Error compensation control method for a multi-branched redundant cooperative self-stabilizing gimbal

Abstract

The present application relates to the technical field of multi-freedom motion control and self-stabilizing platform, and particularly relates to a kind of error compensation control method of self-stabilizing gimbal under multi-branch chain redundancy cooperation, comprising: collecting pose and disturbance signal based on branch chain physical topology, constructing disturbance propagation path, and identifying main error source branch chain and secondary error branch chain;Orthogonal decomposition is carried out to the error component of main error source branch chain, and the error of each secondary error branch chain is mapped to a unified error coordinate system;Extract the time domain characteristics of direction and amplitude error, generate corresponding compensation signal;According to the disturbance intensity ratio, allocate branch chain weight, and construct redundant branch chain cooperative control instruction;Synchronously drive each branch chain to execute compensation, and verify whether the attitude error converges;If not, correct error identification parameters and disturbance model, and feedback to control cycle.The present application can realize multi-branch chain high dynamic coordination and error self-learning compensation, and improve the attitude stability and disturbance resistance of gimbal.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multi-freedom motion control and self-stabilizing platform, and particularly relates to an error compensation control method of a self-stabilizing holder under multi-branch chain redundancy cooperation. BACKGROUND

[0002] With the wide application of multi-freedom execution platforms in unmanned aerial vehicle nacelles, ground image stabilization systems and ocean observation equipment, self-stabilizing holders with multi-branch chain redundancy driving capability gradually become the key structure form of high-precision stabilization control. Such systems are usually composed of multiple spatially distributed driving branches, each branch having independent attitude adjustment capability and feedback path. In the face of external disturbance or structural coupling uncertainty, the overall attitude is quickly adjusted and stably maintained through the cooperative control between the redundant branches. In order to meet the high robustness requirement in complex application scenarios, the self-stabilizing holder control system needs to have comprehensive capabilities such as high-precision error identification, multi-branch chain cooperative compensation, dynamic feedback correction, etc.

[0003] However, the existing error compensation method of multi-branch chain self-stabilizing holder still has several key technical difficulties. On the one hand, most traditional schemes adopt fixed topology assumption or static branch determination rule, lack dynamic modeling capability of disturbance real propagation path among multiple branches, and are prone to problems such as inaccurate error source identification and offset compensation direction; on the other hand, the existing method is mostly based on single-branch error closed-loop control, and it is difficult to realize dynamic fusion and weight coordination of compensation signals between different branches, resulting in insufficient overall compensation capability of the holder in strong coupling disturbance scenarios. In addition, the existing system generally lacks feedback learning and disturbance model self-updating mechanism in the case of non-converged error, and cannot realize continuous optimization and stable iteration under complex disturbance or multi-source error conditions. SUMMARY

[0004] Based on the above purpose, the present application provides an error compensation control method of a self-stabilizing holder under multi-branch chain redundancy cooperation, which provides a multi-branch chain redundancy cooperative control method integrating disturbance path identification, error mapping fusion, weight distribution and feedback learning mechanism, to realize the error compensation control target of high precision, high responsiveness and strong stability.

[0005] An error compensation control method of a self-stabilizing holder under multi-branch chain redundancy cooperation, comprising the following steps:

[0006] S1: based on the physical connection topology of the holder branches, real-time acquisition of the pose signals and disturbance signals of each branch, identification of the dynamic disturbance propagation path, determination of the main error source branch and the secondary error branch;

[0007] S2: orthogonal decomposition of the error components of the main error source branch to obtain the directional error component and the amplitude error component, and mapping of the error components of the secondary error branch to the error coordinate system of the main error source branch;

[0008] S3: generating a direction compensation signal and an amplitude compensation signal according to time domain characteristics of the direction error component and the amplitude error component, the direction compensation signal being used to adjust a driving direction of the branch, and the amplitude compensation signal being used to correct a driving output quantity;

[0009] S4: dynamically allocating a cooperative compensation weight of each branch based on a disturbance intensity ratio of the primary error source branch and the secondary error branch, and generating a redundant branch cooperative control instruction;

[0010] S5: synchronously driving each branch to perform a compensation action according to the redundant branch cooperative control instruction, and verifying whether a gimbal attitude error after compensation converges to a preset threshold in real time;

[0011] S6: if the attitude error does not converge, returning to S1 to update an error source identification result until the error meets a stable condition.

[0012] Optionally, the S1 comprises:

[0013] S11: constructing an adjacency matrix of a motion coupling relationship among the branches based on a physical connection topology of the gimbal branches;

[0014] S12: synchronously and in real time collecting pose signals and disturbance signals of all the branches through sensors arranged at joints of the branches;

[0015] S13: inputting the pose signals and the disturbance signals into a pre-trained disturbance propagation model, combining the adjacency matrix of the physical connection topology, and calculating a transmission gain of the disturbance among the branches;

[0016] S14: constructing a dynamic disturbance propagation path according to the transmission gain, the path representing a diffusion direction and intensity of the disturbance from a source branch to an associated branch;

[0017] S15: identifying an initial disturbance source as the primary error source branch based on a transmission gain extreme value in the dynamic disturbance propagation path, and marking an associated branch affected by the transmission of the initial disturbance source as the secondary error branch.

[0018] Optionally, the S2 comprises:

[0019] S21: extracting an error component of the primary error source branch, and orthogonally decomposing the error component with respect to a driving coordinate system of the branch;

[0020] S22: in the driving coordinate system, defining a tangential projection of the decomposed error component as a direction error component, defining a normal projection of the decomposed error component as an amplitude error component, and establishing an error coordinate system of the primary error source branch;

[0021] S23: Obtain error components of all secondary error branches, and analyze pose transformation relationship between each secondary error branch and error coordinate system of the main error source branch;

[0022] S24: Based on the pose transformation relationship, map the error component of each secondary error branch to the error coordinate system of the main error source branch through a homogeneous transformation matrix, to generate a mapped error component of the unified reference.

[0023] Optionally, the S3 comprises:

[0024] S31: Perform time-frequency analysis on the direction error component and the amplitude error component respectively, and extract time domain features, the time domain features comprising phase offset of the direction error component and envelope fluctuation of the amplitude error component;

[0025] S32: Calculate angle compensation of the driving shaft according to the phase offset of the direction error component, and generate a direction compensation signal in negative feedback relationship with the phase offset;

[0026] S33: Calculate amplitude correction of the driving force according to the envelope fluctuation of the amplitude error component, and generate an amplitude compensation signal in inverse proportional relationship with the envelope fluctuation.

[0027] Optionally, the S4 comprises:

[0028] S41: Calculate disturbance energy integral values of the main error source branch and each secondary error branch respectively, the disturbance energy integral value being a square integral of a disturbance signal in time domain;

[0029] S42: Divide the disturbance energy integral value of each secondary error branch by the disturbance energy integral value of the main error source branch, to obtain a disturbance intensity ratio of the main error source branch and each secondary error branch.

[0030] Optionally, the S4 further comprises:

[0031] S43: According to the disturbance intensity ratio, distribute cooperative compensation weights of each branch in inverse proportional relationship, wherein the weight of the main error source branch is , and the weight of the secondary error branch is , is the disturbance intensity ratio of the i-th secondary branch;

[0032] S44: Fuse the direction compensation signal, the amplitude compensation signal and the corresponding cooperative compensation weight of each branch, to generate a redundant branch cooperative control instruction.

[0033] Optionally, the S5 comprises:

[0034] S51: analyze the redundant branch cooperative control instruction, extract the direction compensation amount and amplitude compensation amount of each branch, and generate corresponding motor driving pulse signals;

[0035] S52: synchronously send the motor driving pulse signals to the execution motors of all branches to drive each branch to synchronously execute compensation actions.

[0036] Optionally, the S5 further comprises:

[0037] S53: real-time acquisition of the gimbal attitude data after compensation through the pose sensors of each branch, and calculation of the current gimbal attitude error;

[0038] S54: comparison of the current gimbal attitude error with a preset threshold value to determine whether the attitude error converges to the preset threshold value range.

[0039] Optionally, the S6 comprises:

[0040] S61: if the attitude error does not converge to the preset threshold value range, record the non-converged attitude error of the current gimbal and the corresponding branch motion state;

[0041] S62: correction of the identification parameters of the primary error source branch and the secondary error branch based on the non-converged attitude error, and update of the disturbance propagation model;

[0042] S63: feedback of the updated disturbance propagation model to S1 to restart the error compensation control cycle.

[0043] The beneficial effects of the present application are:

[0044] In the S1, the present application introduces a dynamic disturbance path identification mechanism based on an adjacency matrix and a disturbance propagation model, which can extract the transfer gain from the real-time collected pose signals and disturbance signals, construct a disturbance diffusion path diagram, and accurately identify the primary error source branch and the secondary error branch. Compared with the traditional experience method or fixed model method, the present application has stronger environmental adaptability and dynamic identification ability, provides accurate branch division basis for subsequent compensation strategies, and effectively avoids the problems of miscompensation or system oscillation caused by misjudgment.

[0045] In the present application, an error coordinate system is constructed in the multi-branch cooperative control scene and an error component orthogonal decomposition is performed, a homogeneous transformation matrix is introduced to map the error component of the secondary error branch to the error coordinate system of the primary error source branch, and the fusion analysis of the error components of different branches under the unified reference is realized. The phase offset of the direction error component and the envelope fluctuation of the amplitude error component are extracted by combining the time-frequency feature extraction algorithm, the dynamic compensation signals of the direction and amplitude are constructed, and the sensitivity and control accuracy of the compensation strategy to different types of errors are significantly enhanced.

[0046] The application constructs a disturbance intensity ratio driven cooperative compensation weight distribution model, ensures that the main error source branch and the secondary error branch participate in compensation according to the contribution degree, generates a redundant branch cooperative control instruction through the weighted fusion of a multi-branch direction compensation signal and an amplitude compensation signal, and realizes system level compensation execution; in an error non-converged scene, the main / secondary error branch identification parameters and the disturbance propagation model are dynamically modified through the associated feedback of the attitude error and the branch motion state, and a feedback closed loop control process is formed, so that the attitude convergence speed and the self-stabilization robustness of the system in a strong disturbance and complex coupling environment are improved. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only illustrate the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0048] Fig. 1 The method flowchart of the embodiment of the present application is shown in the figure.

[0049] Fig. 2 The S5 flowchart of the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0050] The present application will be described in detail below in combination with the drawings and specific embodiments. It should be noted here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and other alternative ways can also be used by those skilled in the art to implement them; and the drawings are only used to more specifically describe the embodiments, and are not intended to specifically limit the present application.

[0051] As shown in the figure, a multi-branch redundant cooperative self-stabilizing gimbal error compensation control method includes the following steps: Figs. 1-2

[0052] S1: based on the physical connection topology of the gimbal branch, real-time acquisition of the pose signal and the disturbance signal of each branch, identification of the dynamic disturbance propagation path, determination of the main error source branch and the secondary error branch;

[0053] S2: orthogonal decomposition of the error component of the main error source branch, obtaining the direction error component and the amplitude error component, and mapping the error component of the secondary error branch to the error coordinate system of the main error source branch;

[0054] S3: generating a direction compensation signal and an amplitude compensation signal according to the time domain characteristics of the direction error component and the amplitude error component, the direction compensation signal being used to adjust the driving direction of the branch, and the amplitude compensation signal being used to correct the driving output.​

[0055] S4: dynamically allocating the collaborative compensation weight of each branch based on the disturbance intensity ratio of the primary error source branch and the secondary error branch, and generating a redundant branch collaborative control instruction;

[0056] S5: synchronously driving each branch to perform compensation actions according to the redundant branch collaborative control instruction, and verifying in real time whether the gimbal attitude error after compensation converges to a preset threshold;

[0057] S6: if the attitude error does not converge, return to S1 to update the error source identification result until the error meets the stability condition.

[0058] S1 includes:

[0059] S11, adjacency matrix construction: first, an adjacency matrix representing the motion coupling relationship between branches is constructed according to the physical connection topology of the gimbal branch, denoted as , wherein n is the number of branches, and the matrix element represents the direct coupling relationship between branch i and branch j. If there is a rigid connection or a collaborative control mechanism between two branches, then:

[0060] ;

[0061] For example, in a typical three-branch redundant gimbal, branch 1 and branch 2, branch 1 and branch 3 have a collaborative driving relationship, but branch 2 and branch 3 have no direct connection, then the adjacency matrix is as follows:

[0062] ;

[0063] S12, pose and disturbance signal acquisition: high-precision sensors are deployed at the key joint positions of each branch to synchronously acquire the pose signal and the disturbance signal of each branch, wherein:

[0064] The pose signal represents the spatial position and attitude of branch i;

[0065] The disturbance signal represents the disturbance force, joint torque and angular velocity on the branch. All data will be synchronously input to the subsequent processing module to ensure the timing consistency between multiple branches.

[0066] S13, disturbance propagation gain calculation: taking the above-mentioned signals as input, combining the adjacency matrix A, inputting a pre-trained disturbance propagation model , calculating the transfer gain matrix of the disturbance between branches, wherein the element represents the intensity of the disturbance transferred from branch i to branch j at time t. Its basic expression is:

[0067] ;

[0068] wherein, represents the set of all branch disturbance signals. The model is fitted based on the graph convolutional neural network (GCN) or disturbance energy attenuation law, which has the modeling ability of dynamic characteristics across branch conduction.

[0069] S14, dynamic disturbance propagation path construction: according to the calculated transfer gain matrix , the dynamic disturbance propagation path graph is constructed , wherein the path direction is determined by , and the path strength is measured by the corresponding gain value. The propagation path graph is a directed weighted graph structure, which truly reflects the diffusion path of disturbance from the source branch to the affected branch.

[0070] For example, if , , it indicates that the disturbance of branch 1 has a high intensity impact on branch 2, and branch 2 further affects branch 3, but the impact degree attenuates step by step.

[0071] S15, error source identification and branch classification: finally, the starting point of the path with the largest transfer gain value is selected from the dynamic disturbance propagation path as the main error source branch , and its calculation method is:

[0072] ;

[0073] that is, the disturbance influence intensity of a branch on all other branches is accumulated, and the maximum one is the initial disturbance source. Those affected by the main error source branch and (wherein is the propagation intensity threshold) are marked as secondary error branches .

[0074] For example, in a three-branch redundant structure, if the disturbance transfer gain matrix is:

[0075] ;

[0076] then the total output disturbance of branch 1 is 0.88+0.73=1.61, which is higher than that of branch 2 and branch 3, and branch 1 is identified as the main error source branch, and the other two are secondary error branches.

[0077] S2 includes:

[0078] S21, error component extraction and orthogonal decomposition: first, the instantaneous error component of the main error source branch is extracted, which is defined as: ;

[0079] wherein, desired pose, actual pose;

[0080] Both poses contain a three-dimensional position vector and an Euler angle or quaternion attitude vector.

[0081] The error component is projected to the driving coordinate system of the main error source branch, denoted as , whose basis vectors represent the driving direction (tangential) and the amplitude direction (normal), respectively. The orthogonal decomposition is expressed as: ;

[0082] where represents the directional error component, represents the amplitude error component, and is maintained to ensure orthogonality.

[0083] S22, error coordinate system establishment: in the driving coordinate system , the tangential direction is taken as the X-axis and the normal direction as the Y-axis to construct a two-dimensional error subspace as the error coordinate system for subsequent unified analysis of multiple branch errors. In this coordinate system, the error of the main error source branch can be represented as a two-dimensional vector: ;

[0084] This coordinate system serves as the target reference, and the error components of all secondary error branches are mapped to this coordinate system through transformation.

[0085] S23, secondary error branch pose transformation analysis: for each secondary error branch , its error component is extracted, and based on the spatial geometric relationship between the main error source branch and the branch , the pose transformation relationship between them is calculated. This transformation relationship can be represented as a quaternion rotation and a translation vector , or a homogeneous transformation matrix is constructed:

[0086] ;

[0087] where is the rotation matrix between the main error source branch and the secondary branch, is the translation vector.

[0088] S24, error homogeneous mapping: the transformation relationship is obtained through offline calibration or real-time estimation, reflecting the spatial mapping of the two coordinate systems after the current mechanism structure deformation. The error component of each secondary error branch is mapped to the error coordinate system of the main error source branch, and the process is as follows:

[0089] The error vector is extended to homogeneous coordinates: ;

[0090] Mapping by homogeneous transformation matrix: ;

[0091] Extract the first 3 dimensions as the mapped error components after transformation: ;

[0092] Thus, the errors of all secondary error branches are expressed in the error coordinate system of the primary error source branch , forming a unified dimension reference, facilitating vector synthesis and weighting in subsequent collaborative compensation strategies.

[0093] Example: Suppose the error vectors of the primary error source branch and the secondary error branch are:

[0094] ;

[0095] The unit vector of the primary error source branch driving direction is , the normal direction is , the rotation matrix is , and the translation is zero. Then the mapped error components are still the original error vectors, the direction error is 1.5, and the amplitude error is 0.5.

[0096] S3 includes:

[0097] S31, time domain feature extraction of direction error component and amplitude error component: first, perform time-frequency analysis on the direction error component and amplitude error component obtained in the unified error coordinate system to extract feature parameters for compensation calculation.

[0098] For , use Hilbert transform to construct the analytic signal:

[0099] ;

[0100] The phase offset is calculated by its instantaneous phase: ;

[0101] Perform envelope extraction on to obtain the envelope fluctuation :

[0102] ;

[0103] Then, within the set analysis window , calculate:

[0104] Phase offset of direction error component ;

[0105] Envelope fluctuation of amplitude error component , wherein std represents standard deviation.

[0106] It should be noted that in the present application, the "direction error component" and the "amplitude error component" can represent the error vector decomposition result at a certain moment, or the continuous change sequence thereof within a certain time window. Specifically:

[0107] "Direction error component" represents the projection component of the error vector in the driving direction of the main error source branch at a certain fixed moment, which is a scalar static quantity;

[0108] "Direction error component over time curve" represents the change trajectory of the direction error component in the time dimension during continuous sampling, and is used to characterize the dynamic evolution behavior of the error;

[0109] Similarly, the "amplitude error component" and its time sequence form represent the instantaneous and dynamic error information in the normal direction.

[0110] Therefore, the time-frequency analysis of the direction error component and the amplitude error component in S31 means that the continuous sampling data of and are processed within a defined time window to extract dynamic features such as phase offset and envelope fluctuation, rather than static error value analysis at a single moment.

[0111] S32, direction compensation signal generation: according to the extracted phase offset , a direction compensation mechanism is constructed to generate a direction compensation signal in negative feedback relationship with the phase offset, which is used to correct the deflection angle of the branch driving direction.

[0112] The compensation angle is calculated as: ;

[0113] wherein is the direction compensation gain coefficient, the value range of which is set according to the sensitivity of the actual control system.

[0114] The direction compensation signal is input to the branch pin driving controller in digital form to directly adjust the actuator attitude angle output.

[0115] Example: if the current phase offset is rad, and the compensation gain coefficient is set to , the direction compensation angle is: ;

[0116] S33, amplitude compensation signal generation: according to the envelope fluctuation , construct the amplitude adjustment mechanism of the driving force output, generate the amplitude compensation signal in inverse proportional relationship with the envelope fluctuation , for adjusting the output thrust of the driver. Its calculation method is:

[0117] Where, is the rated driving amplitude, is the adjustment sensitivity coefficient.

[0118] Example: if =100N, =5, the current envelope fluctuation is , then:

[0119] ;

[0120] The amplitude compensation signal is input to the driver control module, which adjusts the thrust output in real time and suppresses the fluctuation of the force value caused by disturbance coupling.

[0121] S4 includes:

[0122] S41, disturbance energy integral value calculation: square integrate the disturbance signals of the main error source branch and all secondary error branches in the observation period, extract their total disturbance energy as an index to quantify their disturbance action ability. Let:

[0123] The disturbance signal of the main error source branch;

[0124] The disturbance signal of the i-th secondary error branch;

[0125] T is the integral window length;

[0126] Then the corresponding disturbance energy integral values are respectively:

[0127] ;

[0128] ;

[0129] Where, represents the Euclidean norm, that is, the length of the three-dimensional disturbance signal.

[0130] Example: if the sampling time window is 2 seconds, the disturbance signal is a three-dimensional force / torque vector, and the sampling is at 100Hz, then the integral can be realized by numerical approximation (such as trapezoidal method).

[0131] S42, disturbance intensity ratio calculation: to measure the disturbance influence degree of the secondary error branch relative to the main error source branch, the disturbance energy integral value of each secondary error branch is divided by the disturbance energy integral value of the main error source branch to obtain the corresponding disturbance intensity ratio :

[0132] ;

[0133] wherein, if approaches 0, it indicates that the secondary branch is less affected by the main error source disturbance; if , it indicates that the disturbance energy of the secondary branch is strong, which may constitute a significant synergistic compensation factor.

[0134] S43, synergistic compensation weight distribution: according to the aforementioned disturbance intensity ratio , the synergistic compensation weights of each branch are distributed in an inverse proportional manner, so that the secondary branch with stronger disturbance intensity obtains higher compensation participation, and the specific weight distribution formula is as follows:

[0135] The compensation weight of the main error source branch: ;

[0136] The compensation weight of the i-th secondary error branch: ;

[0137] wherein, N is the total number of secondary error branches.

[0138] All weights satisfy the normalization condition: ;

[0139] Example: assuming that a system contains two secondary error branches, the disturbance intensity ratios are , , and the weight calculation is as follows:

[0140] ;

[0141] ;

[0142] ;

[0143] S44, redundant branch synergistic control instruction generation: combine the direction compensation signal and the amplitude compensation signal of each branch with the corresponding synergistic compensation weight to generate the redundant branch synergistic control instruction of the system level by using the weighted synthesis strategy, and the structure is as follows:

[0144] ;

[0145] The control instruction is composed of two components:

[0146] Weighted direction compensation synthesis (for adjusting the branch attitude output);

[0147] Weighted amplitude compensation synthesis (for adjusting the branch thrust or driving force output).

[0148] The control instruction is synchronized and issued to each branch execution unit through the control bus or drive interface, ensuring global consistency and local response sensitivity of the collaborative compensation strategy.

[0149] It should be noted that the direction compensation signal and the amplitude compensation signal are used to describe the error response of the main error source branch in the time dimension. In S44, in order to realize the collaborative fusion control of multiple branches, the compensation signals of the main error source branch and all secondary error branches are identified as , and , (i is the secondary branch number), thereby constructing a unified weighted fusion model. The above symbols represent semantic expansion at different control levels, aiming to distinguish the processing goals of single branch compensation calculation and multi-branch compensation synthesis.

[0150] S5 includes:

[0151] S51, control instruction analysis and pulse signal generation: the redundant branch collaborative control instruction constructed in the foregoing S4 is decomposed into the direction compensation amount and the amplitude compensation amount of each branch, and the motor drive pulse signal that can be used for the actuator is generated accordingly. The specific process is as follows:

[0152] Let the compensation amount of the ith branch be:

[0153] The direction compensation angle is: ;

[0154] The driving force output amplitude is: ;

[0155] The pulse width modulation (PWM) parameter of the driving motor is set as follows:

[0156] The angle control PWM duty cycle is: ;

[0157] The driving amplitude PWM duty cycle is: ;

[0158] Wherein, are the pulse width modulation gain factors of the angle and thrust compensation respectively, : control the output of the execution motor position loop and force loop respectively.

[0159] Example:

[0160] If the branch chain direction compensation angle is rad, the driving amplitude is N, and the gain factor is , ,

[0161] ;

[0162] Convert it into PWM control parameters for microcontroller driving module setting.

[0163] S52, synchronous driving execution: load the motor driving pulse signals corresponding to each branch chain to the respective servo drive controller, and realize the synchronous execution compensation action of all branch chains through the bus scheduling mechanism or high-speed synchronous trigger interrupt mechanism. The control command is coordinated in time sequence through unified timestamp or synchronous trigger signal to ensure the consistency of the attitude adjustment action under the redundant structure.

[0164] S53, attitude error collection and calculation: real-time acquisition of the overall gimbal attitude data after compensation execution through the deployment of the pose sensor (such as inertial measurement unit, rotary encoder, gyroscope) at each branch joint or the end, denoted as:

[0165] ;

[0166] and the system set expected attitude value: ;

[0167] Difference operation is performed to obtain the current gimbal attitude error: ;

[0168] The error can be normalized according to Euclidean norm and converted into attitude error scalar: ;

[0169] S54, attitude error convergence determination: compare the current calculated attitude error with the system set preset error threshold to determine whether the attitude error convergence condition is met: ;

[0170] If it is determined that the condition is met (i.e., the gimbal attitude error has converged), it means that the current round of collaborative compensation has been completed, and the system can enter a stable state or maintain the current control; if the convergence condition is not met, the error source identification and compensation iteration process is restarted in step S6 to ensure the dynamic self-stabilization closed-loop characteristics of the system.

[0171] Example:

[0172] If the current attitude error is = 0.018 rad, and the system set threshold is = 0.02 rad, it is considered that the compensation has reached the stable condition, and the system enters the maintenance mode.

[0173] S6 includes:

[0174] S61, the non-convergent error record is associated with the state: when the current gimbal attitude error calculated in S5 does not meet the convergence condition (i.e. ), the system will automatically record the key state parameters at this moment, including:

[0175] the current non-convergent attitude error vector ;

[0176] the actual motion state of each branch at this moment, including the pose signal driving output (such as angular velocity, driving force, compensation instruction history record;

[0177] The recorded data will be stored in the state cache queue to form an error-driving-response mapping sample pair under abnormal working conditions, which is used for subsequent model correction processing;

[0178] S62, disturbance propagation model correction: based on the above recorded non-convergent attitude error , the identification parameters of the main error source branch and the secondary error branch are adjusted, and the key function structure for calculating the disturbance intensity ratio and the transfer gain in the disturbance propagation model is corrected. The correction method includes two aspects:

[0179] Identification parameter correction: the determination of the main error source branch in the original model is based on the principle of maximum disturbance gain accumulation value: ;

[0180] When the error does not converge, the error weighting term needs to be introduced to correct the identification strategy by combining the current and the corresponding branch disturbance response difference: ;

[0181] wherein, is the error response difference between the branches, is the error response weighting coefficient (empirical setting);

[0182] Disturbance propagation model update: the disturbance propagation model is usually expressed in the form of a graph neural network or an energy recursive function, denoted as: ;

[0183] wherein, is the disturbance transfer gain matrix at the current moment, The gain correction term based on the historical error bias is in the form of: ;

[0184] respectively a step coefficient, a feedback sensitivity factor, an error gain mapping vector.

[0185] S63, feedback to S1 to start a new round of error compensation control cycle: the modified disturbance propagation model and the new error source identification result 、 feedback to S1 to re-execute the adjacency matrix update, dynamic disturbance path reconstruction, and primary / secondary error branch division;

[0186] to start a new round of error compensation control cycle. This mechanism has the ability of self-learning and parameter adaptive adjustment, and is particularly suitable for uncertain control scenarios with multiple source disturbances and multiple path propagation in a strongly coupled redundant structure.

[0187] Supplementary explanation: the system can update the disturbance model iteratively each time the error does not converge, but to prevent overfitting or oscillation, the upper limit of the number of iterations or the threshold of the convergence enhancement coefficient can be set as the termination condition.

[0188] The present application covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present application. In order to make the public have a thorough understanding of the present application, specific details are described in the following preferred embodiments of the present application, and the present application can also be fully understood without the description of these details to those skilled in the art. In addition, in order to avoid unnecessary confusion to the essence of the present application, well-known methods, processes, procedures, elements and circuits, etc. are not described in detail.

[0189] The above is only the preferred embodiment of the present application, and it should be pointed out that for ordinary skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should be regarded as the protection scope of the present application.

Claims

1. A method for error compensation control of a self-stabilized pan / tilt platform under multi-branch redundant coordination, characterized in that: The following steps are involved: S1: Based on the physical connection topology of the gimbal branches, the pose signals and disturbance signals of each branch are collected in real time, the dynamic disturbance propagation path is identified, and the main error source branch and secondary error branch are determined; Said S1 comprises: S11: Based on the physical connection topology of the gimbal branches, an adjacency matrix of the motion coupling relationship between each branch is constructed; S12: Sensors deployed at the joints of each branch chain are used to synchronously collect the posture signals and disturbance signals of all branches in real time; S13: Inputting the posture signal and the disturbance signal into the pre-trained disturbance propagation model, combining the adjacency matrix of the physical connection topology, and calculating the transmission gain of the disturbance between branches; S14: constructing a dynamic disturbance propagation path according to the transfer gain, where the path represents the diffusion direction and intensity of the disturbance from the source branch to the associated branches; S15: Based on the extreme value of the transfer gain in the dynamic disturbance propagation path, the initial disturbance source is identified as the main error source branch, and the associated branches affected by its propagation are marked as secondary error branches; S2: Orthogonalize and decompose the error components of the main error source branch chain to obtain the direction error component and the amplitude error component, and map the error components of the secondary error branch chain to the error coordinate system of the main error source branch chain; S3: generating a direction compensation signal and an amplitude compensation signal according to the time domain characteristics of the direction error component and the amplitude error component, wherein the direction compensation signal is used to adjust the driving direction of the branch chain, and the amplitude compensation signal is used to correct the driving output; S4: Based on the disturbance intensity ratio between the primary error source branch and the secondary error branch, the coordinated compensation weight of each branch is dynamically allocated to generate the coordinated control instructions for the redundant branches. The S4 includes: S41: Calculate the disturbance energy integral value of the main error source branch and each secondary error branch respectively, wherein the disturbance energy integral value is the square integral of the disturbance signal in the time domain; S42: dividing the disturbance energy integral value of each secondary error branch by the disturbance energy integral value of the main error source branch to obtain a disturbance intensity ratio of the main error source branch to each secondary error branch; S43: According to the disturbance intensity ratio, the synergistic compensation weights of each branch are allocated in an inverse proportional relationship, wherein the weight of the main error source branch is , the weight of the secondary error branch is , is the perturbation intensity ratio of the i-th secondary branch; S44: Fusion of the direction compensation signal, amplitude compensation signal and corresponding collaborative compensation weight of each branch to generate redundant branch collaborative control instructions; S5: Synchronously drive each branch chain to perform compensation actions according to the redundant branch chain collaborative control instructions, and verify in real time whether the gimbal attitude error after compensation converges to the preset threshold; S6: If the attitude error has not converged, return to S1 to update the error source identification result until the error meets the stability condition.

2. The error compensation control method for a self-stabilized pan / tilt platform under multi-branch redundant coordination according to claim 1, characterized in that: The S2 includes: S21: extracting the error component of the main error source branch chain, and performing orthogonal decomposition based on the driving coordinate system of the branch chain; S22: In the driving coordinate system, the decomposed tangential projection is defined as a direction error component, the normal projection is defined as an amplitude error component, and an error coordinate system of the main error source branch chain is established; S23: Obtain the error components of all secondary error branches, and analyze the posture transformation relationship between the error coordinate system of each secondary error branch and the main error source branch; S24: Based on the posture transformation relationship, the error components of each secondary error branch are mapped to the error coordinate system of the main error source branch through a homogeneous transformation matrix to generate a mapped error component with a unified reference.

3. The error compensation control method for a self-stabilized pan / tilt platform under multi-branch redundant coordination according to claim 2, characterized in that: The S3 includes: S31: performing time-frequency analysis on the direction error component and the amplitude error component respectively to extract time domain features, wherein the time domain features include a phase offset of the direction error component and an envelope fluctuation of the amplitude error component; S32: Calculating an angle compensation amount of the driving shaft according to the phase offset of the direction error component, and generating a direction compensation signal having a negative feedback relationship with the phase offset; S33: Calculating an amplitude correction amount of the driving force according to the envelope fluctuation amount of the amplitude error component, and generating an amplitude compensation signal that is in inverse proportion to the envelope fluctuation amount.

4. The error compensation control method for a self-stabilized pan / tilt platform under multi-branch redundant coordination according to claim 3, characterized in that: The S5 includes: S51: parsing the redundant branch chain coordinated control instruction, extracting the direction compensation and amplitude compensation of each branch chain, and generating a corresponding motor drive pulse signal; S52: Synchronously sending the motor drive pulse signal to the execution motors of all branches to drive each branch to synchronously execute the compensation action.

5. The error compensation control method for a self-stabilized pan / tilt platform under multi-branch redundant coordination according to claim 4, characterized in that: The S5 further includes: S53: The compensated gimbal attitude data is collected in real time through the attitude sensors of each branch chain, and the current gimbal attitude error is calculated; S54: Compare the current gimbal posture error with a preset threshold value to determine whether the posture error converges to a preset threshold range.

6. The error compensation control method for a self-stabilized pan / tilt platform under multi-branch redundant coordination according to claim 5, characterized in that: The S6 includes: S61: If the attitude error does not converge to a preset threshold range, the unconverged attitude error of the current gimbal and its corresponding branch motion state are recorded; S62: Based on the unconverged posture error, correct the identification parameters of the main error source branch and the secondary error branch, and update the disturbance propagation model; S63: Feedback the updated disturbance propagation model to S1 and restart the error compensation control loop.

Images