Carrier attitude self-adaptive compensation system in complex motion environment

Abstract

The invention relates to the technical field of industrial control systems, and discloses a carrier attitude adaptive compensation system in a complex motion environment, which comprises a distributed adaptive control module, a multi-dimensional state synchronization module and an axial control module, and is characterized in that the axial control module obtains a transmission time delay state, and when the transmission time delay exceeds a threshold interval, the multi-dimensional state synchronization module carries out adaptive compensation on a carrier attitude; a speed residual error sequence is extracted, the zero-crossing frequency of the speed residual error sequence is counted, an attenuation factor is determined according to the zero-crossing frequency so as to reduce the cross-axis compensation amount amplitude, an axial control module converts deviation energy generated by link congestion into a virtual uncertainty boundary parameter and injects the virtual uncertainty boundary parameter into a stability constraint criterion, and dynamic adjustment of the updating step length is completed. According to the method, the uncertainty of a communication link is quantified into robust gain reserve of a control system, the excitation phenomenon caused by high-frequency disturbance is effectively suppressed, and negative definite convergence of a controlled subject attitude control law is guaranteed.

Description

Technical Field

[0001] This invention relates to an adaptive compensation system for carrier attitude under complex motion environments, belonging to the field of industrial control system technology. Background Technology

[0002] Current carriers typically employ a feedback control architecture based on dynamic models in complex motion environments. They utilize preset proportional, derivative, and integral adjustment logic or sliding mode control laws to drive the actuators to generate compensating torques to maintain attitude stability. To address the time-varying characteristics of the controlled object's parameters during operation, adaptive control logic establishes an error mapping relationship between the reference model and the actual state, and corrects the controller's weights in real time, thereby achieving control closed-loop convergence under model mismatch conditions.

[0003] In high-sea-state operating platforms or high-dynamic off-road environments, the external environmental excitations exhibit wide-bandwidth, nonlinear, and transient impact characteristics, causing high-frequency and drastic evolution of the rotational inertia and damping coefficient in the carrier model. When processing the massive amounts of data generated by such abrupt disturbances, the industrial control bus is prone to transient communication congestion and sampling data loss, resulting in cross-cycle command lag in the feedback control loop. Given that external disturbances may cause severe phase reversals within milliseconds, the existing technology's conventional linear extrapolation path based on historical gradients is prone to generating erroneous compensation energy that is completely opposite to the actual physical disturbance direction within the bus blind zone. This energy injection due to prediction failure not only fails to smooth attitude fluctuations but also superimposes in the same direction as the physical impact, causing the system to trigger irreversible transaxial resonance within a specific frequency band, and even inducing physical damage to the controlled object.

[0004] Therefore, the technical problem to be solved by this invention is how to construct a control architecture that can perceive the dispersion characteristics of physical disturbances in real time and realize adaptive yielding and stable takeover of compensation weights when transient congestion occurs in the control bus, so as to solve the problem of attitude instability of the carrier under the condition of nonlinear wideband disturbance superimposed on communication delay. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A carrier attitude adaptive compensation system under complex motion environments, comprising a distributed adaptive control module, a multi-dimensional state synchronization module, and an axial control module: The axial control module includes a dynamic yielding and robust control unit, which is used to obtain the transmission delay state of the cross-axis nonlinear coupling compensation in the multi-dimensional state synchronization module. When the transmission delay exceeds the preset time window threshold range of 10ms to 20ms, the dynamic yielding and robust control unit extracts the velocity residual sequence between the output velocity of the axial reference model and the actual velocity obtained by the state feedback submodule. The dynamic yielding and robust control unit statistically analyzes the zero-crossing frequency of the velocity residual sequence polarity reversal within the sliding time window, and calculates the value of the attenuation factor based on the negative correlation mapping rule between the zero-crossing frequency and the attenuation factor, so as to reduce the amplitude of the cross-axis nonlinear coupling compensation through the attenuation factor. Meanwhile, the dynamic yielding and robust control unit calculates the virtual uncertainty boundary parameters based on the energy lost due to compensation, and injects the virtual uncertainty boundary parameters as a dynamic bias term into the stability constraint criterion of the axial control module to adjust the gain of the adaptive update step size within the communication interruption interval.

[0006] Preferably, the multi-dimensional state synchronization module is used to introduce the feedback signal from the drive end and to establish a dynamic decoupling image model for the motion characteristics of the controlled subject within the multi-dimensional state synchronization module, thereby separating the internal friction component in the motion vector in the physical state space to extract the partial derivative features of the cross-axis nonlinear coupling compensation quantity; the multi-dimensional state synchronization module distributes the partial derivative features to each axial control module through the multicast distribution submodule.

[0007] Preferably, the dynamic yielding and robust control unit is used to obtain the amplitude change rate of the velocity residual sequence after the transmission delay exceeds the time window threshold interval, and establishes attenuation slope constraint logic based on the positive and negative polarity of the amplitude change rate and the zero-crossing frequency to suppress disturbance components with frequencies above 50Hz.

[0008] Preferably, the axial control module is used to calculate the rate of change of the system energy function in real time through stability constraint criteria, and under the dynamic bias effect generated by the virtual uncertainty boundary parameters, forces the rate of change to be maintained in the negative definite interval, so as to ensure the negative definite convergence of the control law of the controlled subject.

[0009] Preferably, the distributed adaptive control module is used to decouple the 6-DOF nonlinear control equations of the controlled subject into multiple low-order single-axis adaptive sub-modules and a central coordination sub-module, and each axis control module shares the tracking error data of its physical axis system through a multi-dimensional state synchronization module.

[0010] Preferably, the axial control module includes a model reference submodule, which performs online correction of the feedforward compensation gain by running the following steps: Step S11: Establish the ideal state reference of the physical axis system to which the controlled subject belongs; Step S12: Calculate the correction amount of the feedforward compensation gain based on the deviation between the actual speed and the ideal state reference.

[0011] Preferably, the dynamic yielding and robust control unit is used to initiate a smooth switching logic based on the residual amplitude envelope when the transmission delay recovers to within the time window threshold range, so that the attenuation factor linearly recovers from the modulation value to unity gain.

[0012] Preferably, the distributed adaptive control module is used to acquire the 3-axis inertial measurement data of the controlled subject in real time, and to establish a three-dimensional nonlinear coupling matrix on the multi-dimensional state synchronization module based on the inertial measurement data.

[0013] Preferably, the system includes embedded processing modules distributed in each physical axis system of the controlled subject. Each embedded processing module is used to carry the corresponding axial control module and completes the logical interaction of control state data through a multi-dimensional state synchronization module.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. In the adaptive compensation of the carrier attitude, the single-axis adaptive node realizes the dimensionality reduction extraction of the dispersion characteristics of external nonlinear disturbances by monitoring the polarity reversal frequency of the velocity state residual sequence in real time, and dynamically adjusts the exponential decay operator according to the reversal frequency. When the external physical impact causes high-frequency phase reversal, the control unit realizes the instantaneous melting of the failure cross-axis compensation by adaptively increasing the attenuation factor, eliminating the risk of energy mis-injection induced by the phase lag of the traditional linear extrapolation logic, thereby ensuring the negative definite convergence of the attitude control law of the carrier under extreme working conditions.

[0015] 2. The cross-decoupling coordination bus introduces current feedback and position feedback signals from the servo end of the actuator to separate the inherent frictional hysteresis component from the geometric motion vector. This processing mechanism improves the data purity of the global nonlinear coupling matrix, enabling the system to accurately extract the partial derivative characteristics of the cross-axis state, avoid the aliasing of intrinsic nonlinear disturbances and external environmental disturbances in the control frequency domain, and ensure the calculation accuracy and determinism of the carrier attitude compensation command in a multi-degree-of-freedom strongly coupled environment.

[0016] 3. By converting the backoff energy difference generated during bus congestion into a virtual uncertainty boundary parameter, and injecting this parameter as a dynamic bias term into the Lyapunov stability constraint equation of the single-axis adaptive node, this logical mapping method physically quantifies the uncertainty of the communication link into the stability gain reserve of the underlying control loop, forces the single-axis adaptive update step size to generate a positive leap in the data blind zone, realizes the seamless degradation switch of the system from a global cooperative mode to a single-axis local limit defense mode, and improves the stability margin of the carrier when facing sudden impacts. Attached Figure Description

[0017] Figure 1 This is a full-scale flowchart of the closed-loop control of the carrier attitude adaptive compensation system of the present invention; Figure 2This is a breakdown diagram of the system functional architecture and key algorithm logic of the present invention.

[0018] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0020] An adaptive compensation system for carrier attitude under complex motion environments includes a distributed adaptive control module, a multi-dimensional state synchronization module, and an axial control module. The axial control module includes a dynamic yielding and robust control unit, which is used to obtain the transmission delay state of the cross-axis nonlinear coupling compensation in the multi-dimensional state synchronization module. When the transmission delay exceeds the preset time window threshold range of 10ms to 20ms, the dynamic yielding and robust control unit extracts the velocity residual sequence between the output velocity of the axial reference model and the actual velocity obtained by the state feedback submodule. The dynamic yielding and robust control unit statistically analyzes the zero-crossing frequency of the velocity residual sequence polarity reversal within the sliding time window, and calculates the value of the attenuation factor based on the negative correlation mapping rule between the zero-crossing frequency and the attenuation factor, so as to reduce the amplitude of the cross-axis nonlinear coupling compensation through the attenuation factor. Meanwhile, the dynamic yielding and robust control unit calculates the virtual uncertainty boundary parameters based on the energy lost due to compensation, and injects the virtual uncertainty boundary parameters as a dynamic bias term into the stability constraint criterion of the axial control module to adjust the gain of the adaptive update step size within the communication interruption interval.

[0021] Preferably, the multi-dimensional state synchronization module is used to introduce the feedback signal from the drive end and to establish a dynamic decoupling image model for the motion characteristics of the controlled subject within the multi-dimensional state synchronization module, thereby separating the internal friction component in the motion vector in the physical state space to extract the partial derivative features of the cross-axis nonlinear coupling compensation quantity; the multi-dimensional state synchronization module distributes the partial derivative features to each axial control module through the multicast distribution submodule.

[0022] Preferably, the dynamic yielding and robust control unit is used to obtain the amplitude change rate of the velocity residual sequence after the transmission delay exceeds the time window threshold interval, and establishes attenuation slope constraint logic based on the positive and negative polarity of the amplitude change rate and the zero-crossing frequency to suppress disturbance components with frequencies above 50Hz.

[0023] Preferably, the dynamic yielding and robust control unit calculates the attenuation factor λ through a nonlinear mapping function, and the calculation rule is as follows: ,in, For zero-crossing frequency, λ is the width of the sliding time window, k is the preset disturbance energy sensitivity coefficient, and the value of λ ranges from 0.1 to 0.9.

[0024] Preferably, the axial control module is used to calculate the rate of change of the system energy function in real time through stability constraint criteria, and under the dynamic bias effect generated by the virtual uncertainty boundary parameters, forces the rate of change to be maintained in the negative definite interval, so as to ensure the negative definite convergence of the control law of the controlled subject.

[0025] Preferably, the distributed adaptive control module is used to decouple the 6-DOF nonlinear control equations of the controlled subject into multiple low-order single-axis adaptive sub-modules and a central coordination sub-module, and each axis control module shares the tracking error data of its physical axis system through a multi-dimensional state synchronization module.

[0026] Preferably, the axial control module includes a model reference submodule, which performs online correction of the feedforward compensation gain by running the following steps: Step S11: Establish the ideal state reference of the physical axis system to which the controlled subject belongs; Step S12: Calculate the correction amount of the feedforward compensation gain based on the deviation between the actual speed and the ideal state reference.

[0027] Preferably, the dynamic yielding and robust control unit is used to initiate a smooth switching logic based on the residual amplitude envelope when the transmission delay recovers to within the time window threshold range, so that the attenuation factor linearly recovers from the modulation value to unity gain.

[0028] Preferably, the distributed adaptive control module is used to acquire the 3-axis inertial measurement data of the controlled subject in real time, and to establish a three-dimensional nonlinear coupling matrix on the multi-dimensional state synchronization module based on the inertial measurement data.

[0029] Preferably, the system includes embedded processing modules distributed in each physical axis system of the controlled subject. Each embedded processing module is used to carry the corresponding axial control module and completes the logical interaction of control state data through a multi-dimensional state synchronization module.

[0030] Example 1: In the control scenario of a vehicle-mounted directed energy weapon base applied in a high-dynamic off-road environment, the carrier is excited by irregular undulations on the ground, generating a spatial six-degree-of-freedom composite disturbance with wide bandwidth and strong nonlinear characteristics. At this time, the moment of inertia and damping coefficient of the controlled object change with high frequency and time with the nonlinear deformation of the suspension system, resulting in a mismatch between the preset fixed gain control parameters and the actual dynamic model; the state acquisition module captures the first spatial state vector of the geometric center of the carrier in real time at a sampling frequency of 1000Hz. The subscript t represents the time index, and the second state feedback quantities, such as the stator current and rotor position of the servo end of the actuator, are collected synchronously. The multi-dimensional state synchronization module uses the second state feedback quantities to establish a dynamic decoupled image model for the motion characteristics of the controlled subject in the physical state space. By retrieving the friction torque characteristic vector table pre-stored in the controller's read-only memory and combining it with the current servo motor's speed polarity and current amplitude for vector superposition calculation, the cross-axis nonlinear coupling compensation quantity is calculated. Previously, feedforward compensation operators were used to counteract the Coulomb friction and viscous damping components generated by the actuator transmission chain, and partial derivative features characterizing the true energy distribution of external physical disturbances were extracted. In this extracted actuator chain, the system constructs nonlinear state transition functions for each independent physical axis based on the pure rigid body dynamic differential equations after stripping the intrinsic friction, and performs partial differential calculations on the equivalent displacement components of the isolated pure external excitation source with respect to the system's multidimensional generalized coordinates. This results in the compilation of a Jacobian directional gradient array matrixed by the partial derivative elements, which is used to accurately measure the spatial partial derivative rate of the uniaxial excitation torque radiating outwards to the surrounding cross degrees of freedom.

[0031] Each axis control module within the distributed adaptive control module independently maintains the ideal state reference of its respective physical axis system, and calculates the single-axis foundation feedforward compensation weight based on the deviation between the actual speed and the ideal state reference. The correction amount, where the subscript i is the axis index; the model reference submodule within each axial control module is based on the local state error. Using the adaptive update law The weights are updated in real time, where, For local state error, It is a positive definite adaptive gain matrix. The regression state vector is used; simultaneously, the multi-dimensional state synchronization module extracts the motion coupling characteristics between each axis in real time and constructs a three-dimensional nonlinear coupling matrix, calculating and generating the cross-axis nonlinear coupling compensation amount used to offset the energy transfer between axes. The subscript ij is the inter-axis coupling index, and the compensation amount is distributed to each axis control module using the multicast distribution submodule. Through distributed node update and decoupling calculation with the central coordination unit, the nonlinear control equation is decoupled into multiple low-order single-axis adaptive submodules.

[0032] When the vehicle-mounted base experiences a transient congestion on the industrial control bus lasting 15ms due to severe vibration, the dynamic yielding and robust control unit in the axial control module detects that the transmission delay of the cross-axis nonlinear coupling compensation exceeds the preset time window threshold range of 10ms to 20ms, and then establishes a sliding time window in the underlying controller. ,in The width of the sliding time window is defined, and the velocity residual sequence between the output velocity of the axial reference model and the actual velocity captured by the state feedback submodule is extracted in real time. In this extraction and statistical process, to remove the interference of invalid flips induced by high-frequency low-amplitude sensor electrical noise on energy assessment from a physical mechanism, the system uses a dead-zone amplitude discrimination mechanism on the acquired velocity residual sequence before performing sequence polarity determination. Microwave ripple signals with amplitudes lower than the rated motion noise threshold of the controlled object (e.g., 0.02 rad / s) are eliminated. Only impact residuals that penetrate the noise energy envelope and exhibit alternating positive and negative polarities are considered valid events carrying real disturbance kinetic energy and allowed to proceed. The dynamic yielding and robust control unit statistically analyzes the sliding time window. Zero-crossing frequency of polarity reversal in the internal velocity residual sequence The attenuation factor λ is calculated using a nonlinear mapping function, and its calculation rule follows... Where k is the preset disturbance energy sensitivity coefficient, and λ ranges from 0.1 to 0.9; the dynamic yielding and robust control unit uses an exponential decay operator containing a decay factor to perform a reduction-order smoothing operation on the historical cross-axis nonlinear coupling compensation amount of the previous effective cycle, generating a transition compensation amount that matches the current external physical disturbance dispersion characteristics to replace the original cross-axis input, thereby quickly reducing the hysteresis compensation energy when the disturbance direction undergoes high-frequency abrupt changes, eliminating the risk of energy mis-injection induced by data hysteresis. Before starting the state bias correction process, the system defines the transient work reference benchmark of the underlying physical axis system, follows the work and kinetic energy equivalence theorem in rigid body systems, and the kinetic energy drift induced by environmental physical impact is equal to the discrete spatial integral of the unmatched torque diagonal displacement. In order to bridge the order of magnitude gap between the information transmission dimension and the underlying work dimension, the system will handle the data interruption caused by bus transient congestion. The interval is equivalently mapped to the blind zone of undriven coasting where the physical axis loses closed-loop control. During this period, the real-time angular velocity measured by state synchronization is discretely accumulated along the time axis of this time delay interruption to deduce the virtual angular displacement span generated by the controlled carrier during this vacuum period. Then, the compensation torque that should have been sent in real time but was stuck in the bus is directly projected onto this virtual span to perform multiplication and addition. The dynamic yielding and robust control unit extracts the cross-axis nonlinear coupling compensation torque recorded in the previous sampling period and performs multiplication and addition with the current angular displacement deviation data of the axis measured by the state feedback submodule to generate the time-domain integral result of the kinetic energy offset amplitude of the controlled carrier. The system introduces the Lyapunov robust adaptive boundary control theory with nonlinear dissipation term, applies the funnel-shaped boundary constraint operator feedforward to the derivative equation of the underlying control law, and uses the set scalar leakage coefficient to cut off the parameter divergence path in the communication blind zone.

[0033] For this robust suppression architecture, the funnel-shaped boundary constraint operator manifests as an error constraint envelope whose mathematical form continuously and exponentially shrinks over time. Its top corresponds to the initial relaxation error margin during the system's large-scale maneuver startup, while its bottom rigidly converges to the carrier's steady-state accuracy limit. The accompanying scalar leakage coefficient is essentially a tiny constant negative damping term (e.g., 0.01) implanted into the conventional weight update integral module. This allows the adaptive matrix to autonomously release its internal integral gain at this minimal rate when it loses its closed-loop error guidance due to data congestion. This effectively intercepts the risk of simple integral divergence at the underlying physical level. Based on the equivalent transformation relationship between physical boundaries and mathematical constraints, the dynamic yielding and robust control unit calculates virtual uncertainty boundary parameters based on the energy lost through compensation. ,in, Virtual uncertainty boundary parameters are used as dynamic bias terms and injected into the stability constraint criterion of the axial control module. The rate of change of the system energy function is calculated in real time using the stability constraint criterion. Under the dynamic bias generated, the rate of change of the axial control module is forced to remain in the negative definite interval, guiding the positive definite adaptive gain matrix. The step size increases positively within the communication interruption interval, transforming the unreliable global decoupling accuracy into a stable gain reserve for the underlying single-axis control loop. When the transmission delay recovers to within the time window threshold range, the dynamic yielding and robust control unit initiates a smooth switching logic based on the residual amplitude envelope, enabling the attenuation factor to linearly recover from the modulation value to unity gain. This ensures a smooth transition of the system from a single-axis local defense mode to a global cooperative mode. This compensation system quantifies the uncertainty state of the communication link into the compensation gain of the control logic. By actively yielding the failed decoupling energy in the bus blind zone and strengthening the adaptive characteristics of the single-axis node, it achieves stability assurance for the controlled entity under the superposition of extreme shocks and signal delays.

[0034] Example 2: The current test platform has a total load mass of 500 kg. It utilizes a servo drive axis to input a vibration power spectral density function covering a frequency range of 0.1 Hz to 200 Hz with a maximum acceleration of 5.5 g to generate random physical excitation. The platform's data acquisition system has 16-bit quantization accuracy, and its sampling frequency is set to 1000 Hz. This sampling frequency is determined based on Nyquist's sampling theorem, which states that the upper limit of the high-frequency physical disturbance of the carrier in the controlled scenario is 200 Hz. To capture the nonlinear characteristics of the signal and prevent signal aliasing, the sampling frequency needs to be within a certain range. To balance system real-time performance and computational load, the perturbation frequency was set to more than five times the upper limit. A 1000Hz sampling frequency was used. Gaussian white noise with a signal-to-noise ratio of 20dB was superimposed on the original signal captured by the state acquisition module, and a 50Hz power frequency interference with an amplitude of 0.05V was introduced to simulate measurement errors under real industrial electromagnetic conditions. The experiment was divided into control group one, control group two, and the sample group of this invention. Control group one used a fixed attenuation factor of 0.5 and did not include inter-axis decoupling logic, while control group two included cross-axis decoupling logic but had a fixed attenuation factor of 0.5. A sliding time window was used. The 50ms window is set to achieve dynamic response while maintaining frequency resolution, avoiding a window that is too long, leading to lag in zero-crossing frequency capture, or a window that is too short, leading to increased statistical variance. A simulated congestion fault is injected at 2.5s of system operation to compensate for the cross-axis nonlinear coupling. The transmission delay increased from the normal 2ms to 15ms. This delay value exceeds the lower limit of the 10ms to 20ms time window threshold range. This threshold setting for congestion failure has the support basis of the underlying servo frequency response characteristics of the controlled object. According to the factory frequency sweep test evaluation, for the electromechanical servo drive platform, command misalignment with an amplitude of less than 10ms can be naturally filtered out by the inherent hysteresis between the rotor and the structure due to the high frequency limitation. However, once the accumulated hysteresis exceeds the critical line of 20ms, the phase reversal of the residual energy will definitely break through the control bandwidth and trigger reverse resonance excitation. Therefore, the upper and lower limit envelope parameters for triggering fuse degradation are established.

[0035] After a 15ms bus congestion, the pitch axis attitude residual amplitude of control group 1 increased from 0.12° to 1.55° and generated high-frequency oscillations; after the delay exceeded the threshold, control group 2, due to the fixed attenuation factor of 0.5, could not respond to the energy reversal caused by the external 150Hz disturbance, resulting in a phase misalignment between the hysteresis compensation and the current physical state, and the system stopped at 3.1s; the sample group of this invention monitored the sliding time window. Zero-crossing frequency of the internal velocity residual sequence The frequency of vibrations has increased from 8 under normal operating conditions to 22. This adapts to the floating-point register operation boundaries of the underlying control unit and suppresses algebraic saturation caused by high-frequency dense excitation. Before inputting parameters, the system executes an adaptive computational scale reconstruction procedure. The dynamic yielding and robust control unit measures the dispersion envelope characteristics of the external environmental excitation in real time. When the vibration index of the target object is in the normal low-frequency band, the reference calibration loop is triggered, locking the dimensionless parameter 0.2 as the variable representing the sliding time window in the algebraic equation. When bus congestion is detected accompanied by a transient excitation zero-crossing frequency increase, the logic control loop dynamically expands the eigenvalue of the operational variable to a constant 0.673 according to the built-in frequency band energy equivalent mapping table. To counteract the phase hysteresis effect in different frequency bands, the operational variable here is broadened to correspond to the disturbance energy sensitivity operator in the aforementioned attenuation calculation equation. The specific constant value of 0.673 is not an empirical estimate. During the initial system debugging phase, after performing excitation frequency sweep on the test prototype's suspension platform under full load, the physical reciprocal of its first-order natural structural period was extracted and linearly normalized to form an engineering calibration baseline. This baseline is used to maintain the damping toughness of the attenuation slope when a sharp increase in frequency is caused by specific dense vibration frequencies. Based on the variable scale logic modulation, the disturbance energy sensitivity coefficient k of this invention's sample group is set to 0.05. According to the formula... The attenuation factor λ is calculated, and this value is modulated from 0.92 to 0.26 in real time, where λ is the attenuation factor and k is the disturbance energy sensitivity coefficient. For zero-crossing frequency, This is the width of the sliding time window; the modulation process compensates for the transaxis nonlinear coupling. The amplitude is reduced to 26% of the original amplitude; at the same time, virtual uncertainty boundary parameters are injected into the system. This makes the positive definite adaptive gain matrix of the single-axis adaptive node... The main diagonal element is increased from 1.5 to 4.2, enhancing the suppression stiffness of the single-axis control loop against uncompensated residual disturbances. The numerical jump of the main diagonal element follows the linear gain expansion law of the underlying stiffness modulation function. The calculation execution mechanism takes the constant value of the diagonal element of 1.5 retained in the previous stable interaction cycle of the system as the starting point, and superimposes the penalty product obtained by the currently extracted virtual uncertainty boundary parameter and the maximum allowable expansion ratio of the servo (identified and set to 1.8 times). Thus, under the rigid algebraic constraint, the dynamic suppression peak value of 4.2 is automatically calculated to ensure that the gain jump has a strict non-divergence upper limit protection. The measured data shows that the maximum value of the attitude residual of the sample group of the present invention is 0.18° in the entire 15ms congestion cycle, and recovers to below 0.10° within 30ms after the congestion is relieved.

[0036] Gradient tests for different disturbance intensities show that when the external disturbance frequency is below 50Hz, the attenuation factor λ is above 0.75, and the system retains most of the decoupling compensation energy. However, when the disturbance frequency approaches 200Hz, the attenuation factor λ exhibits a monotonically decreasing trend and stabilizes in the saturation region around 0.15. This nonlinear variation confirms that the modulation logic of the attenuation factor λ can spontaneously adjust the defense strength according to the dispersion of the external physical disturbance. The carrier attitude adaptive compensation system achieves stable convergence of the control law under extreme conditions, passing through zero-crossing frequencies. The modulation attenuation mechanism avoids phase excitation caused by hysteresis compensation and utilizes virtual uncertainty boundary parameters. To fill the stability gap caused by energy retreat.

[0037] Example 3: In the control scenario of a shipborne satellite communication antenna stabilization base that has been in service for a long time and has experienced mechanical wear, the increased bearing clearance inside the drive mechanism causes irregular drift of the nonlinear frictional torque of the controlled subject. The first spatial state vector obtained by the multi-dimensional state synchronization module... The model error arises between the second-state feedback quantity and the second-state feedback quantity, where Let t be the first spatial state vector and t be the time index. At this point, the system uses the in-situ calibration procedure in the parameter domain to determine the operator values ​​in the dynamic yielding and robust control unit. In the initial stage of system deployment, the dynamic yielding and robust control unit starts the offline calibration mode, using the servo driver to generate a white noise excitation signal with a frequency step of 10Hz and an amplitude equivalent to the rated load torque system of the carrier. .

[0038] In the excitation signal Under the action of dynamic yielding and robust control unit in sliding time window The velocity output difference of the internally acquired axial reference model is used to statistically analyze the zero-crossing frequency of the velocity residual sequence. Resonant frequency of the system The mapping relationship is such that when the system attitude control law is detected to have critical oscillations and zero-crossing frequency... When the system is in the high-level range, the dynamic yielding and robust control unit determines the reference value of the disturbance energy sensitivity coefficient k. The process of determining this reference value is to adjust the coefficient k so that the attenuation factor λ will compensate for the cross-axis nonlinear coupling at the moment when the system experiences phase reversal. The amplitude is suppressed below the upper limit of the energy that can be absorbed within the system's servo frequency band, and the specific calibration value of the disturbance energy sensitivity coefficient k is selected in the range of 0.04 to 0.08.

[0039] When the carrier attitude adaptive compensation system operates in a high-dynamic phase and the control bus experiences congestion with a transmission delay exceeding 20ms, the dynamic yielding and robust control unit calculates the cross-axis nonlinear coupling compensation amount. The time-domain integral deviation ΔE is given, where ΔE is the time-domain integral deviation, representing the decoupling energy that failed to be injected into the axial control module due to communication delay. The dynamic yielding and robust control unit uses an energy mapping operator to convert the deviation ΔE into virtual uncertainty boundary parameters. Virtual uncertainty boundary parameters The computational logic follows a linear positive correlation rule. As the decoupling energy loss increases, the virtual uncertainty boundary parameters... Correspondingly, the virtual uncertainty boundary parameters within the axial control module increase. As a forced bias term, it participates in the real-time calculation of the adaptive update step size; through this parameter injection, the axial control module forces the single-axis basic feedforward compensation weight. The update rate generates a compensatory jump during decoupling failure, utilizing the local high gain of the single-axis loop to offset the lost global cooperative component, where Here, i represents the axis index, and the single-axis basic feedforward compensation weight is used. When the transmission delay state recovers to a safe range of less than 10ms, the dynamic yielding and robust control unit initiates the mode regression procedure. A smooth switching operator based on logic threshold control is used to linearly weight the attenuation factor λ. The specific processing method of the mode regression procedure is as follows: The dynamic yielding and robust control unit establishes a transition time window with a length of 5 sampling periods. At the beginning of the window, the single-axis local defense weight is set to the maximum, and as the sampling count increases, the cross-axis nonlinear coupling compensation amount is gradually increased according to the linear interpolation algorithm. The injection weights are applied until the attenuation factor λ recovers to unity gain 1.0. This process suppresses system shocks induced by sudden intervention of decoupling torque by converting discrete mode transitions into continuous energy loading. The compensation system utilizes the synergistic mechanism of offline calibration and online mapping to eliminate the influence of parameter uncertainty in the nonlinear disturbance model. By converting the decoupling energy residual into a uniaxial robust gain, the stability of the attitude compensation system during the controlled retreat to cooperative recovery process is maintained.

[0040] Example 4: In the deployment scenario of a deep-sea robot attitude compensation system that is sensitive to initial dynamic parameters, due to the differences in fluid damping characteristics at different operating depths, the axial control module determines the initial discrete distribution of the disturbance energy sensitivity coefficient k through offline parameter calibration before the mission starts. This method drives the carrier to complete a frequency sweep pitch motion with an amplitude of 5° and a frequency step from 0.1Hz to 50Hz in a still water environment. The multi-dimensional state synchronization module collects the first spatial state vector with a sampling frequency of 1000Hz. Based on the system identification model, the frequency response characteristics of the controlled entity are calculated, and the dynamic yielding and robust control unit operates within a sliding time window. Internal extraction of the zero-crossing frequency of velocity residuals in each frequency band By comparing experimental data, a zero-crossing frequency was established. The physical correlation between the system phase margin attenuation rate and the physical correlation between the attenuation factor λ and the system phase margin attenuation rate is determined. Based on this, the value of the attenuation factor λ in the 0.5 gain state when the phase margin drops to the critical point of 30° is determined as the reference input of the disturbance energy sensitivity coefficient k, thereby eliminating the parameter mismatch caused by the difference in inertia of different physical entities under specific dispersion environment.

[0041] When a deep-sea robot encounters a long latency of over 20ms due to limited communication bandwidth and physical disturbances exceeding 100Hz caused by external ocean currents, the dynamic yielding and robust control unit uses a discrete integral operator to calculate the cross-axis nonlinear coupling compensation. The amplitude area residual within the time delay period serves as a physical measure of the time-domain integral deviation ΔE. The energy mapping operator uses a linear transformation function based on the kinetic energy equivalence principle to convert the time-domain integral deviation ΔE into virtual uncertainty boundary parameters. The conversion function transforms the lost decoupling torque energy into a compression gain term of the negative qualitative criterion of the Lyapunov function in the single-axis adaptive update law. When the time-domain integral deviation ΔE increases to 30% of the preset full-scale torque energy, the processing unit in the axial control module automatically converts the positive definite adaptive gain matrix... The update step size coefficient is increased to 1.8 times that of the initial state. This feedback method based on physical residual energy enables the system to maintain the convergence of the attitude axis by utilizing the transient stiffness transition of the single-axis control loop during the failure of the decoupling function.

[0042] Example 5: In a cluster deployment scenario consisting of multiple deep-sea payload robots, the multi-dimensional state synchronization module establishes a global logical clock using a time synchronization protocol during the system access phase, and each axial control module broadcasts the upper limit of the rated output torque of the actuator. And the static stability margin value, where As the upper limit of the rated output torque, the multi-dimensional state synchronization module calculates the cluster load consistency weight factor based on the product of the upper limit of the rated output torque and the static stability margin of each unit. The weighting factor is broadcast to each single-axis adaptive node via a broadcast mechanism to serve as the initial strength benchmark for each unit participating in the global decoupling task.

[0043] When a specific execution unit experiences an amplitude shift in its drive-end feedback signal that lasts for more than 5 sampling cycles due to mechanical wear, the multi-dimensional state synchronization module utilizes a sliding time window. The root mean square error within the range determines the effectiveness of the current model, where The sliding time window width is defined as follows: if the root mean square error exceeds a preset threshold of 0.15, the dynamic yielding and robust control unit will compensate for the cross-axis nonlinear coupling of this unit. The injection weight is adjusted to 0.4 times the current cluster mean, and the background identification algorithm is simultaneously started to complete the in-situ correction of the quality matrix of the unit until the zero-crossing frequency of the velocity residual sequence for 100 consecutive sampling periods is obtained. By restoring the system to a normal distribution range, this procedure eliminates the risk of performance drift in industrial control systems during multi-node collaboration by quantifying the individual differences of the perception execution units.

[0044] To ensure the reproducibility of the above adaptive compensation procedure across different physical terminals, the processing unit of the distributed adaptive control module has a main frequency of no less than 300MHz and a hardware acceleration unit that supports single-precision floating-point operations. Its external bus interface follows the real-time industrial Ethernet protocol to maintain a deterministic data exchange cycle. The inertial sensor in the state acquisition module has a zero-bias stability of no less than 0.01° / h and an internal sampling bandwidth of no less than 2000Hz. When the system executes the initial alignment procedure, the multi-dimensional state synchronization module establishes the attitude reference zero point of the controlled carrier by continuously reading 1024 static sampling points and performing sliding window mean filtering, thereby providing physically deterministic initial state constraints for subsequent nonlinear decoupling operations.

[0045] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A carrier attitude adaptive compensation system in a complex motion environment, characterized in that, The distributed adaptive control module, the multi-dimensional state synchronization module, and the axial control module are included. The axial control module includes a dynamic yielding and robust regulation unit, which is configured to obtain a transmission delay state of a cross-axis nonlinear coupling compensation amount in the multi-dimensional state synchronization module. When the transmission delay exceeds a preset 10 ms to 20 ms time window threshold interval, the dynamic yielding and robust regulation unit extracts a speed residual sequence between an axial reference model output speed and an actual speed obtained by a state feedback submodule. The dynamic yielding and robust regulation unit counts a zero-crossing frequency of polarity reversal of the speed residual sequence in a sliding time window, and calculates a value of a decay factor according to a negative correlation mapping rule between the zero-crossing frequency and the decay factor, so as to weaken an amplitude of the cross-axis nonlinear coupling compensation amount by the decay factor. Meanwhile, the dynamic yielding and robust regulation unit calculates a virtual uncertainty boundary parameter according to a compensation amount loss energy, and injects the virtual uncertainty boundary parameter as a dynamic bias term into a stability constraint criterion of the axial control module, so as to adjust a gain of an adaptive update step within a communication interruption interval.

2. The system of claim 1, wherein: The multi-dimensional state synchronization module is configured to introduce a driving end feedback signal, and establish a dynamic decoupling mapping model for a motion characteristic of a controlled subject inside the multi-dimensional state synchronization module, so as to separate an internal friction component in a motion vector in a physical state space, and extract a partial derivative feature of the cross-axis nonlinear coupling compensation amount. The multi-dimensional state synchronization module distributes the partial derivative feature to each axial control module through a multicast distribution submodule.

3. The system of claim 1, wherein: The dynamic yielding and robust regulation unit is configured to obtain a rate of change of an amplitude of the speed residual sequence after the transmission delay exceeds the time window threshold interval, and establish a decay slope constraint logic according to a positive and negative polarity of the rate of change of the amplitude and the zero-crossing frequency, so as to suppress a disturbance component with a frequency higher than 50 Hz.

4. The system of claim 1, wherein: The axial control module is configured to calculate a rate of change of a system energy function in real time through the stability constraint criterion, and maintain the rate of change in a negative definite interval under a dynamic bias effect of the virtual uncertainty boundary parameter, so as to guarantee a negative definite convergence of a control law of the controlled subject.

5. The system of claim 1, wherein: The distributed adaptive control module is configured to decouple a 6-degree-of-freedom nonlinear control equation of the controlled subject into a plurality of low-order single-axis adaptive submodules and a central coordination submodule, and each axial control module shares tracking error data of a physical axis system thereof through the multi-dimensional state synchronization module.

6. The system of claim 1, wherein: The axial control module includes a model reference submodule, which performs online correction on a feedforward compensation gain by performing the following steps: step S11: establishing an ideal state reference of a physical axis system to which the controlled subject belongs; and step S12: calculating a correction amount of the feedforward compensation gain according to a deviation between an actual speed and the ideal state reference.

7. The system of claim 1, wherein: The dynamic yielding and robust regulation unit is configured to start a smooth switching logic based on a residual amplitude envelope when the transmission delay returns to within the time window threshold interval, so that the decay factor linearly recovers from a modulation value to a unit gain.

8. The system of claim 1, wherein: The distributed adaptive control module is configured to obtain 3-axis inertial measurement data of the controlled subject in real time, and establish a three-dimensional nonlinear coupling matrix on the multi-dimensional state synchronization module according to the inertial measurement data.

9. The system of claim 1, wherein: The system includes embedded processing modules distributed across the physical axes of the controlled entity. Each embedded processing module carries the corresponding axial control module and completes the logical interaction of control state data through a multi-dimensional state synchronization module.

Images