A calibration method and system for an integrated servo joint module

Abstract

The present invention relates to a calibration method and system for an integrated servo joint module, and belongs to the field of mechanical arm control technology. The method includes the following steps: sample data acquisition, data preprocessing, taking torque, current and position as inputs of a prediction model, and taking joint angle as output of the model to establish a prediction model; position parameter correction and parameter feedback correction. The present invention integrates high-precision position, current and torque sensors at key positions of the servo joint module to monitor the joint operation status in real time; constructs a dynamic mathematical model of the servo joint module, and realizes accurate compensation of the joint motor position error by predicting the angle of the joint module; and dynamically adjusts the joint control parameters according to the measured data through a fuzzy PID control algorithm and an iteratively optimized adaptive calibration algorithm to minimize the joint tracking error and ensure that the optimal control performance can be achieved under different load and speed conditions.

Description

Technical Field

[0001] The invention belongs to the technical field of robot arm control, and in particular relates to a calibration method and system for an integrated servo joint module. Background Art

[0002] In the current field of robotics and automation, the integrated servo joint module is a core component, and its performance directly affects the accuracy and stability of the entire system. However, the existing servo joint modules face many challenges during the calibration process, including but not limited to low calibration accuracy, complex and time-consuming process, and difficulty in achieving precise control under high dynamic response. Especially in multi-degree-of-freedom robotic arms or complex motion mechanisms, the coordination and accuracy requirements between the joints are extremely high, and traditional calibration methods often cannot meet these high-precision and high-efficiency requirements. Summary of the invention

[0003] In order to solve the above-mentioned problems existing in the prior art, the present invention provides a calibration method and system for an integrated servo joint module, which integrates high-precision position, current and torque sensors at key positions of the servo joint module to monitor the joint operation status in real time, and provide an accurate data basis for subsequent data analysis and correction; utilizes advanced machine learning algorithms and combines actual operation data to construct a dynamic mathematical model of the servo joint module, and achieves accurate compensation for the joint motor position error by predicting the joint module angle; uses a fuzzy PID control algorithm and an iteratively optimized adaptive calibration algorithm to dynamically adjust the joint control parameters according to the measured data to minimize the joint tracking error, and ensure that the optimal control performance can be achieved under different load and speed conditions.

[0004] The purpose of the present invention can be achieved through the following technical solutions:

[0005] A first aspect of the present disclosure provides a calibration method for an integrated servo joint module, comprising the following steps:

[0006] Sample data collection: collect the current, torque, position of the motor, angular velocity and acceleration of the serial robot arm, and calculate the joint angle;

[0007] Data preprocessing: normalize the torque, current, position data and joint angle data;

[0008] Establish a joint angle prediction model: Take torque, current and position as the input of the prediction model, and joint angle as the output of the model to establish a prediction model;

[0009] Position parameter correction: Based on the error calculation between the predicted joint angle and the target joint angle, the joint angle is compensated by adjusting the position of the motor;

[0010] Parameter feedback correction: Based on the collected motor operating parameters, predict the current joint angle position, set the error threshold, and when it exceeds the error threshold range, correct the rotation ratio and re-correct the motor position parameters.

[0011] Furthermore, the calculation of the joint angle specifically includes the following steps:

[0012] Calculate the rotation angle and rotation axis: According to the dynamic data of the inertial measurement unit, the principal component analysis algorithm is used to reduce the dimension to find the rotation axis and the rotation direction of the serial robot arm joint, and the rotation angle and rotation axis are calculated. The formula is as follows:

[0013] ;

[0014] ;

[0015] In the formula, is the rotation angle, is the rotation axis; and are the rotation directions of the main axis and the serial robot joints, cross is the vector cross product operation;

[0016] Calculate the angular velocity and acceleration after rotation: Convert the rotation angle and rotation axis into a rotation quaternion q , use the Rodrigues formula to obtain the rotation matrix from the quaternion R , angular velocity and acceleration To rotate, the formula is as follows:

[0017] ;

[0018] , ;

[0019] In the formula, and are the angular velocity and acceleration after rotation.

[0020] Furthermore, the calculation of the joint angle further includes the steps of:

[0021] Joint angle calculation: The output values ​​of the gyroscope and accelerometer are defined as the motion information of the carrier coordinate system. When the carrier moves, the angular velocity is established. The time updated quaternion q Differential Equations:

[0022] ;

[0023] In the formula, is the derivative of the quaternion, representing the micro-increment of the function, and is numerically solved using the fourth-order Runge-Quinta method;

[0024] Complementary filtering is used to fuse data to estimate the angle of the Z-axis rotation of the robot arm. The accelerometer measurement value is used as the observation value. The extended Kalman filter algorithm is used to fuse the observation data and the estimated data to perform closed-loop management of the measurement error and limit the error to a certain range.

[0025] The quaternion obtained by the above filtering is converted into the Euler angle corresponding to the X / Y / Z axis , , , the formula is as follows:

[0026] ;

[0027] In the formula, , , , is a quaternion.

[0028] Furthermore, the establishment of the joint angle prediction model comprises the following steps:

[0029] The collected data is preprocessed to obtain a data set, which is divided into test samples and training samples at a ratio of 1:4 for model accuracy detection and model training;

[0030] Select the kernel function and build the SVR prediction model;

[0031] Using radial basis function as kernel function, we can get the approximate function f ( x ):

[0032] ;

[0033] In the formula, α、α* is the Lagrange multiplier; K ( xiii , x ) is the kernel function;

[0034] The wolf pack algorithm is used to optimize the penalty factor C and the kernel function parameter g in the kernel function. The update formula of the objective function is:

[0035] ;

[0036] In the formula, , Respectively m A wolf is iterating k +1, k The position at the time; λ is a random number in [-1, 1]; For the intelligent wolf d The attack step length in dimensional space; For the k In the iteration, the prey d The position of the dimensional space;

[0037] Set the number of particles to 40 and the maximum number of walks T max is 400 times, the proportional factor α= 4. Update the scale factor β =6, optimize the penalty factor C and kernel function parameter g of SVR;

[0038] The optimized parameters are written into the SVR neural network, the model is trained and the test samples are input to verify the model accuracy, and the joint angle prediction output is performed through the trained model.

[0039] Furthermore, the SVR prediction model is constructed, wherein the prediction form of SVR is:

[0040] ;

[0041] In the formula, , is the slack variable; ω is the separating hyperplane weight vector; b is the deviation amount; is the input feature vector; is the target value; is the separating hyperplane; C is the penalty factor.

[0042] Furthermore, the position parameter correction comprises the following steps:

[0043] The relationship model between joint angle and motor position is established as follows:

[0044] T J =T m / R a ;

[0045] Among them, T J is the joint angle, T m is the motor position, R a is the rotation ratio;

[0046] The error is calculated based on the predicted joint angle and the target joint angle, and the joint angle error is converted into the motor position error through the relationship model between the joint angle and the motor position;

[0047] Establish the dynamic model of the motor, combine fuzzy control and PID control algorithms, design the fuzzy PID control algorithm, and control the drive motor.

[0048] Furthermore, controlling the drive motor includes the following steps:

[0049] Establish a dynamic model of click: Use torque motor to control the joint, and establish the mathematical model of the motor based on the working principle of DC torque motor and Kirchhoff's law:

[0050] ;

[0051] In the formula, Ua is the armature voltage, Ω is the output angular velocity, Ra is the armature resistance, La is the armature inductance, J is the moment of inertia, Ke is the torque coefficient, Ka is the back electromotive force, s is the Laplace variable;

[0052] Fuzzy PID control algorithm is used to control the torque motor.

[0053] Furthermore, the use of a fuzzy PID control algorithm to control the torque motor comprises the following steps:

[0054] Fuzzy quantification: Establish the mapping relationship between the basic domain and the fuzzy subset domain, assuming that the basic domain is [ a , c ], the fuzzy subset domain is [- n , n ], then its mapping relationship can be expressed as:

[0055] ;

[0056] Determine the membership function: Use the Gaussian membership function with good results as the membership function of fuzzy PID control for calculation:

[0057] ;

[0058] In the formula, are the center of the function and the width of the function respectively;

[0059] Quantization factor and scaling factor settings: Quantization factor and scaling factor are factors that convert input variables from basic domain to fuzzy domain and from fuzzy domain to actual domain. The scaling factor can be set empirically, while the quantization factor can be determined by the following formula:

[0060] ;

[0061] In the formula, For deviation e The basic domain of [-a , a ], For deviation e The corresponding fuzzy subset domain [- n , n ];

[0062] Fuzzy rule construction: reasonable settings Kp , Ki , Kd Parameters are helpful for the system to reduce steady-state error and adjustment time, and reduce overshoot; by analyzing Kp , Ki , Kd Three parameter characteristics to construct fuzzy rules:

[0063] Position Deviation| e | When it is large, in order to improve the system's fast tracking capability, a larger Kp ; At the same time, in order to avoid the deviation from increasing instantly, a smaller Kd ; In order to avoid overshoot, a smaller Ki ;

[0064] Position Deviation| e |When the size is medium, in order to reduce the system overshoot, a smaller Kp ; In order to avoid large oscillations in the system, a smaller Kd ; In order to ensure the corresponding response speed, Ki The size should be moderate;

[0065] Position Deviation| e | is small, in order to improve the control accuracy, a larger Kp ; In order to reduce the static error and shorten the adjustment time, Ki .

[0066] Fuzzy reasoning and defuzzification: Fuzzy reasoning is performed using the commonly used minimum operation algorithm, which is mathematically expressed as:

[0067] ;

[0068] The centroid method is used for defuzzification, and its mathematical expression is:

[0069] ;

[0070] In the formula, is the membership function of the output;

[0071] Dynamic compensation position deviation: the output will be K p , K i ,K d Input to the PID controller to complete online adjustment of PID parameters, and the PID controller dynamically compensates for the motor position error.

[0072] A second aspect of the present disclosure provides a calibration system for an integrated servo joint module, which is applied to a calibration method for an integrated servo joint module as described above, and includes a data acquisition module, an angle prediction module, a position compensation module, and a parameter correction module;

[0073] The data acquisition module is used to collect the current, torque and position data of the motor operation, and collect the angular velocity and acceleration of the serial manipulator by installing an inertial measurement unit;

[0074] The angle prediction module is used to construct a joint angle prediction model and use the collected torque, current and position data of the motor operation to predict the angle of the joint module.

[0075] As a preferred technical solution of the present invention, the position compensation module is used to convert the error of the joint angle into the motor position error, and then compensate the motor position error through the fuzzy PID control algorithm;

[0076] The parameter correction module is used to make a judgment by using the error size after the joint angle adjustment, and re-correct the rotation ratio when it exceeds a set threshold range.

[0077] The beneficial effects of the present invention are:

[0078] The present invention first collects the operating state parameters of the motor under load conditions, and then calculates the joint angle by installing an inertial measurement unit, thereby providing real-time data for the dynamic calibration of the joint module during operation; constructs a data set to train and test the prediction model, and continuously optimizes the control algorithm and compensation mechanism; obtains the state of the joint angle when the motor is running, and converts it into the motor error of the joint module for fuzzy PID control algorithm compensation, thereby calibrating the joint module of the robot arm, avoiding the system from being disturbed by the outside world, and still maintaining a good control effect, thereby improving the robustness and stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0080] Picture 1 A schematic diagram of the steps of a calibration method for an integrated servo joint module provided in Embodiment 1 of the present invention;

[0081] Picture 2 A schematic diagram of the steps of establishing a joint angle prediction model provided in the first embodiment of the present invention;

[0082] Picture 3 A schematic diagram of the steps of position parameter correction provided in the first embodiment of the present invention;

[0083] Picture 4 A schematic structural diagram of a calibration system for an integrated servo joint module provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0084] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0085] Embodiment 1

[0086] This embodiment provides a calibration method for an integrated servo joint module, such as Picture 1 As shown, the following steps are included:

[0087] S1. Sample data collection: collect the current, torque, position of the motor, angular velocity and acceleration of the serial robot arm, and calculate the joint angle;

[0088] It can be understood that the torque is collected by the torque transmitter and then uploaded to the host computer through the data acquisition card. The current and position error data can be viewed and saved through the motor's supporting HMServo software.

[0089] It should be noted that the angular velocity and acceleration of the serial robot arm are measured by setting up an inertial measurement unit, and the spatial posture of adjacent links is measured using a gyroscope and an accelerometer to calculate the joint angle.

[0090] Calculating joint angles involves the following steps:

[0091] S11. Calculate the rotation angle and rotation axis: According to the dynamic data of the inertial measurement unit, the principal component analysis (PCA) algorithm is used to reduce the dimension to find the rotation axis and the rotation direction of the serial robot joint, and the rotation angle and rotation axis are calculated. The formula is as follows:

[0092] ;

[0093] ;

[0094] In the formula, is the rotation angle, is the rotation axis; and are the rotation directions of the main axis and the serial robot joints, cross is the vector cross product operation;

[0095] S12. Calculate the angular velocity and acceleration after rotation: convert the rotation angle and rotation axis into a rotation quaternion q , use the Rodrigues formula to obtain the rotation matrix from the quaternion R , angular velocity and acceleration To rotate, the formula is as follows:

[0096] ;

[0097] , ;

[0098] In the formula, and are the angular velocity and acceleration after rotation.

[0099] It can be understood that by mapping the coordinate system of the inertial measurement unit to the joint coordinate system of the serial robot, finding the rotation axis through the PCA algorithm, and then performing rotation compensation on the data, the initial installation posture is aligned with the joint axis posture of the serial robot, thereby solving the error problem between the inertial measurement unit and the robot.

[0100] S13, joint angle calculation: define the output values ​​of the gyroscope and accelerometer as the motion information of the carrier coordinate system, and establish the angular velocity based on the motion of the carrier. The time updated quaternion q Differential Equations:

[0101] ;

[0102] In the formula, is the derivative of the quaternion, representing the micro-increment of the function, and is numerically solved using the fourth-order Runge-Quinta method;

[0103] This embodiment is based on the frequency response advantage of the sensor and adopts complementary filtering to fuse data to estimate the angle of the Z-axis rotation of the robot arm. On the other hand, because the X-axis rotation of the robot arm is affected by gravity, the accelerometer measurement value is used as the observation value. Through the extended Kalman filter algorithm, the observation data and the estimated data are fused, and the measurement error is closed-loop managed, which limits the error to a certain range and solves the gyroscope drift error problem in the system process.

[0104] The quaternion obtained by the above filtering is converted into the Euler angle corresponding to the X / Y / Z axis , , , the formula is as follows:

[0105] ;

[0106] In the formula, , , , is a quaternion.

[0107] It can be understood that after the above steps, the initial coordinate system is first calibrated by the principal component analysis method, and then the joint angle is estimated by the fusion algorithm.

[0108] S2. Data preprocessing: Since the collected data ranges are different, it may increase the subsequent neural network training time and reduce the prediction accuracy, so the data is normalized. That is, the torque, current, position data and joint angle range are normalized to [-1, 1].

[0109] S3. Establish a joint angle prediction model: Take torque, current and position as the input of the prediction model, and joint angle as the output of the model to establish a prediction model. Picture 2 As shown, the following steps are included:

[0110] S31. Preprocess the collected data to obtain a data set, and divide the data set into test samples and training samples at a ratio of 1:4 for model accuracy detection and model training.

[0111] S32, select a kernel function and construct an SVR prediction model, where the prediction form of SVR is:

[0112] ;

[0113] In the formula, , is the slack variable; ω is the separating hyperplane weight vector; b is the deviation amount; is the input feature vector; is the target value; is the separating hyperplane; C is the penalty factor.

[0114] Using radial basis function (RBF) as the kernel function, we can get the approximate function f ( x ):

[0115] ;

[0116] In the formula, α、α* is the Lagrange multiplier; K ( xiii , x ) is the kernel function.

[0117] SVR is an important application branch of support vector machine (SVM), which can handle small sample data, nonlinearity and time series problems well and has strong generalization ability. Different from traditional regression methods, SVR introduces insensitive loss factor ε , when the absolute value of the difference between the predicted value and the target value is less than ε The prediction is considered correct and no loss is calculated.

[0118] S33, using the wolf pack algorithm to optimize the penalty factor C and the kernel function parameter g in the kernel function, the update formula of the objective function is:

[0119] ;

[0120] In the formula, , Respectively m A wolf is iterating k +1, k The position at the time; λ is a random number in [-1, 1]; For the intelligent wolf d The attack step length in dimensional space; For the k In the iteration, the prey d dimensional space position.

[0121] Set the number of particles (intelligent wolves) to 40 and the maximum number of wandering times T max is 400 times, the proportional factor α= 4. Update the scale factor β =6, and optimize the penalty factor C and kernel function parameter g of SVR.

[0122] S34. Write the optimized parameters into the SVR neural network, train the model and input the test samples to verify the model accuracy, and use the trained model to predict and output the joint angle.

[0123] S4, position parameter correction: According to the error calculation between the predicted joint angle and the target joint angle, the joint angle compensation is completed by adjusting the position of the motor, such as Picture 3 As shown, the following steps are included:

[0124] S41. Establish the relationship model between joint angle and motor position as shown below:

[0125] T J =T m / R a ;

[0126] Among them, T Jis the joint angle, T m is the motor position, R a is the rotation ratio.

[0127] It should be noted that the rotation ratio in this embodiment is obtained by actual measurement of a given robot arm joint module, including friction loss, load transmission and transmission efficiency. The specific measurement process uses a magnetic powder brake to connect with the motor through a coupling, simulates a variable load condition by adjusting the current of the magnetic powder brake, and obtains the joint angle in the above manner by installing an inertial measurement unit, thereby calculating the rotation ratio. In the later error adjustment process, if the joint angle error obtained by the current rotation ratio conversion is large, the rotation ratio is corrected.

[0128] S42, calculating an error according to the predicted joint angle and the target joint angle, and converting the joint angle error into a motor position error through a relationship model between the joint angle and the motor position;

[0129] S43, establishing a motor dynamic model, combining fuzzy control and PID control algorithms, designing a fuzzy PID control algorithm, and controlling the drive motor, including the following steps:

[0130] S431. Establish a dynamic click model: Use a torque motor to control the joints. According to the working principle of the DC torque motor and Kirchhoff's law, establish the mathematical model of the motor:

[0131] ;

[0132] In the formula, Ua is the armature voltage, Ω is the output angular velocity, Ra is the armature resistance, La is the armature inductance, J is the moment of inertia, Ke is the torque coefficient, Ka is the back electromotive force, s is the Laplace variable.

[0133] S432, using a fuzzy PID control algorithm to control the torque motor, including the following steps:

[0134] S4321, fuzzy quantification: Establish the mapping relationship between the basic domain and the fuzzy subset domain, assuming that the basic domain is [ a , c ], the fuzzy subset domain is [- n , n ], then its mapping relationship can be expressed as:

[0135] ;

[0136] S4322, determine the membership function: This embodiment uses the Gaussian membership function with good effect as the membership function of the fuzzy PID control for calculation:

[0137] ;

[0138] In the formula, are the center and width of the function respectively.

[0139] S4323, Quantization factor and proportional factor settings: Quantization factor and proportional factor are factors that convert input variables from basic domain to fuzzy domain and from fuzzy domain to actual domain. Setting reasonable quantization factor and proportional factor is conducive to ensuring system stability. Usually, the proportional factor can be set through experience, and the quantization factor can be determined by the following formula:

[0140] ;

[0141] In the formula, For deviation e The basic domain of [- a , a ], For deviation e The corresponding fuzzy subset domain [- n , n ] in the .

[0142] S4324, Fuzzy rule construction: reasonable setting Kp , Ki , Kd Parameters are helpful for the system to reduce steady-state error and adjustment time, and reduce overshoot. Kp , Ki , Kd Three parameter characteristics to construct fuzzy rules:

[0143] Position Deviation| e | When it is large, in order to improve the system's fast tracking capability, a larger Kp ; At the same time, in order to avoid the deviation from increasing instantly, a smaller Kd ; In order to avoid overshoot, a smaller Ki .

[0144] Position Deviation| e |When the size is medium, in order to reduce the system overshoot, a smaller Kp ; In order to avoid large oscillations in the system, a smaller Kd ; In order to ensure the corresponding response speed, Ki The size should be moderate.

[0145] Position Deviation| e| is small, in order to improve the control accuracy, a larger Kp ; In order to reduce the static error and shorten the adjustment time, Ki .

[0146] S4325, fuzzy reasoning and defuzzification: This embodiment adopts the commonly used minimum operation algorithm to perform fuzzy reasoning, and its mathematical expression is:

[0147] ;

[0148] The centroid method is used for defuzzification, and its mathematical expression is:

[0149] ;

[0150] In the formula, is the membership function of the output.

[0151] S4326, dynamic compensation position deviation: the output K p , K i , K d Input to the PID controller to complete online adjustment of PID parameters, and the PID controller dynamically compensates for the motor position error.

[0152] S5, parameter feedback correction: predict the current joint angle position based on the collected motor operating parameters, set the error threshold, and correct the rotation ratio when it exceeds the error threshold range, and re-correct the motor position parameters.

[0153] It is understandable that the motor operating parameters here include the current, torque and position of the motor.

[0154] In this embodiment, the position error of the motor is corrected by overall calculation of the joint angle of the robot arm, thereby reducing the complexity of analyzing influencing factors such as joint nonlinear characteristics, friction and backlash. Correction is performed indirectly through the control strategy, which simplifies the system design and debugging process and improves the motion accuracy, stability and adaptability of the joint module.

[0155] Embodiment 2

[0156] This embodiment also provides a calibration system for an integrated servo joint module, such as Picture 4 As shown, it includes a data acquisition module, an angle prediction module, a position compensation module and a parameter correction module.

[0157] The data acquisition module is used to collect the current, torque and position data of the motor operation, and collect the angular velocity and acceleration of the series robot arm by installing an inertial measurement unit.

[0158] It should be noted that the torque is collected through the torque transmitter and then uploaded to the host computer through the data acquisition card. The current and position error data can be viewed and saved through the motor-matching HMServo software; the inertial measurement unit is set on the left side of the robotic arm in turn, and the X-axis of the inertial measurement unit corresponds to the opposite direction of gravity.

[0159] The angle prediction module is used to construct a joint angle prediction model and use the collected torque, current and position data of the motor operation to predict the angle of the joint module.

[0160] The position compensation module is used to convert the joint angle error into the motor position error, and then compensate the motor position error through the fuzzy PID control algorithm.

[0161] The parameter correction module is used to make a judgment by using the error size after the joint angle adjustment, and re-correct the rotation ratio when it exceeds a set threshold range.

[0162] It should be noted that the rotation ratio correction corrects the rotation ratio by adjusting the ratio of the motor position converted by the error before adjustment to the motor position converted by the error after adjustment, so as to reduce the conversion error caused by mechanical wear, inertia and other factors.

[0163] The present invention first collects the operating state parameters of the motor under load conditions, and then calculates the joint angle by installing an inertial measurement unit, thereby providing real-time data for the dynamic calibration of the joint module during operation; constructs a data set to train and test the prediction model, and continuously optimizes the control algorithm and compensation mechanism; obtains the state of the joint angle when the motor is running, and converts it into the motor error of the joint module for fuzzy PID control algorithm compensation, thereby calibrating the joint module of the robot arm, avoiding the system from being disturbed by the outside world, and still maintaining a good control effect, thereby improving the robustness and stability of the system.

[0164] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A calibration method for an integrated servo joint module, characterized in that: The following steps are involved: Sample data collection: collect the current, torque, position of the motor, angular velocity and acceleration of the serial robot arm, and calculate the joint angle; Data preprocessing: normalize the torque, current, position data and joint angle data; Establish a joint angle prediction model: Take torque, current and position as the input of the prediction model, and joint angle as the output of the model to establish a prediction model; Position parameter correction: Based on the error calculation between the predicted joint angle and the target joint angle, the joint angle is compensated by adjusting the position of the motor; Parameter feedback correction: Based on the collected motor operating parameters, predict the current joint angle position, set the error threshold, and when it exceeds the error threshold range, correct the rotation ratio and re-correct the motor position parameters.

2. The calibration method of an integrated servo joint module according to claim 1, characterized in that: The calculation of the joint angle specifically includes the following steps: Calculate the rotation angle and rotation axis: According to the dynamic data of the inertial measurement unit, the principal component analysis algorithm is used to reduce the dimension to find the rotation axis and the rotation direction of the serial robot arm joint, and the rotation angle and rotation axis are calculated. The formula is as follows: ； ； In the formula, is the rotation angle, is the rotation axis; and are the rotation directions of the main axis and the serial robot joints, cross is the vector cross product operation; Calculate the angular velocity and acceleration after rotation: Convert the rotation angle and rotation axis into a rotation quaternion q , use the Rodrigues formula to obtain the rotation matrix from the quaternion R , angular velocity and acceleration To rotate, the formula is as follows: ； ， ； In the formula, and are the angular velocity and acceleration after rotation.

3. The calibration method of an integrated servo joint module according to claim 2, characterized in that: The calculation of the joint angle also includes the steps of: Joint angle calculation: The output values ​​of the gyroscope and accelerometer are defined as the motion information of the carrier coordinate system. When the carrier moves, the angular velocity is established. The time updated quaternion q Differential Equations: ； In the formula, is the derivative of the quaternion, representing the micro-increment of the function, and is numerically solved using the fourth-order Runge-Quinta method; Complementary filtering is used to fuse data to estimate the Z-axis rotation angle of the robot arm. The accelerometer measurement value is used as the observation value. The extended Kalman filter algorithm is used to fuse the observation data and the estimated data to perform closed-loop management of the measurement error. The quaternion obtained by filtering is converted into the Euler angle corresponding to the X / Y / Z axis , , , the formula is as follows: ； In the formula, , , , is a quaternion.

4. The calibration method of an integrated servo joint module according to claim 1, characterized in that: The method of establishing a joint angle prediction model comprises the following steps: The collected data is preprocessed to obtain a data set, which is divided into test samples and training samples at a ratio of 1:4 for model accuracy detection and model training; Select the kernel function and build the SVR prediction model; Using radial basis function as kernel function, we can get the approximate function f ( x ): ； In the formula, α、α* is the Lagrange multiplier; K ( xi , x ) is the kernel function, b is the deviation amount; The wolf pack algorithm is used to optimize the penalty factor C and the kernel function parameter g in the kernel function. The update formula of the objective function is: ； In the formula, , Respectively m A wolf is iterating k +1, k The position at the time; λ is a random number in [-1, 1]; For the intelligent wolf d The attack step length in dimensional space; For the k In the iteration, the prey d The position of the dimensional space; Set the number of particles to 40 and the maximum number of walks T max is 400 times, the proportional factor α= 4. Update the scale factor β =6, optimize the penalty factor C and kernel function parameter g of SVR; The optimized parameters are written into the SVR neural network, the model is trained and the test samples are input to verify the model accuracy, and the joint angle prediction output is performed through the trained model.

5. The calibration method of an integrated servo joint module according to claim 4, characterized in that: The SVR prediction model is constructed, wherein the prediction form of SVR is: ； In the formula, , is the slack variable; ω is the separating hyperplane weight vector; b is the deviation amount; is the input feature vector; is the target value; is the separating hyperplane; C is the penalty factor.

6. The calibration method of an integrated servo joint module according to claim 1, characterized in that: The position parameter correction comprises the following steps: The relationship model between joint angle and motor position is established as follows: T J =T m / R a ; Among them, T J is the joint angle, T m is the motor position, R a is the rotation ratio; The error is calculated based on the predicted joint angle and the target joint angle, and the joint angle error is converted into the motor position error through the relationship model between the joint angle and the motor position; Establish the dynamic model of the motor, combine fuzzy control and PID control algorithms, design the fuzzy PID control algorithm, and control the drive motor.

7. The calibration method of an integrated servo joint module according to claim 6, characterized in that: Controlling the drive motor includes the following steps: Establish a dynamic model of click: Use torque motor to control the joint, and establish the mathematical model of the motor based on the working principle of DC torque motor and Kirchhoff's law: ； In the formula, Ua is the armature voltage, Ω is the output angular velocity, Ra is the armature resistance, La is the armature inductance, J is the moment of inertia, Ke is the torque coefficient, Ka is the back electromotive force, s is the Laplace variable; Fuzzy PID control algorithm is used to control the torque motor.

8. The calibration method of an integrated servo joint module according to claim 6, characterized in that: The method of controlling the torque motor by using a fuzzy PID control algorithm comprises the following steps: Fuzzy quantification: Establish the mapping relationship between the basic domain and the fuzzy subset domain, assuming that the basic domain is [ a , c ], the fuzzy subset domain is [- n , n ], then its mapping relationship can be expressed as: ； Determine the membership function: Use the Gaussian membership function as the membership function of the fuzzy PID control for calculation: ； In the formula, are the center of the function and the width of the function respectively; Quantization factor and proportional factor setting: Quantization factor and proportional factor are factors that convert input variables from basic domain to fuzzy domain and from fuzzy domain to actual domain. The proportional factor is set through experience, and the quantization factor is determined by the following formula: ； In the formula, For deviation e The basic domain of [- a , a ], For deviation e The corresponding fuzzy subset domain [- n , n ]; Fuzzy rule construction: reasonable settings Kp , Ki , Kd Parameters are helpful for the system to reduce steady-state error and adjustment time, and reduce overshoot; by analyzing Kp , Ki , Kd Three parameter characteristics, construct fuzzy rules; Fuzzy reasoning and defuzzification: Fuzzy reasoning is performed using the commonly used minimum operation algorithm, which is mathematically expressed as: ； The centroid method is used for defuzzification, and its mathematical expression is: ； In the formula, is the membership function of the output; Dynamic compensation position deviation: the output will be K p , K i , K d Input to the PID controller to complete online adjustment of PID parameters, and the PID controller dynamically compensates for the motor position error.

9. A calibration system for an integrated servo joint module, applied to a calibration method for an integrated servo joint module as claimed in any one of claims 1 to 8, characterized in that: It includes a data acquisition module, an angle prediction module, a position compensation module and a parameter correction module; The data acquisition module is used to collect the current, torque and position data of the motor operation, and collect the angular velocity and acceleration of the serial manipulator by installing an inertial measurement unit; The angle prediction module is used to construct a joint angle prediction model and use the collected torque, current and position data of the motor operation to predict the angle of the joint module.

10. The calibration system of the integrated servo joint module according to claim 9, characterized in that: The position compensation module is used to convert the joint angle error into the motor position error, and then compensate the motor position error through the fuzzy PID control algorithm; The parameter correction module is used to make a judgment by using the error size after the joint angle adjustment, and re-correct the rotation ratio when it exceeds a set threshold range.

Images