A Robot Load Dynamics Parameter Identification Method Based on Current Dynamics

Abstract

This invention relates to a method for identifying robot load dynamic parameters based on current dynamics, within the field of robot dynamic parameter calibration. It addresses the problem that existing methods for identifying load dynamic parameters suffer from errors in obtaining joint torque constants, leading to increased joint torque prediction errors and consequently, accumulated errors in load dynamic parameter identification. This invention constructs a novel model for load dynamic parameter identification and proposes a weighted iterative identification method for estimating current-level dynamic parameters, which helps obtain weighted least-squares solutions and eliminate outliers. Furthermore, a zero-velocity continuous nonlinear joint combination friction model is introduced during the current-level dynamic parameter identification process to improve overall identification accuracy. This invention is primarily used for identifying robot current-level dynamic parameters when joint torques are not readily available or are difficult to obtain.

Description

Technical Field

[0001] This invention relates to the field of robot dynamics parameter calibration. Background Technology

[0002] As the load capacity of robots increases, the proportion of current or torque generated by the load in the actuators also increases, thereby enhancing their importance in robotics. For example, accurate load dynamics parameters help to accurately predict joint currents or torques, improve robot motion control performance, and enhance human-robot safe interaction capabilities.

[0003] When a load is fixedly connected to the robot's end effector, there are two main identification methods. One is to use force or torque sensors mounted on the robot's joints to identify the load's dynamic characteristics. However, the high cost, large size, and weight make this method impractical in some situations. The load is fixedly connected to the end effector, which can be considered part of the robot's end effector linkage. Therefore, another method is to identify the load's dynamic parameters by identifying the robot's own dynamic parameters. This method is more widely used, specifically as follows:

[0004] Swevers et al. presented a classic load identification model in their paper. For loads connected to robot joint moments and links, this method distinguishes the load from the end effector, thus reducing identification errors. Khalil et al. proposed two load identification methods. One method involves having the robot track the same reference trajectory twice: once without load and then with load. The difference in joint moments between the two methods is attributed to the load. The other method uses weighted least squares to simultaneously identify the robot's dynamic parameters and load parameters, but does not consider frictional variations. Gaz et al. proposed a static identification method based on the linear correlation between changes in robot static parameters and changes in load static parameters. Bahloul et al. attempted to identify the load's dynamic parameters using nonlinear optimization techniques, applying basic physical constraints such as ensuring the load mass is positive. Xu et al. proposed an identification method based on dual-weighting techniques to improve the accuracy of payload dynamic estimation.

[0005] The algorithms described above provide a sound theoretical foundation and valuable reference for identifying load dynamic parameters. However, these methods all rely on joint torque signals to identify load dynamic parameters. In reality, most industrial robots are not equipped with joint torque sensors; joint torque can only be predicted by multiplying the actuator current by the joint torque constant. However, the joint torque constants provided by manufacturers typically have an uncertainty of about 10%. Even if some precise identification methods are introduced to obtain these constants, errors still exist. Therefore, an increase in joint torque prediction error will lead to an accumulation of errors in load dynamic parameter identification. Summary of the Invention

[0006] The purpose of this invention is to address the problem that existing load dynamic parameter identification methods have errors in obtaining joint torque constants, which increases the error in joint torque prediction and leads to the accumulation of load dynamic parameter identification errors. This invention provides a robot load dynamic parameter identification method based on electrodynamics.

[0007] A method for identifying robot load dynamic parameters based on current dynamics, the method comprising:

[0008] S1. Constructing a joint assembly friction model for the robot. ,right Optimization was performed to determine the robot's trajectory execution under load at each sampling time. Corresponding combination coefficients , and The value of is determined based on the sampling time. , and Construct the corresponding linear regression matrix ;in, For the joint velocity exponential vector, Stribeck's velocity vector and The arctangent coefficient vector;

[0009] When the robot executes the excitation trajectory under load, it constructs a linear regression matrix at each sampling moment based on a set of joint coordinates collected at each sampling moment. Each set of joint coordinates includes the angular positions of all joints. ,speed and acceleration ;

[0010] S2, based on each sampling time... and Constructing a matrix ;

[0011] Collect all joint current data corresponding to both states at each sampling time when the robot executes the excitation trajectory under loaded and unloaded conditions, and merge the data from both states to form a set of joint current data corresponding to the current sampling time. ;

[0012] S3, N sampling times Stack them in chronological order to form a matrix. N sampling times Stack them in chronological order to form a matrix. ;in, This represents the total number of robot joints. for The number of elements in the middle, For dimension The real space, For dimension The real number space;

[0013] S4, according to and Solving the constructed dynamic parameter identification model OLS solution ,according to Determine the covariance matrix The initial value; where, For current-level load dynamic parameters, The parameter represents the frictional change in the current level caused by the load. for OLS solution, for OLS solution;

[0014] S5. Initialize the joint data weighting vector Weighted vector of stacked joint data ; , , for A column vector of rows, where all elements of the column vector are 1; for The row vector of the column, where all elements of the row vector are 1;

[0015] S6. Using the covariance matrix To each and By weighting, the normalized motor current vector is obtained. and normalized linear regression matrix , , S7, according to and Calculate the matrix ,according to and Calculate the matrix ;

[0016] S8, according to and ,calculate ;

[0017] S9, according to , and Calculate the standardized current residual vector ;

[0018] S10, according to renew According to the threshold and renew Through the updated renew ;

[0019] S11, Judgment If convergence has occurred, output the value from step S8. If the result is negative, return to step S6.

[0020] Preferably, in step S1, a joint combination friction model of the robot is constructed. The implementation method is as follows:

[0021] First, design the first Joint friction model ,and

[0022] , ;

[0023] in, , , and Let represent the static friction coefficient, Coulomb friction coefficient, viscous friction coefficient, and friction offset coefficient of the j-th joint, respectively. Let the velocity of the j-th joint be... It is the velocity index of the j-th joint. It is the arctangent coefficient of the j-th joint. It is the Stribeck velocity of the j-th joint. It is a natural constant;

[0024] Secondly, based on all joint friction models, a joint combination friction model is constructed. ,in, .

[0025] Preferably, in step S1, when determining each sampling time... Corresponding combination coefficients , and The method for determining the value is as follows:

[0026] S1-1. Acquiring a set of joint current data at the current sampling moment during the process of the robot executing the excitation trajectory under load. And a set of joint coordinates corresponding to the current sampling time; among which, each set of joint current data Includes all joint current data under load at the current sampling time.

[0027] S1-2. Based on a set of joint coordinates corresponding to the current sampling time, construct the linear regression matrix for that sampling time. ;

[0028] S1-3, according to , and Estimate the triboelectric current at the current sampling time. ;

[0029] For the reason The vector of current inertia parameters formed by the mass of each link, the first moment of each link, the Coriolis force of each link, the centrifugal force of each link, the gravity of each link, and the rotational inertia of the rotor of each joint motor in the robot.

[0030] S1-4, to Apply constraints and optimize the solution. , and Specifically:

[0031] ;

[0032] in, Let j be the set of optimization variables for the j-th joint. ;

[0033] , It is the velocity index of the j-th joint;

[0034] , It is the arctangent coefficient of the j-th joint;

[0035] , It is the Stribeck velocity of the j-th joint;

[0036] It is a diagonal matrix with constant coefficients for joint moments, and , This is the vector formed by all the collected joint torques.

[0037] Preferably, the expression for the dynamic parameter identification model constructed in step S4 is: .

[0038] Preferably, in step S4, the solution... OLS solution The implementation method is as follows: .

[0039] Preferably, in step S4, according to Determine the covariance matrix The initial value is implemented as follows:

[0040] according to Determine the initial value of the current residual vector. ,and , For dimension The real number space;

[0041] according to Obtain the covariance matrix initial value .

[0042] Preferably, in step S6, , .

[0043] Preferably, in step S8, .

[0044] Preferably, in step S9, , .

[0045] Preferably, in step S10, according to renew The implementation method is as follows:

[0046] ;

[0047] According to the threshold and renew The implementation method is as follows:

[0048] ;

[0049] in, Let be a function of a vector of size Nn×1, and let Exceeding the threshold The element will be set to 0 if it is not set to 0, otherwise it will be set to 1;

[0050] Through the updated renew The implementation method is as follows:

[0051] .

[0052] Advantages of this invention:

[0053] This invention proposes a method for identifying robot load dynamic parameters based on current dynamics. This invention utilizes current instead of joint torque to identify load dynamic parameters. The method of this invention is superior to comparative methods in predicting effective load current dynamic parameters. It exhibits a small root mean square error (RMSE) for both concentric and eccentric loads.

[0054] On the one hand, the method proposed in this invention is derived from electrodynamics, thus eliminating the need for joint torque constants and avoiding joint torque estimation errors caused by inaccurate identification of joint torque constants, thereby reducing cumulative identification errors. Therefore, unlike methods based on torque dynamics, the load identification method based on current level dynamics proposed in this invention exhibits superior accuracy in load dynamic estimation.

[0055] On the other hand, the proposed iterative weighted estimation helps mitigate the negative impact of outliers in the measurement results, which also improves the identification accuracy of the effective load current dynamics. Furthermore, using a continuous nonlinear joint combination friction model at the velocity reversal point effectively avoids the problem of abrupt frictional force changes during joint velocity reversal, contributing to improved overall identification accuracy of the robot's current dynamics. Attached Figure Description

[0056] Figure 1 This is a flowchart of the robot load dynamics parameter identification method based on current dynamics described in this invention;

[0057] Figure 2 This is a schematic diagram of the end effector loading of the UR 10 robot; where, Figure 2 (a) is a schematic diagram of the robot end effector when no load is applied. Figure 2 (b) is a schematic diagram of the structure when a concentric load is applied to the end effector of the robot. Figure 2 (c) is a schematic diagram of the structure when an eccentric load is applied to the end effector of the robot;

[0058] Figure 3 This is a reconstructed diagram of the current dynamics of a concentric load; where, Figure 3 (a) to Figure 3 (f) are comparison charts of the predicted and measured values ​​of the joint currents from the 1st to the 6th joints, respectively; the joint current is the current driving the joint motor, which includes friction current (i.e., the current that overcomes friction), link inertia current (the current that overcomes the inertia of the link), joint inertia current (the current that overcomes the joint inertia when no load is applied and friction is not considered), and load current (the current that overcomes the resistance of the load on the joint).

[0059] Figure 4 This is a schematic diagram of the standardized current residual distribution of each joint obtained using the method of this invention when δ=2.5;

[0060] Figure 5 This is a schematic diagram of the standardized current residual distribution of each joint obtained using the method of this invention when δ=3. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0063] Specific Implementation Method 1: Combination Figure 1 This embodiment describes a robot load dynamics parameter identification method based on current dynamics, which includes:

[0064] S1. Constructing a joint assembly friction model for the robot. ,right Optimization was performed to determine the robot's trajectory execution under load at each sampling time. Corresponding combination coefficients , and The value of is determined based on the sampling time. , and Construct the corresponding linear regression matrix ;in, For the joint velocity exponential vector, Stribeck's velocity vector and The arctangent coefficient vector;

[0065] When the robot executes the excitation trajectory under load, it constructs a linear regression matrix at each sampling moment based on a set of joint coordinates collected at each sampling moment. Each set of joint coordinates includes the angular positions of all joints. ,speed and acceleration ;

[0066] S2, based on each sampling time... and Constructing a matrix ;

[0067] Collect all joint current data corresponding to both states at each sampling time when the robot executes the excitation trajectory under loaded and unloaded conditions, and merge the data from both states to form a set of joint current data corresponding to the current sampling time. ;

[0068] S3, N sampling times Stack them in chronological order to form a matrix. N sampling times Stack them in chronological order to form a matrix. ;in, This represents the total number of robot joints. for The number of elements in the middle, For dimension The real space, For dimension The real number space;

[0069] S4, according to and Solving the constructed dynamic parameter identification model OLS solution ,according to Determine the covariance matrix The initial value; where, For current-level load dynamic parameters, The parameter represents the frictional change in the current level caused by the load. for OLS solution, for OLS solution;

[0070] S5. Initialize the joint data weighting vector Weighted vector of stacked joint data ; , , for A column vector of rows, where all elements of the column vector are 1; for The row vector of the column, where all elements of the row vector are 1;

[0071] S6. Using the covariance matrix To each and By weighting, the normalized motor current vector is obtained. and normalized linear regression matrix , , ;

[0072] As an example, ,

[0073] S7, according to and Calculate the matrix ,according to and Calculate the matrix ;

[0074] S8, according to and ,calculate As an example, ;

[0075] S9, according to , and Calculate the standardized current residual vector As an example, ;

[0076] S10, according to renew According to the threshold and renew Through the updated renew ;

[0077] S11, Judgment If convergence has occurred, output the value from step S8. If the result is negative, return to step S6.

[0078] To mitigate the impact of measurement noise or model uncertainty on outliers in the identification results, a data weighting vector P and a matrix P extended from P are used. e To calculate and ,in, , In specific applications, if P and P e If an element in the array is 0, it means... and If the corresponding element is an outlier, its value will be set to 0; otherwise, if it is a non-outlier, it will remain unchanged. After removing outliers, and Convert them respectively and 。 operator This indicates that corresponding elements are multiplied.

[0079] In practical applications, the construction of a linear regression matrix can be achieved using existing technologies.

[0080] These are the robot's current-level dynamic parameters, including current-level load dynamic parameters. and the frictional variation parameters caused by the load at the current level In specific applications, load dynamic parameters Specifically, this may include the mass of each link, the first moment of each link, the centrifugal force of each link, the Coriolis force of each link, the gravity of each link, the moment of inertia of each joint, the static friction parameters of each joint, the Coulomb friction parameters of each joint, the viscous friction parameters of each joint, the friction offset coefficient information of each joint, and friction variation parameters. Specifically, this may include each joint. , and (j=1……n).

[0081] In this embodiment, current is used instead of joint torque to identify load dynamic parameters. The method of this invention is superior to comparative methods in predicting effective load current dynamic parameters. It exhibits a small root mean square error (RMSE) for both concentric and eccentric loads. Specifically, a new model for load dynamic parameter identification is derived based on current dynamics; a weighted iteration-based identification method is proposed to estimate the dynamic parameters at the current level, which helps to obtain weighted least squares solutions and eliminate outliers. Furthermore, a zero-velocity continuous nonlinear joint combination friction model is introduced in the current-level dynamic parameter identification process to improve overall identification accuracy. The current-level dynamic parameters of the robot are identified using the method of this invention. Then, using this and the dynamic basis parameter vector of the robot current level under no-load conditions The estimated value of the joint current was predicted, and the specific formula 9 was discussed.

[0082] This invention introduces a zero-velocity continuous nonlinear joint combination friction model to improve overall identification accuracy, and provides the joint combination friction model for constructing the robot in step S1. The implementation method is as follows:

[0083] First, for the j-th joint, the Stribeck friction model with a velocity exponent is used to characterize the viscous friction phenomenon, and the arctangent function is used instead of the sign function to improve the discontinuity of the friction model at the velocity reversal point. For joint j, a joint friction model is proposed and used as follows:

[0084] Design No. Joint friction model ,and

[0085] , ;

[0086] in, , , and Let represent the static friction coefficient, Coulomb friction coefficient, viscous friction coefficient, and friction offset coefficient of the j-th joint, respectively. Let the velocity of the j-th joint be... It is the velocity index of the j-th joint. It is the arctangent coefficient of the j-th joint. It is the Stribeck velocity of the j-th joint. It is a natural constant;

[0087] Secondly, based on all joint friction models, a joint combination friction model is constructed. ,in, .

[0088] Specifically, in step S1, the sampling time is determined. Corresponding combination coefficients , and The method for determining the value is as follows:

[0089] S1-1. Acquiring a set of joint current data at the current sampling moment during the process of the robot executing the excitation trajectory under load. And a set of joint coordinates corresponding to the current sampling time; among which, each set of joint current data Includes all joint current data under load at the current sampling time.

[0090] S1-2. Based on a set of joint coordinates corresponding to the current sampling time, construct the linear regression matrix for that sampling time. ;

[0091] S1-3, according to , and Estimate the triboelectric current at the current sampling time. ;

[0092] For the reason The vector of current inertia parameters formed by the mass, first moment, Coriolis force, centrifugal force, gravity, and rotor inertia of each link in the robot. For current-level dynamic parameters that include load dynamics and frictional changes;

[0093] S1-4, to Apply constraints and optimize the solution. , and Specifically:

[0094] ;

[0095] in, Let j be the set of optimization variables for the j-th joint. ;

[0096] , It is the velocity index of the j-th joint;

[0097] , It is the arctangent coefficient of the j-th joint;

[0098] , It is the Stribeck velocity of the j-th joint;

[0099] It is a diagonal matrix with constant coefficients for joint moments, and , This is the vector formed by all the collected joint torques.

[0100] In this preferred embodiment, the determination is given. , and The way to achieve the value of is by solving optimization equations, which can better handle the nonlinear parameter space and obtain a greater convergence speed.

[0101] In step S4, the expression for the constructed dynamic parameter identification model is: The derivation process is as follows:

[0102] For a serial robot with n rigid links, considering the dynamic characteristics of both the links and the motor rotor, the dynamic differential equation can be described in the following concise form:

[0103] (1);

[0104] In the formula, Let M(q) be the robot's coordinates, representing the robot's joint angular position, velocity, and acceleration, respectively. M(q) is the inertia matrix, and I... a It is the equivalent inertia between the motor rotor and the transmission system. Let g(q) be the matrix consisting of the centrifugal force and the Coriolis force of the link, and g(q) be the gravity vector of the link. For the j-th link, its mass, centrifugal force, Coriolis force, and gravity are respectively... , and . It is the vector of joint friction torque. It is the joint output torque vector.

[0105] For the j-th joint, the Stribeck friction model with a velocity exponent is used to characterize the viscous friction phenomenon, and the arctangent function is used instead of the sign function to improve the discontinuity of the friction model at the velocity reversal point. For joint j, a friction model is proposed and used as follows:

[0106] (2);

[0107] In the formula, , , and Let represent the static friction coefficient, Coulomb friction coefficient, viscous friction coefficient, and friction offset coefficient of the j-th joint, respectively, all of which are linear friction coefficients. It is the frictional torque of the j-th joint. It is the velocity index of the j-th joint. It is the Stribeck velocity of the j-th joint. It is the arctangent coefficient of the j-th joint. , , These are collectively referred to as the nonlinear friction coefficient of the j-th joint.

[0108] Equation (2) is a scalar equation. For a robot with n degrees of freedom, the friction scalar equations for each joint can be combined into a matrix form (j=1,……,n), i.e. Then, substituting it into equation (1) and linearizing the resulting formula, equation (1) can be written in the following linear form:

[0109] (3);

[0110] In the formula, It is the base parameter set. It is the regression matrix of robot dynamics torque level, which is The function is a full-rank matrix. Here, , , They are , , The stacked column vectors, and these three parameters are collectively referred to as the combined friction coefficient. r represents the number of columns in the L matrix, that is... The number of basis parameters included. For simplicity, each matrix will henceforth appear in the form of its first letter, e.g., L will replace [previous matrix]. .

[0111] Due to motor current There is usually a linear relationship between the joint torque τ and the joint torque τ:

[0112] (4);

[0113] In the formula, It is a diagonal matrix of constant coefficients for joint torques. For a certain model of robot, it can be obtained by measuring the torque τ of each joint and the motor current i during robot operation, and then calculating it using equation (4). By substituting equation (4) into equation (1), and then dividing each term on the right side of equation (1) by K, we obtain an expression for the robot motor current i. This expression contains information about all joint currents, and the expression for the joint current of the j-th joint is:

[0114] (5);

[0115] In the formula, They are M and I respectively a And the j-th row element of C, It is the j-th element of g. It is the element at the j-th diagonal position of K.

[0116] As can be seen from equation (3), equation (5) can also be linearized in a similar way:

[0117] (6);

[0118] In the formula, It is the current of the motor at the j-th joint. It can be obtained by extracting the j-th row element of L in equation (3). .

[0119] For a robot with n joint degrees of freedom, these n scalar equations can be expressed as:

[0120] ;

[0121] Subsequently, by stacking the above n scalar equations, the matrix form of equation (6) can be obtained:

[0122] (7);

[0123] If the linearly dependent columns in all matrices of equation (7) are removed, equation (7) can be further simplified:

[0124] (8);

[0125] In the formula, It is the set of fundamental parameters at the current level. H is a regression matrix at the current level. Similar to L, H is also a full-rank matrix. r represents the number of elements in H. Similar to L, H is also a full-rank matrix. Similar functional relationships also exist between them.

[0126] If the robot executes the same excitation trajectory twice, one with a load on its end effector and the other without, the following set of dynamic equations can be obtained:

[0127] (9);

[0128] In the formula, subscripts a and b represent the two cases of no load and load, respectively. Under the two cases of executing the excitation trajectory, the measured robot joint currents are respectively... and . and These are regression matrices calculated based on the joint coordinates of the excitation trajectory under two different conditions. It is the dynamic basis parameter vector of the robot at the current level under no-load conditions, including the robot's inertial basis parameter vector under no-load conditions. Friction base parameter vector under no load ,and . These are the load dynamic parameters at the current level, which include the mass of each link, the first moment of each link, the centrifugal force of each link, the Coriolis force of each link, the gravity of each link, the moment of inertia of each joint, the static friction parameters of each joint, the Coulomb friction parameters of each joint, the viscous friction parameters of each joint, and the friction offset coefficient of each joint. is its corresponding regression matrix. c is... The number of elements in it. These are the frictional change parameters caused by the load at the current level. Each joint has three dynamic frictional parameters. For the j-th joint, these three parameters are the joint's... , and Divide by respectively As a result, It is its corresponding regression matrix.

[0129] If the robot executes the same excitation trajectory whether carrying a load or not, then theoretically... Combining it with equation (9) and rearranging it, we can obtain the following formula:

[0130] (10);

[0131] The above formula can be expressed as the product of two matrices:

[0132] (11);

[0133] In the formula, equal , equal .in, Still about The function, It is about The function, It is about The function.

[0134] Equation (11) is a dynamic parameter identification model at a single sampling time. However, for practical applications, this invention uses data from multiple sampling times for identification. Therefore, by substituting equation (11), we can transform it into:

[0135] (12);

[0136] Equation (12) is the current-level dynamic parameter identification model established in this invention. For multiple moments The stacking, For multiple moments The stacking.

[0137] When the robot executes an excitation trajectory with and without a load, after collecting N sets of sampled data (joint currents, joint coordinates), the collected motor current vector is... and the calculated regression matrix It can be listed in the following ways:

[0138] (13);

[0139] In the formula, to These represent the motor current vectors measured from the 1st to the Nth time. to These represent the regression matrices for the 1st to Nth measurements, respectively.

[0140] In practical applications, in step S4, the solution... OLS solution The implementation method is as follows: By using generalized inverse matrix Solve Avoid Irreversible problems, and obtainable The least squares solution.

[0141] In step S4, according to Determine the covariance matrix The initial value is implemented as follows:

[0142] according to Determine the initial value of the current residual vector. ,and , For dimension The real number space;

[0143] according to Obtain the covariance matrix initial value .

[0144] In step S10, according to renew The implementation method is as follows:

[0145] ;

[0146] According to the threshold and renew The implementation method is as follows:

[0147] ;

[0148] in, Let be a function of a vector of size Nn×1, and let Exceeding the threshold Elements that are set to 0 will be set to 0, otherwise they will be set to 1; in practical applications, the threshold will be... Set to 2.5;

[0149] Through the updated renew The implementation method is as follows:

[0150] .

[0151] Verification experiment:

[0152] The method proposed in this invention was verified using a UR 10 robot. The UR 10 robot has six joints, and the end effector of the UR 10 robot under no load, concentric load, and eccentric load conditions are as follows: Figure 2 of (a), Figure 2 (b) Figure 2 As shown in (c).

[0153] Using the method proposed in this invention, the current-level load dynamic parameters of the robot end effector under concentric and eccentric loads are identified respectively. and the frictional variation parameters caused by the load at the current level The results are shown in Table 1 and Table 2, respectively.

[0154] Table 1 Load dynamic parameters and friction variation parameters when the robot end effector is subjected to a concentric load.

[0155]

[0156]

[0157] Table 2 Load dynamic parameters and frictional variation parameters when the robot end effector is subjected to an eccentric load.

[0158]

[0159]

[0160] From Tables 1 and 2 above, to This represents the load current dynamic parameters of the robot's six joints. to The parameters represent the frictional changes in the current level caused by the load at the robot's six joints. From the data given in Tables 1 and 2, it can be seen that the standard deviation of all parameter identification results is controlled within 0.01, therefore the identification results are relatively reliable.

[0161] To compare the dynamic parameter identification method proposed in this invention with the methods used in the prior art (named Method 1 (identifying robot dynamic parameters using joint torque), Method 2 (identifying robot dynamic parameters and load parameters simultaneously using weighted least squares method), and Method 3 (identifying robot dynamic parameters based on dual weighting technology) respectively), the RMSE of joint currents measured by the four methods was compared under concentric and eccentric loads at the end effector, as shown in Tables 3 and 4 respectively.

[0162] Table 3. Current Residuals (RMSE) Identified by Four Different Methods for Concentric Loads

[0163] Table 4. Identifying the Current Residual (RMSE) of Eccentric Loads Using Four Different Methods

[0164]

[0165] As can be seen from Tables 3 and 4, for each joint, the current residual identified by the method of the present invention is the smallest compared to other methods. Therefore, the method of the present invention has the highest accuracy compared to the other three methods.

[0166] Figure 3 The current on the vertical axis represents the robot joint current under load. .from Figure 3It can be seen that, without considering measurement noise, the predicted value and the measured value can be matched well. Therefore, the method proposed in this invention can accurately predict the current of each joint of the robot when it is carrying a concentric load.

[0167] Figure 4 and Figure 5 The given threshold The distribution of standardized current residuals at each joint under different values, based on Figure 4 and Figure 5 It can be seen that when the threshold When the threshold is set to 2.5, the normal probability plot does not exhibit "heavy tails" but rather forms a near-straight line with a unit slope within the threshold range, indicating that the threshold setting is appropriate. Conversely, when the value is set to 3.5, the normal probability plot exhibits "heavy tails," indicating that the normalized residuals do not fully follow a normal distribution, and the threshold setting is unreasonable.

[0168] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A method for identifying robot load dynamic parameters based on current dynamics, characterized in that, The method includes: S1. Constructing a joint assembly friction model for the robot. ,right Optimization was performed to determine the robot's trajectory execution under load at each sampling time. Corresponding combination coefficients , and The value of is determined based on the sampling time. , and Construct the corresponding linear regression matrix ;in, For the joint velocity exponential vector, Stribeck's velocity vector and The arctangent coefficient vector; When the robot executes the excitation trajectory under load, it constructs a linear regression matrix at each sampling moment based on a set of joint coordinates collected at each sampling moment. Each set of joint coordinates includes the angular positions of all joints. ,speed and acceleration ; S2, based on each sampling time... and Constructing a matrix ; Collect all joint current data corresponding to both states at each sampling time when the robot executes the excitation trajectory under loaded and unloaded conditions, and merge the data from both states to form a set of joint current data corresponding to the current sampling time. ; S3, N sampling times Stack them in chronological order to form a matrix. N sampling times Stack them in chronological order to form a matrix. ;in, This represents the total number of robot joints. for The number of elements in the middle, For dimension The real space, For dimension The real number space; S4, according to and Solving the constructed dynamic parameter identification model OLS solution ,according to Determine the covariance matrix The initial value; where, For current-level load dynamic parameters, The parameter represents the frictional change in the current plane caused by the load. for OLS solution, for OLS solution; S5. Initialize the joint data weighting vector Weighted vector of stacked joint data ; , , for A column vector of rows, where all elements of the column vector are 1; for The row vector of the column, where all elements of the row vector are 1; S6. Using the covariance matrix To each and By weighting, the normalized motor current vector is obtained. and normalized linear regression matrix , , ; S7, according to and Calculate the matrix ,according to and Calculate the matrix ; S8, according to and ,calculate ; S9, according to , and Calculate the standardized current residual vector ; S10, according to renew According to the threshold and renew Through the updated renew ; S11, Judgment If convergence has occurred, output the value from step S8. If the result is negative, return to step S6.

2. The method for identifying robot load dynamic parameters based on current dynamics according to claim 1, characterized in that, In step S1, a joint combination friction model of the robot is constructed. The implementation method is as follows: First, design the first Joint friction model ,and ， ； in, , , and Let represent the static friction coefficient, Coulomb friction coefficient, viscous friction coefficient, and friction offset coefficient of the j-th joint, respectively. Let the velocity of the j-th joint be... It is the velocity index of the j-th joint. It is the arctangent coefficient of the j-th joint. It is the Stribeck velocity of the j-th joint. It is a natural constant; Secondly, based on all joint friction models, a joint combination friction model is constructed. ,in, .

3. The method for identifying robot load dynamic parameters based on current dynamics according to claim 2, characterized in that, In step S1, determine the sampling time. Corresponding combination coefficients , and The method for obtaining the value is as follows: S1-1. Acquiring a set of joint current data at the current sampling moment during the process of the robot executing the excitation trajectory under load. And a set of joint coordinates corresponding to the current sampling time; among which, each set of joint current data Includes all joint current data under load at the current sampling time; S1-2. Based on a set of joint coordinates corresponding to the current sampling time, construct the linear regression matrix for that sampling time. ; S1-3, according to , and Estimate the triboelectric current at the current sampling time. ; For the reason The vector of current inertia parameters formed by the mass of each link, the first moment of each link, the Coriolis force of each link, the centrifugal force of each link, the gravity of each link, and the rotational inertia of the rotor of each joint motor in the robot. S1-4, to Apply constraints and optimize the solution. , and Specifically: ； in, Let j be the set of optimization variables for the j-th joint. ; , It is the velocity index of the j-th joint; , It is the arctangent coefficient of the j-th joint; , It is the Stribeck velocity of the j-th joint; It is a diagonal matrix with constant coefficients for joint moments, and , This is the vector formed by all the collected joint torques.

4. The method for identifying robot load dynamic parameters based on current dynamics according to claim 1, characterized in that, In step S4, the expression for the constructed dynamic parameter identification model is: .

5. The method for identifying robot load dynamic parameters based on current dynamics according to claim 1, characterized in that, In step S4, the solution is... OLS solution The implementation method is as follows: .

6. The method for identifying robot load dynamic parameters based on current dynamics according to claim 1, characterized in that, In step S4, according to Determine the covariance matrix The initial value is implemented as follows: according to Determine the initial value of the current residual vector. ,and , For dimension The real number space; according to Obtain the covariance matrix initial value .

7. The method for identifying robot load dynamic parameters based on current dynamics according to claim 1, characterized in that, In step S6, , .

8. The method for identifying robot load dynamic parameters based on current dynamics according to claim 1, characterized in that, In step S8, .

9. The method for identifying robot load dynamic parameters based on current dynamics according to claim 1, characterized in that, In step S9, , .

10. The method for identifying robot load dynamic parameters based on current dynamics according to claim 1, characterized in that, In step S10, according to renew The implementation method is as follows: ； According to the threshold and renew The implementation method is as follows: ； in, Let be a function of a vector of size Nn×1, and let Exceeding the threshold The element will be set to 0 if it is not set to 0, otherwise it will be set to 1; Through the updated renew The implementation method is as follows: 。

Images