A robot friction parameter identification method and device, a terminal and a medium

Abstract

The application discloses a robot friction parameter identification method and device, a terminal and a medium. The method comprises the following steps: acquiring a dynamic speed threshold of a robot joint; identifying a static friction coefficient, a coulomb friction coefficient and a critical speed in a low-speed section where the speed of the robot joint is lower than the dynamic speed threshold; and identifying a viscous friction coefficient in a medium-high speed section where the speed of the robot joint is higher than the dynamic speed threshold, in combination with the obtained static friction coefficient, the coulomb friction coefficient and the critical speed. The application has the advantage that higher-precision robot friction parameters can be obtained, thereby facilitating the construction of a high-precision robot friction model.

Description

Technical Field

[0001] This application relates to the field of robotics technology, and in particular to a method, device, terminal and medium for identifying robot friction parameters. Background Technology

[0002] In the field of robot motion control, friction is a key nonlinear factor affecting motion accuracy, dynamic response, and energy efficiency. Robot friction parameter identification, as a core technology for solving this problem, provides crucial support for achieving high-precision robot control by establishing accurate friction models and solving for key parameters.

[0003] In the prior art, patent CN 114800519 B proposes a method for identifying the dynamic parameters of a six-degree-of-freedom industrial robot considering friction. This patent identifies the gravity, inertial, Coriolis, and friction terms in the dynamic model together, without using a separate method for identifying friction. Due to the coupling effect of the gravity, inertial, and Coriolis terms, the actual friction model obtained is of poor accuracy and cannot be applied to scenarios requiring friction compensation. Patent CN114952858 B proposes a method and system for tracking the trajectory of an industrial robot based on friction compensation control. This patent uses a Coulomb-viscous model for friction compensation, which cannot consider the Stribeck effect at low speeds, resulting in low accuracy of the friction model at low speeds.

[0004] Traditional friction compensation models often use global fitting parameters for low-speed and high-speed segments when identifying them. Taking Stribeck friction as an example, the high-speed segment data overwhelmingly affects the Stribeck features in the low-speed segment. Using a global identification strategy without segmentation often results in low model fitting accuracy. Summary of the Invention

[0005] This application provides a method, device, terminal, and medium for identifying robot friction parameters. Its advantage is that it can obtain higher precision robot friction parameters, which is beneficial for constructing high-precision robot friction models.

[0006] The proposed solution is as follows:

[0007] On the one hand, this application provides a method for identifying robot friction parameters, including the following steps:

[0008] S1: Obtain the dynamic velocity threshold of the robot joints;

[0009] S2: Identify the static friction coefficient in the low-speed range when the robot joint speed is below the dynamic speed threshold. Coulomb friction coefficient and critical velocity ;

[0010] S3: In the medium-to-high speed range where the robot joint speed is higher than the dynamic speed threshold, combine the static friction coefficient obtained in step S2. Coulomb friction coefficient and critical velocity Identify the coefficient of viscous friction .

[0011] Furthermore, it also includes step S4: using the static friction coefficient Coulomb friction coefficient Critical speed and viscous friction coefficient Construct the Stribeck friction model of the robot joint.

[0012] Furthermore, step S1 includes the following steps:

[0013] When other joints are at rest, apply a duration T to the target joint. The linear sweep frequency is shown in equation (1):

[0014] (1)

[0015] Set the sampling frequency to f for joint velocity Sampling is performed every [time]. Data was sampled once, with sampling time points i being respectively , , ..., A total of samples were collected. One data point;

[0016] Record the actual speed of the joint For speed signals The target joint acceleration is obtained by filtering and performing first-order difference. ;

[0017] Define window time as ,make sure If it is divisible, then the number of sampling points in the window From the sampling point Beginning, to End, iterate through and count the current time. forward The degree of fluctuation in internal acceleration is shown in equation (2):

[0018] (2)

[0019] From point Beginning, to Finally, the gradient of the velocity is calculated by iterating through the data, as shown in equation (3):

[0020] (3)

[0021] Calculate each step according to formulas (2) and (3). and The value; setting the acceleration The threshold is Set the speed gradient The threshold is Select and The point i corresponding to the value is marked as a candidate segmentation point;

[0022] The minimum velocity value among all valid candidate points is taken as the segmentation threshold. The dynamic velocity threshold for this joint is shown in equation (4):

[0023] (4).

[0024] By combining time-domain statistics with kinematic gradients to detect velocity segmentation points using the dynamic threshold calculation method described above, cost-effective hardware is not required. Only multiplication and addition operations are needed, greatly reducing computational complexity and significantly improving noise resistance (sliding variance naturally suppresses high-frequency noise).

[0025] Furthermore, in step S1, the obtained segmentation thresholds are... Make corrections:

[0026] set up Let be the inertia of joint j, and α be the correction factor. The average inertia of multiple joints of the robot is used to correct the dynamic velocity threshold of each joint based on the inertia, as shown in Equation (5):

[0027] (5).

[0028] The above scheme introduces a joint load inertia ratio coefficient to dynamically correct the speed threshold, which can automatically adapt to changes in mechanical load and effectively solve the problem of inconsistent thresholds caused by load differences among multiple joints.

[0029] Furthermore, step S2 also includes the step of: obtaining joint friction torque data, the steps of which are as follows:

[0030] Without external force, control a certain joint of the robot to perform multiple back-and-forth trapezoidal velocity planning movements with different speeds in the forward and reverse directions within the motion range [-Q, Q]. Extract the uniform motion portion of the joint and collect n sets of joint friction torque, joint angle, and joint velocity. For the i-th data, equation (9) can be used to obtain...

[0031] (10)

[0032] The frictional torque of each joint, , These represent the joint torques collected when the positive and negative velocities reach the i-th joint angle, respectively.

[0033] According to formula (10), in Within a certain range, gradually increasing uniform forward and reverse motion is applied to the target joint; Speed ​​range, collect m sets of data, the first time collect data at a constant speed. torque at ° / s and uniform speed torque at ° / s And so on, the sampling is performed at a constant speed during the m-th time. torque and uniform speed torque Similarly, continue collecting torque data m+1 times up to n times, ensuring at least 20 times between m+1 and n; collect uniform velocity data on the nth time. torque at ° / s and uniform speed torque at ° / s Therefore, the individual frictional torque for each instance can be calculated. .

[0034] By using the above-mentioned method for obtaining joint friction torque data, friction torque can be extracted from complex dynamic torques, and the friction torque is easy to obtain.

[0035] Furthermore, in the low-speed range where the robot joint speed is below the dynamic speed threshold, the static friction coefficient is identified without considering the viscous friction term in the Stribeck friction model. Coulomb friction coefficient and critical velocity ;

[0036] Collect m sets of joint friction torque data and construct the regression matrix as shown in equation (11):

[0037] (11)

[0038] Next, determine , and Initial values:

[0039] set up and To approximate the mean torque at zero speed from both the positive and negative directions, the initial value of the static friction coefficient is calculated using a bidirectional approximation method. As shown in equation (12):

[0040] (12)

[0041] In equation (12), , ;

[0042] Calculate m sets of torque data The average value of the average value can be used to obtain the initial value of the Coulomb friction force. As shown in equation (13):

[0043] (13);

[0044] get and Then, calculate according to formula (14). initial value By measuring m sets of data, the difference in torque between adjacent sets is calculated and divided by the velocity to obtain multiple sets of data. And select the largest set and substitute it into equation (14) for calculation:

[0045] (14)

[0046] Get , and initial value , and Then, the parameters of equation (11) are estimated using the nonlinear least squares method, and the parameters of the Stribeck friction model are identified. , and .

[0047] Furthermore, step S3 includes the following steps:

[0048] When the robot joint speed is above the dynamic speed threshold in the medium-to-high speed range, the already identified static friction coefficient will be... Coulomb friction coefficient and critical velocity Substituting this as a constant into the Stribeck friction model, only the viscous friction coefficient is identified. ;

[0049] Collect n sets of joint friction torque data, of which m sets are low-speed data, nm sets are medium-speed data, and nm>20;

[0050] The regression matrix is ​​constructed based on n sets of data as shown in equation (15):

[0051] (15)

[0052] Next, determine The initial value is obtained according to formula (16). Linear regression was performed on the data to obtain the viscous friction coefficient. initial value :

[0053] (16)

[0054] Get initial values Then, the parameters of equation (15) were estimated using the nonlinear least squares method, and the parameters of the Stribeck friction model were identified. .

[0055] The advantages of segmented identification of friction parameters mentioned above include the following:

[0056] 1) Accurately capture nonlinear characteristics:

[0057] Identifying the static friction coefficient at low speeds Coulomb friction coefficient and critical velocity To avoid interference from high-speed data in the static friction zone, the focus is on the viscous friction coefficient in the medium to high speed range. Fitting improves parameter independence.

[0058] 2) Strong anti-interference ability:

[0059] Speed ​​ranges are segmented by dynamic thresholds (such as the AVG method) to automatically adapt to changes in load and lubrication, thereby reducing the impact of noise.

[0060] 3) High computational efficiency:

[0061] Phased optimization reduces parameter coupling and avoids the complexity and divergence risk of global nonlinear fitting.

[0062] 4) Ease of use:

[0063] The physical meaning of the parameters at each stage is clear, which facilitates fault diagnosis (e.g.) An error message indicating lubrication failure was displayed.

[0064] In another aspect, this application provides a robot friction parameter identification device, comprising:

[0065] The dynamic velocity threshold acquisition unit acquires the dynamic velocity threshold of the robot joints; while other joints are stationary, a linear sweep velocity of duration T is applied to the target joint, and the joint velocity is sampled by setting a sampling frequency; the joint velocity is recorded, and first-order difference is performed to obtain the target joint acceleration; the fluctuation degree of acceleration and the velocity gradient at each acquisition time are traversed and statistically analyzed, and the points corresponding to the acceleration fluctuation degree and velocity gradient values ​​exceeding the preset threshold are selected and marked as candidate segment points; the minimum velocity value among all candidate points is taken as the dynamic velocity threshold of the joint;

[0066] The torque acquisition unit is used to collect multiple sets of robot joint friction torque data;

[0067] The parameter identification unit constructs a regression matrix using robot joint friction torque data in the low-speed range where the robot joint speed is below the dynamic speed threshold, and identifies the static friction coefficient, Coulomb friction coefficient, and critical speed. Based on the identified static friction coefficient, Coulomb friction coefficient, critical speed, and robot joint friction torque data, it constructs a regression matrix to identify the viscous friction coefficient.

[0068] In another aspect, this application provides a terminal including a memory and a processor, wherein the memory stores a computer program, and when the computer program is invoked and executed by the processor, it implements the method described above.

[0069] In another aspect, this application provides a computer-readable medium storing a computer program that, when executed by a computer, implements the method described above.

[0070] In summary, the beneficial effects of this application are as follows:

[0071] 1. By identifying different parameters of the friction model separately, this application can avoid the coupling effects of gravity, inertial force and Coriolis force terms, and obtain friction model parameters with high accuracy.

[0072] 2. The Stribeck friction model can be used to account for the Stribeck effect in the low-speed range. The friction model has high accuracy in the low-speed range and can be used in situations where low speed is required, while also taking into account medium and high-speed situations.

[0073] 3. Abandoning the drawbacks of traditional fixed segmentation points, we design an adaptive segmentation rule based on dynamic velocity / acceleration thresholds. This dynamic segmentation method can accurately capture the characteristic changes of different velocity ranges, especially in the low-speed range and the inversion transition zone, which greatly improves the modeling accuracy and effectively reduces the errors generated by traditional methods in these areas.

[0074] 4. A piecewise fitting method is adopted to avoid interference from high-speed data in the static friction zone and improve parameter independence; at the same time, phased optimization reduces parameter coupling and avoids the complexity and divergence risk of global nonlinear fitting. Attached Figure Description

[0075] Figure 1 This is a flowchart of the robot friction parameter identification method of this application. Detailed Implementation

[0076] The specific embodiments of this application are described in detail below with reference to the accompanying drawings.

[0077] The following detailed embodiments use a typical six-joint industrial robot as an example. It should be understood that the solution proposed in this application is also applicable to other types of robots. In a typical six-joint industrial robot, the torque required for the motor to overcome friction accounts for more than 20% of the total torque, making it essential to identify and compensate for frictional torque. Taking the Stribeck friction model as an example, when using the Stribeck friction model for identification, the inflection point speeds of the low-speed and medium-to-high-speed segments of the six joints are not the same, making existing technical solutions difficult to apply.

[0078] One specific embodiment of this application provides a method for identifying robot friction parameters, referencing... Figure 1 This includes the following steps:

[0079] S1: Obtain the dynamic velocity threshold of the robot joints.

[0080] As mentioned above, when using the Stribeck friction model for identification, the inflection point velocities of the low-speed and medium-to-high-speed segments of the 6 joints are not the same. This embodiment proposes a "dynamic threshold segmentation method based on the joint criterion of acceleration variance and velocity gradient," abbreviated as "AVG dynamic threshold method" (Acceleration Variance-Gradient Method). Taking a target joint as an example, the specific steps are as follows:

[0081] When other joints are at rest, apply a duration T to the target joint. The linear sweep frequency is shown in equation (1):

[0082] (1)

[0083] Set the sampling frequency to f (in Hz) for joint velocity Sampling is performed every [time]. Data was sampled once, with sampling time points i being respectively , , ..., A total of samples were collected. One data point;

[0084] Record the actual speed of the joint For speed signals The target joint acceleration is obtained by filtering and performing first-order difference. ;

[0085] Define window time as ,make sure If it is divisible, then the number of sampling points in the window From the sampling point Beginning, to End, iterate through and count the current time. forward The degree of fluctuation in internal acceleration is shown in equation (2):

[0086] (2)

[0087] From point Beginning, to Finally, the gradient of the velocity is calculated by iterating through the data, as shown in equation (3):

[0088] (3)

[0089] Calculate each step according to formulas (2) and (3). and The value; setting the acceleration The threshold is Set the speed gradient The threshold is Select and The point i corresponding to the value is marked as a candidate segmentation point. In this embodiment, for ;

[0090] The minimum velocity value among all valid candidate points is taken as the segmentation threshold. The dynamic velocity threshold for this joint is shown in equation (4):

[0091] (4).

[0092] set up Let be the inertia of joint j, and α be a correction factor, where α takes a positive value less than 1. The average inertia of multiple joints of the robot is used to correct the dynamic velocity threshold of each joint based on the inertia, as shown in Equation (5):

[0093] (5).

[0094] The dynamic threshold calculation method described above combines time-domain statistics with kinematic gradients to detect velocity segmentation points. This eliminates the need for expensive hardware, requiring only multiplication and addition operations, significantly reducing computational complexity and improving noise immunity (sliding variance naturally suppresses high-frequency noise). Furthermore, by introducing a joint load inertia proportionality coefficient to dynamically correct the velocity threshold, the method automatically adapts to changes in mechanical load, effectively resolving the inconsistency in thresholds caused by load differences across multiple joints.

[0095] Before proceeding, we will first introduce the principle of frictional torque separation. For a 6-joint industrial robot, its dynamic equation is shown in equation (6):

[0096] (6)

[0097] Equation (1) describes the relationship between various forces and torques during the robot's motion, where Let q be the frictional torque at each joint. , The position, velocity, and acceleration of each joint are recorded separately.

[0098] Define motion state 1 as Motion state 2 is The torques corresponding to states 1 and 2 are as follows. and These can be expressed as equations (7) and (8), respectively:

[0099] (7)

[0100] (8)

[0101] because ,and Then, from (7)-(8), we can obtain:

[0102] (9)

[0103] Equation (9) shows that when the designed motion state is met, the difference in torque between the two motion states is exactly the current velocity. This corresponds to twice the frictional torque value, thus enabling the extraction of frictional torque from complex dynamic torques.

[0104] Based on the above principles, a method for obtaining joint friction torque data is proposed as follows:

[0105] For simplicity, without external force, the robot's joint is controlled to perform multiple back-and-forth trapezoidal velocity planning movements with different speeds in the forward and reverse directions within the motion range [-Q, Q]. The uniform motion portion of the joint is selected, and the joint torque, joint angle, and joint velocity are collected n times. For the i-th data, equation (9) yields:

[0106] (10)

[0107] The frictional torque of each joint, , These represent the joint torques collected when the positive and negative velocities reach the i-th joint angle, respectively.

[0108] According to formula (10), in Within a certain range, gradually increasing uniform forward and reverse motion is applied to the target joint; Speed ​​range, collect m sets of data, the first time collect data at a constant speed. torque at ° / s and uniform speed torque at ° / s And so on, the sampling is performed at a constant speed during the m-th time. torque and uniform speed torque Similarly, continue collecting torque data m+1 times up to n times, ensuring at least 20 times between m+1 and n; collect uniform velocity data on the nth time. torque at ° / s and uniform speed torque at ° / s Therefore, the individual frictional torque for each instance can be calculated. .

[0109] The friction model is segmented and its parameters are identified, including steps S2 and S3.

[0110] S2: Identify the static friction coefficient in the low-speed range when the robot joint speed is below the dynamic speed threshold. Coulomb friction coefficient and critical velocity ;

[0111] In the low-speed range where the robot joint speed is below the dynamic speed threshold, the static friction coefficient is identified without considering the viscous friction term in the Stribeck friction model. Coulomb friction coefficient and critical velocity ;

[0112] Collect m sets of joint friction torque data according to the method described in equation (10), and construct the regression matrix as shown in equation (11):

[0113] (11)

[0114] Next, determine , and Initial values:

[0115] set up and To approximate the mean torque at zero speed from both the positive and negative directions, the initial value of the static friction coefficient is calculated using a bidirectional approximation method. As shown in equation (12):

[0116] (12)

[0117] In equation (12), , ;

[0118] Calculate m sets of torque data The average value of the average value can be used to obtain the initial value of the Coulomb friction force. As shown in equation (13):

[0119] (13);

[0120] get and Then, calculate according to formula (14). initial value By measuring m sets of data, the difference in torque between adjacent sets is calculated and divided by the velocity to obtain multiple sets of data. And select the largest set and substitute it into equation (14) for calculation:

[0121] (14)

[0122] Get , and initial value , and Then, the parameters of equation (11) are estimated using the nonlinear least squares method, and the parameters of the Stribeck friction model are identified. , and .

[0123] S3: In the medium-to-high speed range where the robot joint speed is higher than the dynamic speed threshold, combine the static friction coefficient obtained in step S2. Coulomb friction coefficient and critical velocity Identify the coefficient of viscous friction .

[0124] When the robot joint speed is above the dynamic speed threshold in the medium-to-high speed range, the already identified static friction coefficient will be... Coulomb friction coefficient and critical velocity Substituting this as a constant into the Stribeck friction model, only the viscous friction coefficient is identified. ;

[0125] Collect n sets of joint friction torque data, of which m sets are low-speed data, nm sets are medium-speed data, and nm>20;

[0126] The regression matrix is ​​constructed based on n sets of data as shown in equation (15):

[0127] (15)

[0128] Next, determine The initial value is obtained according to formula (16). Linear regression was performed on the data to obtain the viscous friction coefficient. initial value :

[0129] (16)

[0130] Get initial values Then, the parameters of equation (15) were estimated using the nonlinear least squares method, and the parameters of the Stribeck friction model were identified. .

[0131] At this point, the parameters of the Stribeck friction model have been fully identified. , , and .

[0132] S4: By static friction coefficient Coulomb friction coefficient Critical speed and viscous friction coefficient Construct the Stribeck friction model of the robot joint.

[0133] The advantages of segmented identification of friction parameters mentioned above include the following:

[0134] 1) Accurately capture nonlinear characteristics:

[0135] Identifying the static friction coefficient at low speeds Coulomb friction coefficient and critical velocity To avoid interference from high-speed data in the static friction zone, the focus is on the viscous friction coefficient in the medium to high speed range. Fitting improves parameter independence.

[0136] 2) Strong anti-interference ability:

[0137] Speed ​​ranges are segmented by dynamic thresholds (such as the AVG method) to automatically adapt to changes in load and lubrication, thereby reducing the impact of noise.

[0138] 3) High computational efficiency:

[0139] Phased optimization reduces parameter coupling and avoids the complexity and divergence risk of global nonlinear fitting.

[0140] 4) Ease of use:

[0141] The physical meaning of the parameters at each stage is clear, which facilitates fault diagnosis (e.g.) An error message indicating lubrication failure was displayed.

[0142] In another specific embodiment, the above method is used to specifically describe an industrial robot as follows.

[0143] 1. Dynamic speed threshold calculation

[0144] With all other joints stationary, a linear sweep rate from 1° / s to 100° / s is applied to the target joint for 10 seconds. The formula for the velocity change over time is: The sampling frequency is set to 1000Hz, and every [time / percentage]... A total of 10,001 data points were collected in one data collection session.

[0145] During the data acquisition process, the actual speed of the joints was accurately recorded. The acceleration signal is filtered, and the target joint acceleration is obtained through first-order difference operation. Define the window time. This indicates the number of sampling points within the window. From sampling points From the beginning until End, according to the formula Statistical analysis of each current moment The degree of acceleration fluctuation in the first 0.1 seconds.

[0146] At the same time, from point Beginning, to End, according to the formula Iterate through the gradient of the statistical velocity. (Typically 0.5° / s) 2 ).

[0147] Calculate one by one to as well as to Choose a value that satisfies and Point i is identified and marked as a candidate segmentation point. Among all valid candidate points, the minimum velocity value is selected as the segmentation threshold. Then, adjustments are made based on the joint inertia. Assume the inertia of joint j is... The average inertia of the 6 joints is Let α = 0.2, and use the formula... The corrected speed threshold is obtained.

[0148] 2. Identification of segmented parameters for the friction model

[0149] 2.1 Low-speed segment identification

[0150] At 1° / s to In the velocity segment, 20 sets of data were collected (m=20). Based on the principle of frictional torque separation, the robot joints were controlled to perform multiple back-and-forth trapezoidal velocity planning movements at different speeds in the forward and reverse directions within the motion range [-Q, Q]. Data was collected from the uniform motion portion. For example, the torque during the first uniform forward motion at 1° / s was collected. Torque during uniform motion at -1° / s in the opposite direction This process continues until a uniform speed is achieved. Torque during forward and reverse motion.

[0151] A regression matrix is ​​constructed based on the collected data, and the initial value of the static friction coefficient is calculated using the two-way approximation method. ,Right now ,in , Calculate 20 sets of torque data. The average value is used to obtain the initial value of the Coulomb friction coefficient. Multiple groups are obtained by calculating the difference in torque between adjacent groups and dividing it by the velocity using measured data. Select the largest set and substitute it into the formula. Calculate the initial value of the critical velocity. Finally, the nonlinear least squares method was used to estimate the parameters of the regression matrix, thereby identifying the static friction coefficient in the low-speed range. Coulomb friction coefficient and critical velocity .

[0152] 2.2 Identification of medium and high speed sections

[0153] exist Continue collecting data up to the 100° / s speed range (ensuring at least 20 data collections between m+1 and n). Then, process the identified low-speed parameters. , and Substitute these values ​​as constants into the Stribeck friction model. Data is collected and a regression matrix is ​​constructed using a method similar to that used for the low-speed section. Linear regression is then performed on the data between [m+1,n], according to the formula... Calculate the initial value of the viscous friction coefficient. Then, by using the nonlinear least squares method for estimation, the viscous friction coefficient was finally identified. .

[0154] Another specific embodiment of this application provides a robot friction parameter identification device, including:

[0155] The dynamic velocity threshold acquisition unit acquires the dynamic velocity threshold of the robot joints; while other joints are stationary, a linear sweep velocity of duration T is applied to the target joint, and the joint velocity is sampled by setting a sampling frequency; the joint velocity is recorded, and first-order difference is performed to obtain the target joint acceleration; the fluctuation degree of acceleration and the velocity gradient at each acquisition time are traversed and statistically analyzed, and the points corresponding to the acceleration fluctuation degree and velocity gradient values ​​exceeding the preset threshold are selected and marked as candidate segment points; the minimum velocity value among all candidate points is taken as the dynamic velocity threshold of the joint;

[0156] The torque acquisition unit is used to collect multiple sets of robot joint friction torque data;

[0157] The parameter identification unit constructs a regression matrix using robot joint friction torque data in the low-speed range where the robot joint speed is below the dynamic speed threshold, and identifies the static friction coefficient, Coulomb friction coefficient, and critical speed. Based on the identified static friction coefficient, Coulomb friction coefficient, critical speed, and robot joint friction torque data, it constructs a regression matrix to identify the viscous friction coefficient.

[0158] Another specific embodiment of this application provides a terminal, including a memory and a processor. The memory stores a computer program, and when the computer program is called and executed by the processor, it implements the robot friction parameter identification method as described above.

[0159] Another specific embodiment of this application provides a computer-readable medium storing a computer program, which, when executed by a computer, implements the robot friction parameter identification method as described above.

[0160] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several modifications and improvements can be made without departing from the inventive concept of this application, and these all fall within the protection scope of this application.

Claims

1. A method for identifying robot friction parameters, characterized in that, Includes the following steps: S1: Obtain the dynamic velocity threshold of the robot joints; S2: Identify the static friction coefficient in the low-speed range when the robot joint speed is below the dynamic speed threshold. Coulomb friction coefficient and critical velocity ; S3: In the medium-to-high speed range where the robot joint speed is higher than the dynamic speed threshold, combine the static friction coefficient obtained in step S2. Coulomb friction coefficient and critical velocity Identify the coefficient of viscous friction ; Step S3 includes the following steps: When the robot joint speed is above the dynamic speed threshold in the medium-to-high speed range, the already identified static friction coefficient will be... Coulomb friction coefficient and critical velocity Substituting this as a constant into the Stribeck friction model, only the viscous friction coefficient is identified. ; Collect n sets of joint friction torque data, of which m sets are low-speed data, nm sets are medium-speed data, and nm>20; The regression matrix is ​​constructed based on n sets of data as shown in equation (15): (15) Next, determine The initial value is obtained according to formula (16). Linear regression was performed on the data to obtain the viscous friction coefficient. initial value : (16) Where i represents the i-th set of collected data. Let be the frictional torque of the i-th joint. This represents the joint velocity data for the i-th group; Get initial values Then, the parameters of equation (15) were estimated using the nonlinear least squares method, and the parameters of the Stribeck friction model were identified. ; S4: By static friction coefficient Coulomb friction coefficient Critical speed and viscous friction coefficient Construct the Stribeck friction model of the robot joint.

2. The robot friction parameter identification method according to claim 1, characterized in that, Step S1 includes the following steps: When other joints are at rest, apply a duration T to the target joint. The linear sweep frequency is shown in equation (1): (1) Set the sampling frequency to f for joint velocity Sampling is performed every [time]. Data was sampled once, and a total of samples were collected. One data point; Record the actual speed of the joint For speed signals The target joint acceleration is obtained by filtering and performing first-order difference. ; Define window time as The number of sampling points in the window From the sampling point Beginning, to End, iterate through and count the current time. forward The degree of fluctuation in internal acceleration is shown in equation (2): （2） From point Beginning, to Finally, the gradient of the velocity is calculated by iterating through the data, as shown in equation (3): (3) Calculate each step according to formulas (2) and (3). and The value; setting the acceleration The threshold is Set the speed gradient The threshold is Select and The point corresponding to the value of i is marked as a candidate segmentation point; Take the minimum velocity value among all valid candidate points as the segmentation threshold. The dynamic velocity threshold for this joint is shown in equation (4): (4)。 3. The robot friction parameter identification method according to claim 2, characterized in that, In step S1, the obtained segmented thresholds are... Make corrections: set up Let be the inertia of joint j, and α be the correction factor. The average inertia of multiple joints of the robot is used to correct the dynamic velocity threshold of each joint based on the inertia, as shown in Equation (5): （5）。 4. The robot friction parameter identification method according to claim 3, characterized in that, Step S2 also includes the step of: obtaining joint friction torque data, the steps of which are as follows: Without external force, control a certain joint of the robot to perform multiple back-and-forth trapezoidal velocity planning movements with different speeds in the forward and reverse directions within the motion range [-Q, Q]; extract the uniform motion part of the joint and collect the joint friction torque, joint angle and joint velocity n times respectively; For the i-th data, we get (10) The frictional torque of each joint, , These represent the joint torques collected when the positive and negative velocities reach the i-th joint angle, respectively. According to formula (10), in Within a certain range, gradually increasing uniform forward and reverse motion is applied to the target joint; Speed ​​range, collect m sets of data, the first time collect data at a constant speed. torque at ° / s and uniform speed torque at ° / s And so on, the sampling is performed at a constant speed during the m-th time. torque and uniform speed torque Similarly, continue collecting torque data m+1 times up to n times, ensuring at least 20 times between m+1 and n; collect uniform velocity data on the nth time. torque at ° / s and uniform speed torque at ° / s Therefore, the individual frictional torque for each instance can be calculated. .

5. The robot friction parameter identification method according to any one of claims 1-4, characterized in that, In the low-speed range where the robot joint speed is below the dynamic speed threshold, the static friction coefficient is identified without considering the viscous friction term in the Stribeck friction model. Coulomb friction coefficient and critical velocity ; Collect m sets of joint friction torque data and construct the regression matrix as shown in equation (11): （11） Next, determine , and Initial values: set up and To approximate the mean torque at zero speed from both the positive and negative directions, the initial value of the static friction coefficient is calculated using a bidirectional approximation method. As shown in equation (12): (12) In equation (12), , ; Calculate m sets of torque data The average value of the average value can be used to obtain the initial value of the Coulomb friction force. As shown in equation (13): （13）； get and Then, calculate according to formula (14). initial value By measuring m sets of data, the difference in torque between adjacent sets is calculated and divided by the velocity to obtain multiple sets of data. And select the largest set and substitute it into equation (14) for calculation: (14) Get , and initial value , and Then, the parameters of equation (11) are estimated using the nonlinear least squares method, and the parameters of the Stribeck friction model are identified. , and .

6. A robot friction parameter identification device, characterized in that, To implement the steps in the method as described in any one of claims 1-5, including: The dynamic velocity threshold acquisition unit acquires the dynamic velocity threshold of the robot joints; while other joints are stationary, a linear sweep velocity of duration T is applied to the target joint, and the joint velocity is sampled by setting a sampling frequency; the joint velocity is recorded, and first-order difference is performed to obtain the target joint acceleration; the fluctuation degree of acceleration and the velocity gradient at each acquisition time are traversed and statistically analyzed, and the points corresponding to the acceleration fluctuation degree and velocity gradient values ​​exceeding the preset threshold are selected and marked as candidate segment points; the minimum velocity value among all candidate points is taken as the dynamic velocity threshold of the joint; The torque acquisition unit is used to collect multiple sets of robot joint friction torque data; The parameter identification unit constructs a regression matrix using robot joint friction torque data in the low-speed range where the robot joint speed is below the dynamic speed threshold, and identifies the static friction coefficient, Coulomb friction coefficient, and critical speed. Based on the identified static friction coefficient, Coulomb friction coefficient, critical speed, and robot joint friction torque data, it constructs a regression matrix to identify the viscous friction coefficient.

7. A terminal, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is invoked and executed by the processor, it implements the method as described in any one of claims 1-5.

8. A computer-readable medium, characterized in that, The computer-readable medium stores a computer program that, when executed by a computer, implements the method as described in any one of claims 1-5.

Images