea based on piecewise time delay identification and Padé approximation. Background Technology

[0002] As collaborative operations of unmanned systems extend to maritime scenarios, cross-domain collaborative missions between unmanned aerial vehicles (UAVs) and unmanned vessels place higher demands on the dynamic landing capabilities of multi-rotor UAVs on moving platforms (such as unmanned surface vessels, ship stern decks, and floating platforms at sea). Unlike landing on a land-based static platform, landing on a sea-based dynamic platform faces two types of coupled challenges: First, the target platform itself is in six-degree-of-freedom random motion. The forced excitation of the waves causes deck heave, roll, pitch, and yaw of the hull, making the landing reference target constantly moving and exhibiting significant frequency band characteristics determined by the wave spectrum. Second, there is a non-negligible input time delay in the UAV's execution channel. From the time the command is issued by the ground station to the time the blade generates effective thrust, it needs to go through multiple serial links, including wireless communication links, flight control firmware processing, ESC commutation, brushless motor electromechanical response, and blade aerodynamic setup. The accumulated time delay of each link is usually on the order of tens to hundreds of milliseconds. At the terminal landing phase, this is enough to cause significant phase lag in the position control loop or even closed-loop instability. Under conditions close to the sea surface, the blade aerodynamic setup time is further increased by the water surface effect.

[0003] Existing technologies for input time delay compensation for shipborne or general unmanned aerial vehicles (UAVs) mainly fall into two categories. One category is the Padé approximation-based time delay compensation method, which uses rational transfer function approximation to equivalently transform a time-delayed control system into a time-delay-free system for design. However, in this type of technology, the total time delay is directly substituted into the design as a known constant, without disclosing how to decompose and identify the time delays of each physical source segment, resulting in a lack of portability under different hardware / firmware combinations; the compensator order and bandwidth are fixed throughout the entire process, neither changing with sea state nor with mission phase, leading to insufficient compensation margin under conditions of long time delays, strong sea states, or high accuracy requirements at the end of landing; the controlled objects are mostly fixed-wing shipborne aircraft or single attitude loops, without addressing the comprehensive processing of multi-level time delays of the position loop, attitude loop, and propeller aerodynamics of the multi-rotor UAV's moving platform; and there is a lack of structural coupling between the compensation and prediction of the moving platform's future attitude, resulting in an inconsistency between the compensation time window and the prediction time window, leading to a systematic deviation between the deck attitude aimed at by the compensated command when it reaches the actuator and the actual attitude. Another type is the prediction feedback method based on Smith predictors or state predictors. This type of method is more sensitive to model mismatch and time delay mismatch, and existing public technologies also do not explicitly incorporate sea state spectrum information and mission phase information into the online parameter tuning of the predictor / compensator.

[0004] Furthermore, existing literature on longitudinal deck motion compensation methods for carrier-based aircraft compensates for deck motion rather than input time delay, while patents on autonomous landing flight control systems for carrier-based unmanned aerial vehicles remain at the system / device level and do not disclose specific time delay compensation algorithms.

[0005] Therefore, in the dynamic landing scenario of multi-rotor UAVs on a maritime platform, how to achieve engineerable segmented time delay identification, how to construct a Padé compensator that can adapt to sea conditions and mission stages, how to form a structural coupling with the future pose prediction of the moving deck with a time window consistent with the time delay, and how to maintain closed-loop stability even when the time delay identification value and the actual time delay are mismatched have become key technical problems that urgently need to be solved. Summary of the Invention

[0006] The purpose of this invention is to provide a method for compensating for the time delay of UAVs at sea based on piecewise time delay identification and Padé approximation, so as to improve the landing accuracy and reliability of dynamic landing missions at sea under conditions of long time delay, strong sea state, and gust disturbance, without relying on known nominal time delay.

[0007] To achieve the above objectives, the present invention employs the following technical solution: A method for compensating for UAV landing time delay at sea based on piecewise time delay identification and Padé approximation includes: The UAV control channel is segmented into five segments: communication link, flight control processing, ESC commutation, motor electromechanical and propeller aerodynamics. Segmented time delays are identified to obtain segmented time delay parameters and the total composite time delay is obtained based on these parameters. An online time delay adaptive module is set up to determine the current smooth time delay estimate online by cross-correlation calculation between command acceleration and measured acceleration. Based on the obtained total composite time delay, a Padé compensator with variable order is constructed. The Padé order of the Padé compensator is automatically selected based on the control loop operating bandwidth and smooth time delay estimation, and the position loop and attitude loop adopt their own independent Padé order. Construct a sea state / stage adaptive module; execute sea state adaptive rules based on the current sea state spectrum parameters, and adjust the Padé order, compensation bandwidth, and center frequency of the embedded wave frequency notch filter set in the compensation loop online; execute stage adaptive rules based on the stage division of the landing mission; ensure continuous output of the Padé compensator through state mapping at the moment of switching; The expected attitude of the deck is predicted so that the time window for the deck forward prediction is equal to the time delay window of the Padé compensator. The prediction result is coupled to the reference input of the Padé compensator. The control signal after feedforward coupling is activated by the UAV actuator to realize the dynamic landing of the UAV on the moving platform.

[0008] Furthermore, a pseudo-random binary sequence of throttle or attitude commands is injected into the flight control system from the ground or test bench as an excitation. The attitude loop response output by the flight control system is collected, and the equivalent pure time delay between the communication link segment and the flight control processing segment is extracted using least-squares system identification or frequency domain identification methods. and ; Under bench testing conditions, a step-superimposed frequency sweep signal is sent to the ESC as excitation, the motor speed response is collected, and the equivalent pure time delay of the ESC commutation is identified and obtained. With the first-order time constant of the motor , and by Extracting the electromechanical equivalent pure time delay of the motor ; Frequency sweep thrust tests were conducted on the blade unit in a near-water surface or water surface simulated environment. Thrust curves were collected to obtain the equivalent pure time delay of the blade aerodynamics. ; Finally, the total time delay is synthesized. : .

[0009] Furthermore, the online time-delay adaptive module continuously calculates the cross-correlation function between the current command acceleration and the measured acceleration. : ; in, This is the time offset. The current moment; This is the cross-correlation window length; It is an integral dummy variable; Indicates at time Acceleration command issued; Indicates at time The measured acceleration was as follows; When the cross-correlation function When taking the maximum value, the corresponding time offset This is denoted as the optimal time offset. , Optimal time offset As an instantaneous time delay estimate It is then smoothed using a first-order low-pass filter: ; in, A smooth time delay estimate for the current moment; This is the preset low-pass smoothing coefficient; This is a smoothed estimate from the previous time step; Smoothing time delay estimation Total time delay of substitution synthesis As internal parameters of the Padé compensator.

[0010] Furthermore, the operating bandwidth of the control loop is calculated. With smooth time delay estimation product ; in accordance with The value is automatically selected from the Padé order: when First-order Padé is used: ; when Second-order Padé is used: ; when At this time, a third-order or higher Padé is used, or it degenerates into a state predictor based on numerical integration; among which, This is the Laplace transform term representing a time-delay element in control theory. It is a natural constant; For the complex frequency domain variables in the Laplace transform; The time delay parameter corresponding to the time delay element is taken as a smoothed time delay estimate. ; and The preset Padé level switching threshold; Construct Padé compensators for the position loop and attitude loop respectively, where the equivalent time delay of the attitude loop is... The equivalent time delay of the position loop is the total time delay of the synthesis. The Padé order of the two Padé compensators can be chosen independently.

[0011] Furthermore, the Smith predictor controller is embedded into the Padé compensator, thus enabling a time-delay-free nominal model. Parallel to the actual controlled object The difference between the two outputs is fed back to the input of the control algorithm module of the flight control system, so that the closed-loop characteristic equation no longer contains... Item; among which For the attitude / position control transfer function of the UAV, This indicates a pure time delay element present at the input.

[0012] Furthermore, sea state spectrum parameters, including significant wave height, are acquired in real time. With spectral peak period ; Execute sea state adaptive rules: when At that time, the Padé order is taken as first order, and the compensation bandwidth is taken as... ;when At that time, the Padé order is upgraded to second order, and the compensation bandwidth is taken as... And the center frequency is embedded in the compensation loop of the Padé compensator. Wavelength notch filter; center frequency Based on the spectral peak period Online updates: ; when At that time, the Padé order remains second-order or is upgraded to third-order, and the frequency notch filter is upgraded to one that can be tracked online. A drift-adaptive notch filter; where, and To preset sea state classification thresholds, and These are the design values ​​for compensation bandwidth under low and medium sea states, respectively.

[0013] Furthermore, the landing mission is divided into five phases: far approach, alignment, low-altitude hovering, terminal descent, and touchdown confirmation; in the far approach phase, the Padé order is taken as first order, with reference to the forward shift factor. Take a constant value lower than the nominal value; in the alignment stage, take the Padé order as first order. The Padé order is raised to the nominal value; during the low-altitude hovering phase, the Padé order is raised to second order and the wave frequency notch filter is activated. Maintain the nominal value; during the final descent phase, the Padé order is taken as second order. Take a value greater than the nominal value; during the ground contact confirmation phase, the Padé compensator degenerates into a Taylor forward shift of the reference signal: ; in Indicates the forward projection of the future by Taylor. Expected reference position at any given moment; Indicates the reference position at the current moment; Indicates the reference speed at the current moment; The reference acceleration at the current moment; the reference forward coefficient Used in forward push replace ; The equivalent time delay parameter is taken as the smoothed time delay estimate. .

[0014] Furthermore, a window length of [value missing] is used to measure the deck pose sequence for the UAV's upcoming landing. The current position of the deck can be obtained by sliding polynomial fitting or Kalman filtering. ,speed With acceleration The estimate; Smooth time delay estimation As a prediction time window, the extrapolation deck is used with a second-order Taylor expansion. Expected pose at time : ; Expected pose The time delay window eliminated by the Padé compensator is sent as a reference input to the Padé compensator so that it is numerically equal to the time window predicted by the deck push-forward. Finally, the control signal generated by the joint correction calculation of the set of adjustment parameters output by the sea state / stage adaptive module received by the Padé compensator is sent to the UAV actuator to take effect, thus completing the dynamic landing of the moving platform.

[0015] Furthermore, during the drone landing process, continuous monitoring of time lag mismatch is conducted. When the mismatch is greater than the robustness margin The online time-delay adaptive module is re-identified upon triggering; among which, The optimal time offset is obtained by calculating the cross-correlation between the current command acceleration and the measured acceleration when the current command acceleration is triggered again. Assignment; Let be the robustness margin coefficient determined by the small gain theorem.

[0016] A terminal device includes a processor, a memory, and a computer program stored in the memory; when the processor executes the computer program, it implements the UAV sea landing time delay compensation method based on segmented time delay identification and Padé approximation.

[0017] A computer-readable storage medium storing a computer program; when executed by a processor, the computer program implements the UAV sea landing time delay compensation method based on piecewise time delay identification and Padé approximation.

[0018] Compared with the prior art, the present invention has the following technical features: 1. Engineerable segmented time delay identification: By decomposing the total time delay into five physical sources and providing standardized identification protocols for each segment, the dependence of existing Padé compensation methods on known constants is eliminated, significantly improving the portability of the compensator under different hardware / firmware combinations.

[0019] 2. Dual Adaptation for Sea State and Stage: The Padé order and compensation bandwidth are adjusted in real time according to the sea state spectrum parameters and mission stage. It retains the simplicity of low-order compensators in low sea states and provides sufficient compensation margin in strong sea states and terminal landing phase, overcoming the shortcomings of compensators that are fixed throughout the entire process in existing technologies.

[0020] 3. Time window coupling: The Padé compensation time window is strictly equal to the dynamic deck prediction time window, so that the deck attitude targeted when the compensated command reaches the actuator is consistent with the actual attitude, eliminating the systematic landing deviation caused by the decoupling of the two.

[0021] 4. Quantifiable robust margin: Through the online time delay adaptive module and the robust margin formula based on the small gain theorem, the closed loop can still remain stable even if there is a mismatch between the time delay identification value and the actual time delay.

[0022] 5. Specialized for marine scenarios: This invention explicitly covers marine characteristics such as blade aerodynamic lag increment when close to the sea surface and wave spectrum embedded wave frequency trapping, forming a clear technical distinction from land landing and general quadcopter time delay compensation methods. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of a dynamic landing scenario for a marine unmanned aerial vehicle (UAV) or unmanned vessel, which is applied in this invention. Figure 2 This is a flowchart illustrating the overall process of the method of the present invention. Figure 3 A schematic diagram of a five-segment time delay model for the UAV execution channel; Figure 4 This is a structural block diagram of the variable-step Padé compensator of the present invention; Figure 5 A schematic diagram illustrating the coupling relationship between the dynamic deck prediction time window and the Padé compensation time window. Detailed Implementation

[0024] This invention provides a method for compensating for UAV landing time delay at sea based on piecewise time delay identification and Padé approximation. See [link to relevant documentation]. Figure 1 and Figure 2 Specifically, it includes the following steps: Step 1: The UAV control channel is segmented into five segments: communication link, flight control processing, ESC commutation, motor electromechanical and propeller aerodynamics. The segmented time delay parameters are obtained and the total composite time delay is obtained based on them. An online time delay adaptive module is set up to determine the current smooth time delay estimate online by cross-correlation calculation between command acceleration and measured acceleration.

[0025] Step 1.1: Jointly identify the communication link and the flight control processing segment.

[0026] The ground or test bench injects pseudo-random binary sequence (PRBS) throttle or attitude commands into the flight controller as excitation, collects the attitude loop response output by the flight controller, and extracts the equivalent pure time delay between the communication link segment and the flight controller processing segment using least squares system identification or frequency domain identification methods. and ;in, The transmission delay of command signals to the flight control input via the communication link. The processing time delay from reading instructions from the flight controller firmware to generating attitude control signals.

[0027] Step 1.2: Joint identification of ESC and motor electromechanical section.

[0028] Under bench testing conditions, a step-superimposed sweep frequency signal is sent to the ESC as excitation, the motor speed response is collected, and the commutation time delay of the ESC is identified. With the first-order time constant of the motor , and by Extracting equivalent pure time delay ;in, The time delay between the ESC receiving the PWM command and generating the phase current. The first-order time constant of the electromechanical response of the motor. To and The corresponding equivalent pure time delay estimate.

[0029] Step 1.3: Identify the aerodynamic section of the propeller blade separately.

[0030] Frequency sweep thrust tests were conducted on the blade unit in a near-water surface or water surface simulated environment to collect thrust settling curves and obtain the blade aerodynamic settling time constant. With equivalent pure time delay Among these factors, the surface effect occurs when the water surface is close to the sea surface, therefore... This includes the aerodynamic lag increment caused by the water surface effect when the propeller blades are operating close to the sea surface.

[0031] Step 1.4: Determine the total synthesis time delay.

[0032] Finally, the total time delay is synthesized. : ; And in five-dimensional vectors The latency is stored in the form of a delay fingerprint for the current drone hardware / firmware combination. This synthesizes the total latency. The nominal design basis value will be used as the underlying control structure and prediction model of the system, and will be directly used in the definition of the equivalent time delay of the position loop in step 2.3, and in the construction of the time delay-free nominal model of the Smith predictor in step 2.4.

[0033] Step 1.5: Configure the online time-delay adaptive module.

[0034] The online time-delay adaptive module continuously calculates the acceleration of the current command. Compared with measured acceleration cross-correlation function : ; in, This is the time offset. The current moment; This is the cross-correlation window length; Let be the dummy variable for integration, representing any time point within the integration interval; Indicates at time Acceleration command issued; Indicates at time The measured acceleration was as follows; It is an integral infinitesimal element.

[0035] When the cross-correlation function When taking the maximum value, the corresponding time offset This is denoted as the optimal time offset. , Optimal time offset As an instantaneous time delay estimate It is then smoothed using a first-order low-pass filter: ; in, A smooth time delay estimate for the current moment; This is the preset low-pass smoothing coefficient; This is a smoothed estimate from the previous time step.

[0036] Step 1.5 Real-time estimation of smoothed time delay Replace the total time delay obtained in step 1.4 As real-time dynamic compensation parameters for downstream compensators (i.e., the internal parameters of the Padé compensator to be constructed in subsequent step 2), this improves the portability of the compensator under different hardware / firmware combinations; this online estimation process does not rely on the nominal value delay in step 1.4. Instead, it directly measures the actual physical time delay by comparing the command with the measured response, and further eliminates the slow drift caused by battery voltage, motor temperature rise, blade wear, etc. through online adaptive testing.

[0037] In step 1, the total synthesis time delay is calculated. By decomposing the physical source, high reliability can be obtained through the same standardized protocol under different hardware / firmware combinations. The parameters significantly improve the portability of the time delay compensator; and through online adaptation, it further eliminates the slow drift caused by battery voltage, motor temperature rise, blade wear, etc.

[0038] Step 2: Based on the obtained total synthetic time delay, construct a Padé compensator with a variable order; the Padé order of the Padé compensator is automatically selected based on the control loop operating bandwidth and smooth time delay estimation, and the position loop and attitude loop adopt their own independent Padé order.

[0039] Step 2.1: Calculate the operating bandwidth of the control loop (the position closed-loop control channel and attitude closed-loop control channel within the UAV flight control system). With smooth time delay estimation product ; That is, the equivalent time delay estimate corresponding to the loop.

[0040] Step 2.2, based on The value is automatically selected from the Padé order: when First-order Padé is used: ; when Second-order Padé is used: ; when At this time, a third-order or higher Padé is used, or it degenerates into a state predictor based on numerical integration; among which, This is the Laplace transform term representing a time-delay element in control theory. It is a natural constant; For the complex frequency domain variables in the Laplace transform; These are the time delay parameters corresponding to this time delay element; in this scheme, the parameters are... Substitute the value into the smoothed time delay estimate obtained in step 1.5 , which serves as the dynamic time delay input for compensation calculation at the current moment; and The preset Padé level switching threshold, .

[0041] Step 2.3: Construct Padé compensators for the position loop (i.e., the feedback control loop controlling the UAV's three-dimensional spatial position) and the attitude loop (i.e., the feedback control loop controlling the UAV's roll, pitch, and yaw attitude), respectively. The equivalent time delay of the attitude loop is... The equivalent time delay of the position loop is the total composite time delay obtained in step 1.4. The Padé order of the two Padé compensators is selected independently according to step 2.2.

[0042] Step 2.4: Embed the Smith predictor controller into the Padé compensator, i.e., make it a delay-free nominal model. Parallel to the actual controlled object The difference between the two outputs (error signal) is fed back to the input of the control algorithm module (i.e., controller) of the flight control system, so that the closed-loop characteristic equation no longer contains the error signal. Item; among which For the attitude / position control transfer function of the UAV, This indicates a pure time delay element present at the input.

[0043] In step 2, the Padé order is automatically selected based on the current operating bandwidth and the identified time delay, which avoids the unnecessary complexity of high-order compensators under low operating conditions and ensures sufficient compensation accuracy under high operating conditions. The position loop and attitude loop adopt independent orders to match the total time delay decomposition with the multi-level control structure.

[0044] Step 3: Construct the sea state / stage adaptive module; execute the sea state adaptive rules based on the current sea state spectrum parameters, and adjust the Padé order, compensation bandwidth, and center frequency of the embedded wave frequency notch filter set in the compensation loop online; execute the stage adaptive rules based on the stage division of the landing mission; ensure the continuous output of the Padé compensator through state mapping at the moment of switching.

[0045] The processing procedure of the sea state / stage adaptive module is as follows: Step 3.1: Acquire sea state spectrum parameters in real time; the sea state spectrum parameters include at least the significant wave height. With spectral peak period ;in, Characterizes the energy level of sea surface waves. The dominant frequency components characterizing sea surface waves, the sea state spectrum parameters can be provided by shipborne wave sensors, meteorological forecast data, or visual / radar wave inversion modules.

[0046] Step 3.2, execute the sea state adaptive rule; when At that time, the Padé order is taken as first order, and the compensation bandwidth is taken as... ;when At that time, the Padé order is upgraded to second order, and the compensation bandwidth is taken as... And the center frequency is embedded in the compensation loop (i.e., the feedback / feedforward compensation loop used to cancel the input time delay) of the Padé compensator. A wave frequency notch filter; the center frequency of the wave frequency notch filter. Based on the peak period obtained in real time in step 3.1 The calculation is dynamically performed and updated online according to the following formula: ; when At that time, the Padé order remains second-order or is upgraded to third-order, and the frequency notch filter is upgraded to one that can be tracked online. A drift-adaptive notch filter; where, and To preset sea state classification thresholds, for example Corresponding to the light to moderate wave limit in the International Sea State Scale. The corresponding range is from medium to large waves. The specific value can be determined by the wind and wave resistance capability of the UAV and simulation calibration. and These are the design values ​​for compensation bandwidth under low sea state (0 to 3) and moderate sea state (3 to 6), respectively. .

[0047] Step 3.3, execute the phase adaptive rules; divide the landing mission into five phases: far approach, alignment, low-altitude hovering, terminal descent, and touchdown confirmation; in the far approach phase, take the Padé order of first order, referring to the forward shift coefficient. Take a constant value lower than the nominal value; in the alignment stage, take the Padé order as first order. Increase to the nominal value (e.g., set to 1.0); during the low-altitude hovering phase, the Padé order is increased to second order and the wave frequency notch filter from step 3.2 is activated. Maintain the nominal value; during the final descent phase, the Padé order is taken as second order. Take a value greater than the nominal value; during the ground contact confirmation phase, the Padé compensator degenerates into a Taylor forward shift of the reference signal: ; in Indicates the future, after Taylor's projection. The expected reference position at any given moment is the ultimate goal of the guidance commands during the ground contact confirmation phase; Indicates the reference position at the current moment; Indicates the reference speed at the current moment; The reference acceleration at the current moment; the reference forward coefficient Used in forward push replace ; This is the equivalent time delay parameter. In actual real-time control, this parameter is the smoothed time delay estimate obtained in step 1.5. The actual input serves as the basis for calculating the step size of the forward position.

[0048] The reference forward coefficient It only applies to the Taylor forward of the reference signal in the adaptive rule of this stage, without changing the Padé compensation time window in step 2 and the dynamic deck prediction time window in step 4.

[0049] Step 3.4, Bumpless Transfer: When the Padé order or the center frequency of the notch filter changes as described in Step 3.2, or when the landing mission phase changes as described in Step 3.3, the internal state vector of the Padé compensator before the switch (i.e., the internal state space memory variable of the compensator) will be used for the currently running Padé approximation time delay compensator. Through structural continuity mapping Assign to the internal state vector of the compensator after the switch This ensures that the Padé compensator output remains continuous during the switching instant; that is: ; The structural continuous mapping By constraining the continuity of the compensator output and its several derivatives between the old and new order compensators at the moment of switching, the output of the compensator is made continuous at the moment of switching through algebraic solution or pseudo-inverse construction of the new order state equation, thus achieving disturbance-free switching.

[0050] The set of adjustment parameters calculated and output in real time by the sea state / stage adaptive module includes: the currently required Padé order, the compensation bandwidth value, and the center frequency of the wave frequency notch filter. and reference forward coefficients These parameters will serve as dynamic correction inputs to the Padé compensator in subsequent steps to construct the final control signal.

[0051] In this step, the Padé order and compensation bandwidth are adjusted in real time according to the sea state spectrum parameters and mission phase. The simplicity of the low-order compensator is preserved in low sea states, and sufficient compensation margin is provided in strong sea states and terminal landing phase. The seamless switching is achieved through state mapping at the moment of switching, avoiding abrupt changes in the actuator caused by internal state jumps of the compensator.

[0052] Step 4: Predict the expected attitude of the deck so that the time window for the deck forward prediction is equal to the time delay window of the Padé compensator; couple the prediction result to the reference input of the Padé compensator, and the control signal after feedforward coupling is activated by the UAV actuator to realize the dynamic landing of the UAV on the moving platform.

[0053] Step 4.1, create a window with a length of [missing information - likely a value or value] for the pose measurement sequence of the deck (moving platform) where the UAV will land. The current position of the deck can be obtained by sliding polynomial fitting or Kalman filtering. ,speed With acceleration The estimate; of which This represents the number of sampling points for the sliding window.

[0054] Step 4.2, using the smoothed time delay estimate obtained in step 1.5 As a prediction time window, the extrapolation deck is used with a second-order Taylor expansion. Expected pose at time: ; in, Indicates the elapsed time window The expected attitude of the deck after extrapolation.

[0055] Step 4.3, set the expected pose The time delay window eliminated by the Padé compensator is sent as a reference input to the Padé compensator constructed in step 2, so that the time window predicted by the Padé compensator is numerically equal to that predicted by the deck push-forward, that is, the following condition is met: ; Among them, among them, This indicates the length of the time delay window that the Padé compensator design aims to eliminate; This represents the length of the time window for the deck advance prediction, which is directly derived from the smoothed time delay estimate output in step 1.5. Assignment, i.e. This satisfies the strict equality relationship described above.

[0056] Finally, the control signal (i.e., the pulse width modulation (PWM) control command for the UAV motors) generated after the Padé compensator receives the set of adjustment parameters (Padé order, compensation bandwidth, notch filter frequency, and forward coefficient) output by the sea state / stage adaptive module in step 3 and performs joint correction calculations is sent to the UAV actuator to take effect, completing the dynamic landing of the moving platform; during the execution process, time delay mismatch is continuously monitored. When the mismatch is greater than the robustness margin The online time-delay adaptive module described in step 1.5 is re-identified; wherein, The optimal time offset is obtained from the cross-correlation calculation of the current command acceleration and the measured acceleration in step 1.5 when the trigger is triggered again. Assign a value as the engineering replacement quantity for the current actual time delay; Let be the robustness margin coefficient determined by the small gain theorem.

[0057] In step 4 above, the Padé compensation time window is strictly equal to the dynamic deck prediction time window, so that the deck pose targeted when the compensated command reaches the actuator is consistent with the real pose, eliminating the systematic landing deviation caused by the decoupling of the two; the robust margin formula based on the small gain theorem ensures that the closed loop can still remain stable when there is a mismatch between the time delay identification value and the real time delay.

[0058] By performing steps 1 to 4 above, it is possible to achieve stable dynamic landing of UAVs under the combined conditions of long time delay, dynamic platform movement and sea wave disturbance under the conditions of a moving platform at sea.

[0059] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for compensating for UAV landing time delay at sea based on piecewise time delay identification and Padé approximation, characterized in that, include: The UAV control channel is segmented into five segments: communication link, flight control processing, ESC commutation, motor electromechanical and propeller aerodynamics. Segmented time delays are identified to obtain segmented time delay parameters and the total composite time delay is obtained based on these parameters. An online time delay adaptive module is set up to determine the current smooth time delay estimate online by cross-correlation calculation between command acceleration and measured acceleration. Based on the obtained total composite time delay, a Padé compensator with variable order is constructed. The Padé order of the Padé compensator is automatically selected based on the control loop operating bandwidth and smooth time delay estimation, and the position loop and attitude loop adopt their own independent Padé order. Construct a sea state / stage adaptive module; execute sea state adaptive rules based on the current sea state spectrum parameters, and adjust the Padé order, compensation bandwidth, and center frequency of the embedded wave frequency notch filter set in the compensation loop online; execute stage adaptive rules based on the stage division of the landing mission; ensure continuous output of the Padé compensator through state mapping at the moment of switching; The expected attitude of the deck is predicted so that the time window for the deck forward prediction is equal to the time delay window of the Padé compensator. The prediction result is coupled to the reference input of the Padé compensator. The control signal after feedforward coupling is activated by the UAV actuator to realize the dynamic landing of the UAV on the moving platform.

2. The method for compensating for UAV sea landing time delay based on piecewise time delay identification and Padé approximation according to claim 1, characterized in that, The ground or test bench injects a pseudo-random binary sequence of throttle or attitude commands into the flight controller as an excitation, collects the attitude loop response output by the flight controller, and extracts the equivalent pure time delay between the communication link segment and the flight controller processing segment using least squares system identification or frequency domain identification methods. and ; Under bench testing conditions, a step-superimposed frequency sweep signal is sent to the ESC as excitation, the motor speed response is collected, and the equivalent pure time delay of the ESC commutation is identified and obtained. With the first-order time constant of the motor , and by Extracting the electromechanical equivalent pure time delay of the motor ; Frequency sweep thrust tests were conducted on the blade unit in a near-water surface or water surface simulated environment. Thrust curves were collected to obtain the equivalent pure time delay of the blade aerodynamics. ; Finally, the total time delay is synthesized. : 。 3. The method for compensating for UAV sea landing time delay based on piecewise time delay identification and Padé approximation according to claim 1, characterized in that, The online time-delay adaptive module continuously calculates the cross-correlation function between the current command acceleration and the measured acceleration. : ； in, This is the time offset. The current moment; This is the cross-correlation window length; It is an integral dummy variable; Indicates at time Acceleration command issued; Indicates at time The measured acceleration was as follows; When the cross-correlation function When taking the maximum value, the corresponding time offset This is denoted as the optimal time offset. , Optimal time offset As an instantaneous time delay estimate It is then smoothed using a first-order low-pass filter: ； in, A smooth time delay estimate for the current moment; This is the preset low-pass smoothing coefficient; This is a smoothed estimate from the previous time step; Smoothing time delay estimation Total time delay of substitution synthesis As internal parameters of the Padé compensator.

4. The method for compensating for UAV sea landing time delay based on piecewise time delay identification and Padé approximation according to claim 1, characterized in that, Calculate the operating bandwidth of the control loop With smooth time delay estimation product ; in accordance with The value is automatically selected from the Padé order: when First-order Padé is used: ； when Second-order Padé is used: ； when At this time, a third-order or higher Padé is used, or it degenerates into a state predictor based on numerical integration; among which, This is the Laplace transform term representing a time-delay element in control theory. It is a natural constant; For the complex frequency domain variables in the Laplace transform; The time delay parameter corresponding to the time delay element is taken as a smoothed time delay estimate. ; and The preset Padé level switching threshold; Construct Padé compensators for the position loop and attitude loop respectively, where the equivalent time delay of the attitude loop is... The equivalent time delay of the position loop is the total time delay of the synthesis. The Padé order of the two Padé compensators can be chosen independently.

5. The method for compensating for UAV sea landing time delay based on piecewise time delay identification and Padé approximation according to claim 1, characterized in that, Embedding the Smith predictor controller into the Padé compensator enables a time-delay-free nominal model. Parallel to the actual controlled object ; The difference between the two outputs is fed back to the input of the control algorithm module of the flight control system, so that the closed-loop characteristic equation no longer contains... Item; among which For the attitude / position control transfer function of the UAV, This indicates a pure time delay element present at the input.

6. The method for compensating for UAV sea landing time delay based on piecewise time delay identification and Padé approximation according to claim 1, characterized in that, Real-time acquisition of sea state spectrum parameters, including significant wave height With spectral peak period ; Execute sea state adaptive rules: when At that time, the Padé order is taken as first order, and the compensation bandwidth is taken as... ;when At that time, the Padé order is upgraded to second order, and the compensation bandwidth is taken as... And the center frequency is embedded in the compensation loop of the Padé compensator. Wavelength notch filter; center frequency Based on the spectral peak period Online updates: ; when At that time, the Padé order remains second-order or is upgraded to third-order, and the frequency notch filter is upgraded to one that can be tracked online. A drift-adaptive notch filter; where, and To preset sea state classification thresholds, and These are the design values ​​for compensation bandwidth under low and medium sea states, respectively.

7. The method for compensating for UAV sea landing time delay based on piecewise time delay identification and Padé approximation according to claim 1, characterized in that, The landing mission is divided into five phases: far approach, alignment, low-altitude hovering, terminal descent, and touchdown confirmation. In the far approach phase, the Padé order is taken as first order, referencing the forward shift factor. Take a constant value lower than the nominal value; in the alignment stage, take the Padé order as first order. The Padé order is raised to the nominal value; during the low-altitude hovering phase, the Padé order is raised to second order and the wave frequency notch filter is activated. Maintain the nominal value; during the final descent phase, the Padé order is taken as second order. Take a value greater than the nominal value; during the ground contact confirmation phase, the Padé compensator degenerates into a Taylor forward shift of the reference signal: ； in Indicates the future projected forward by Taylor. Expected reference position at any given moment; Indicates the reference position at the current moment; Indicates the reference speed at the current moment; Indicates the reference acceleration at the current moment; The reference forward coefficient Used in forward push replace ; The equivalent time delay parameter is taken as the smoothed time delay estimate. .

8. The method for compensating for UAV sea landing time delay based on piecewise time delay identification and Padé approximation according to claim 1, characterized in that, A window length of [value] is used to measure the deck pose sequence of the UAV before its landing. The current position of the deck can be obtained by sliding polynomial fitting or Kalman filtering. ,speed With acceleration The estimate; Smooth time delay estimation As a prediction time window, the extrapolation deck is used with a second-order Taylor expansion. Expected pose at time : ； Expected pose The time delay window eliminated by the Padé compensator is sent as a reference input to the Padé compensator so that it is numerically equal to the time window predicted by the deck push-forward. Finally, the control signal generated by the joint correction calculation of the set of adjustment parameters output by the sea state / stage adaptive module received by the Padé compensator is sent to the UAV actuator to take effect, thus completing the dynamic landing of the moving platform.

9. The method for compensating for UAV sea landing time delay based on piecewise time delay identification and Padé approximation according to claim 1, characterized in that, Continuous monitoring of time lag mismatch during drone landing execution When the mismatch is greater than the robustness margin The online time-delay adaptive module is re-identified upon triggering; among which, The optimal time offset is obtained by calculating the cross-correlation between the current command acceleration and the measured acceleration when the current command acceleration is triggered again. Assignment; Let be the robustness margin coefficient determined by the small gain theorem.

10. A terminal device, comprising a processor, a memory, and a computer program stored in the memory; characterized in that, When the processor executes the computer program, it implements the UAV sea landing time delay compensation method based on piecewise time delay identification and Padé approximation as described in any one of claims 1-9.

Images