Harmonic reducer vibration suppression method and system based on hybrid parameter identification

Abstract

The invention discloses a harmonic reducer vibration suppression method based on hybrid parameter identification, and relates to the technical field of flexible robot motion control, and the method comprises the steps: building a harmonic reducer vibration compensation model based on the physical characteristics of a harmonic reducer; static parameters of the harmonic reducer vibration compensation model are obtained through off-line identification; dynamic parameters of the harmonic reducer vibration compensation model are obtained through online identification; and calculating a feed-forward compensation torque of the motor based on the harmonic reducer vibration compensation model, superposing the feed-forward compensation torque to a motor torque instruction, and outputting a final motor control signal. According to the method, direct feed-forward compensation is adopted instead of an observation or filtering method, time delay is avoided, resistance changes can be responded in real time, the accuracy of phase compensation is guaranteed, meanwhile, the phase margin of a control system is not reduced, and the robustness of the control system is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of flexible robot motion control, in particular to a harmonic reducer vibration suppression method and system based on hybrid parameter identification. BACKGROUND

[0002] As a kind of precision transmission device, harmonic reducer is widely used in high-precision fields such as robots and aerospace. However, due to the special structure of the harmonic reducer with a difference of 2 between the inner and outer teeth, the wave generator will produce uneven resistance during movement. Specifically, there will be two periodic resistance fluctuations of sine wave every rotation of the wave generator. This resistance fluctuation will cause periodic tracking errors, and the periodic tracking errors of each joint will be superimposed on the end of the mechanical arm, which is characterized as vibration. This vibration is particularly evident when the mechanical arm is running at high speed, seriously affecting the motion accuracy and stability of the flexible robot.

[0003] Currently, the main technical solutions for harmonic reducer 2 times vibration suppression are as follows: CN115629533A: A disturbance observer is used to estimate the harmonic torque and perform feedforward compensation. This method estimates the disturbance in real time through the observer, but there is a estimation lag problem, which leads to inaccurate compensation timing and reduces the phase margin of the control system, affecting the stability of the system.

[0004] CN116872208A: A notch filter is designed based on the harmonic oscillation frequency. This method identifies the vibration frequency and performs filtering, but the notch filter introduces phase delay and has poor adaptability to frequency changes, which is not effective in variable speed motion.

[0005] CN114415521A: A method combining modeling and observer. This method establishes a full dynamics model of the flexible joint, but involves too many parameters, the model accuracy is limited, and still relies on the observer for compensation, which cannot completely eliminate the time lag effect.

[0006] The above existing technologies all have a common defect: compensation time lag. Whether it is a disturbance observer or a notch filter, it will introduce a certain phase delay, which will reduce the phase margin of the control system and affect the stability and robustness of the system. Especially in high-speed, high-precision application scenarios, this time lag will greatly reduce the vibration suppression effect, and may even exacerbate the amplitude. SUMMARY

[0007] The technical problem to be solved by the present application is to overcome the shortcomings of the prior art and provide a harmonic reducer vibration suppression method and system based on hybrid parameter identification.

[0008] To solve the above technical problems, the technical solution of the present application is as follows: A method for suppressing vibration in a harmonic reducer based on hybrid parameter identification includes: A vibration compensation model for harmonic reducers is constructed based on their physical characteristics. The static parameters of the vibration compensation model of the harmonic reducer are obtained by offline identification, and the static parameters include amplitude parameters. The dynamic parameters of the vibration compensation model of the harmonic reducer are obtained by online identification, and the dynamic parameters include the phase angle; The feedforward compensation torque of the motor is calculated based on the vibration compensation model of the harmonic reducer, and the feedforward compensation torque is superimposed on the motor torque command to output the final motor control signal.

[0009] As a preferred embodiment of the harmonic reducer vibration suppression method based on hybrid parameter identification described in this invention, the harmonic reducer vibration compensation model is as follows: ; in, This corresponds to the second harmonic. Position of the motor; Phase angle; Let be the amplitude as a function of the load, and , and For amplitude parameters; This is the feedback torque of the motor.

[0010] As a preferred embodiment of the harmonic reducer vibration suppression method based on hybrid parameter identification described in this invention, wherein: obtaining the static parameters of the harmonic reducer vibration compensation model through offline identification includes: Control the target joint of the robot to move along a preset trajectory, and collect the feedback torque and position data of the motor; The collected data were divided into n groups according to the magnitude of the motor feedback torque. For each group of data, the model was... Perform least-squares fitting to obtain the amplitude under the corresponding load. and phase angle And calculate the arithmetic mean of the feedback torque for each group of motors. ; Based on n sets of amplitudes The arithmetic mean of the motor feedback torque By fitting a linear relationship using the least squares method, the amplitude parameters of the amplitude-load model are obtained. and ; For n sets of phase angles The arithmetic mean is taken to obtain the initial phase angle value used for online identification. .

[0011] As a preferred embodiment of the harmonic reducer vibration suppression method based on hybrid parameter identification described in this invention, wherein: the model Perform least-squares fitting to obtain the amplitude under the corresponding load. and phase angle include: For the model Linearization is performed to obtain ,in, , The length of the sampled data, This is the difference between the actual torque and the theoretical torque of the motor. Let be the regression matrix, and , Let be the parameter vector to be identified, and ; According to the formula and Calculate amplitude and phase angle .

[0012] As a preferred embodiment of the harmonic reducer vibration suppression method based on hybrid parameter identification described in this invention, wherein: obtaining the dynamic parameters of the harmonic reducer vibration compensation model through online identification includes: Get the motor position in the current control cycle and motor feedback torque ; Motor torque was calculated using a full dynamics model. And calculate the difference between it and the motor feedback torque. , ; The amplitude at the current moment is calculated based on the amplitude-load model. ; Calculate the regression matrix ,and ; Calculate the gain matrix for the current control cycle. ,and ,in, R is the covariance matrix of the estimation error calculated in the previous control cycle, and R is the noise variance of the Gaussian white noise of the motor. Calculate the parameter vector to be identified ,and ,in, This represents the parameter vector identification result of the previous control cycle; Through formula Calculate the current phase angle .

[0013] As a preferred embodiment of the harmonic reducer vibration suppression method based on hybrid parameter identification described in this invention, wherein: in the process of obtaining the motor position of the current control cycle... and motor feedback torque Previously, it also included: Determine if the current state is the first control cycle; if so, then... , ,and , If not, skip this step.

[0014] As a preferred embodiment of the harmonic reducer vibration suppression method based on hybrid parameter identification described in this invention, the method includes: calculating the feedforward compensation torque of the motor based on the harmonic reducer vibration compensation model, and superimposing the feedforward compensation torque into the motor torque command to output the final motor control signal, including: Based on the current phase angle Calculate feedforward compensation torque ,and ; feedforward compensation torque Motor torque calculated with the full dynamics model Superimposed, the target feedforward torque is obtained. ,Right now It then outputs the final motor control signal.

[0015] As a preferred embodiment of the harmonic reducer vibration suppression method based on hybrid parameter identification described in this invention, wherein: in the process of using the formula Calculate the current phase angle Following that, it also includes: Through formula Calculate the covariance matrix of the current control cycle estimation error, where, For a unit matrix, the initial value is ; The covariance matrix of the estimation error calculated in the previous control cycle. Updated to The parameter vector identification results of the previous control cycle Updated to .

[0016] As a preferred embodiment of the harmonic reducer vibration suppression method based on hybrid parameter identification described in this invention, after calculating the feedforward compensation torque of the motor based on the harmonic reducer vibration compensation model, superimposing the feedforward compensation torque into the motor torque command, and outputting the final motor control signal, the method further includes: Wait for the next control cycle to start, and after jumping to the next control cycle, obtain the motor position of the current control cycle again. and motor feedback torque Then, proceed with the subsequent steps in sequence.

[0017] The present invention also provides a harmonic reducer vibration suppression system based on hybrid parameter identification, comprising: The model building module is used to build a vibration compensation model for harmonic reducers based on their physical characteristics. The offline identification module is used to obtain the static parameters of the vibration compensation model of the harmonic reducer through offline identification, and the static parameters include amplitude parameters. An online identification module is used to obtain the dynamic parameters of the harmonic reducer vibration compensation model through online identification, and the dynamic parameters include the phase angle; The torque compensation module is used to calculate the feedforward compensation torque of the motor based on the vibration compensation model of the harmonic reducer, and to superimpose the feedforward compensation torque into the motor torque command to output the final motor control signal.

[0018] The beneficial effects of this invention are: (1) This invention adopts direct feedforward compensation instead of observation or filtering methods, abandons the idea of ​​observer and notch filter, and directly performs feedforward compensation based on physical model. The biggest advantage of this scheme is that it avoids time delay, can respond to resistance changes in real time, ensures the accuracy of compensation phase, and does not reduce the phase margin of the control system, thus ensuring the robustness of the control system.

[0019] (2) This invention adopts a hybrid parameter identification strategy to balance computational complexity and adaptability. It can not only track the changes in vibration amplitude caused by load changes and the phase changes caused by velocity changes, but also maintain good performance under different working conditions. At the same time, compared with the full-parameter online identification method, this invention determines most parameters offline and only updates one parameter in the online stage, minimizing the number of online identification parameters, reducing the real-time computational burden and parameter drift risk, and reducing the computational burden while ensuring adaptability and reliability of online identification results.

[0020] (3) In terms of suppressing high-frequency vibration caused by the second harmonic resistance of the harmonic reducer, the present invention can reduce the amplitude by about 40% and maintain the original level of phase margin. The system stability is significantly better than the prior art. At the same time, the real-time calculation burden is small, no high-performance processor is required, and the system hardware cost is reduced. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A schematic flowchart of the harmonic reducer vibration suppression method based on hybrid parameter identification provided by the present invention; Figure 2 This is a schematic diagram of the resistance fluctuation characteristics of a harmonic reducer. Figure 3 A flowchart illustrating step S102 in the harmonic reducer vibration suppression method based on hybrid parameter identification provided by the present invention. Figure 4 This is a flowchart illustrating steps S103 and S104 of the harmonic reducer vibration suppression method based on hybrid parameter identification provided by the present invention. Figure 5 This is a schematic diagram illustrating the compensation effect of the amplitude-load model using experimental data. Figure 6 for Figure 5 A magnified view of a portion of the image; Figure 7 A comparison chart showing the compensation effect of RLS online phase identification using experimental data; Figure 8 This is a comparison chart showing the vibration suppression effect using experimental data; Figure 9 A schematic diagram of the harmonic reducer vibration suppression system based on hybrid parameter identification provided by the present invention. Detailed Implementation

[0023] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0024] In the aforementioned background technologies, both disturbance observers and notch filters introduce a certain phase delay, which reduces the phase margin of the control system and affects its stability and robustness. Especially in high-speed, high-precision applications, this time delay significantly reduces vibration suppression effectiveness and may even exacerbate amplitude problems. This invention provides a vibration suppression method for harmonic reducers based on hybrid parameter identification. This method establishes a mathematical model of the second harmonic resistance of the harmonic reducer, adds feedforward to the real-time control system to compensate for the second harmonic resistance, and determines amplitude-related parameters offline and updates phase parameters online. This achieves an optimal balance between computational efficiency and adaptability, effectively suppressing high-frequency vibrations caused by the harmonic reducer without reducing the phase margin of the control system.

[0025] See Figure 1 The vibration suppression method for harmonic reducers based on hybrid parameter identification provided in this application specifically includes the following steps: Step S101: Construct a vibration compensation model for the harmonic reducer based on its physical characteristics.

[0026] Specifically, during motor operation, the harmonic reducer will generate resistance to the motor, thereby affecting the motor's control performance. Figure 2 This diagram illustrates the resistance fluctuation characteristics of a harmonic reducer. It shows in detail how, under the uniform motion trajectory of the CSP, the motor (the wave generator of the harmonic reducer) experiences resistance from two sinusoidal waves during one rotation. This resistance causes fluctuations in the actual speed, resulting in tracking errors. The tracking errors of each joint are superimposed on the end effector of the robotic arm, manifesting as end-effector vibration.

[0027] To compensate for the aforementioned resistance, a vibration compensation model for the harmonic reducer is established based on the physical characteristics of the harmonic reducer, relating the resistance to the motor position and load torque, in order to improve the motor's control performance.

[0028] The vibration compensation model for the harmonic reducer established in this application is as follows: ; in, This corresponds to the second harmonic, and the unit is cycles per revolution. The motor position is in radians. This is the phase angle, in radians; Let be the amplitude as a function of the load, and , and For amplitude parameters; This is the feedback torque of the motor.

[0029] Step S102: Obtain the static parameters of the harmonic reducer vibration compensation model through offline identification, and the static parameters include amplitude parameters.

[0030] Specifically, this application divides the parameters of the harmonic reducer vibration compensation model into static parameters and dynamic parameters. The static parameters are obtained through an offline identification strategy, i.e., vibration data under different loads are obtained through systematic experiments, and amplitude-related parameters are obtained by fitting using the least squares method. See [link to relevant documentation]. Figure 3 The specific process is as follows: Step S102a: Data Sampling. First, define the sampling trajectory of the robot target joint. Then, control the flexible robot target joint to run according to the preset sampling trajectory and collect motor feedback torque and motor position data.

[0031] It should be noted that the motor is controlled to rotate at a constant speed for multiple revolutions during sampling to ensure that the complete harmonic cycle signal is acquired. Simultaneously, the sampling period must cover different loads to ensure an accurate amplitude-load relationship is established.

[0032] Step S102b: Data Grouping and Parameter Identification. The collected data is divided into n groups based on the magnitude of the motor feedback torque. For each group of data, the model is... Perform least-squares fitting to obtain the amplitude under the corresponding load. and phase angle And calculate the arithmetic mean of the feedback torque for each group of motors. , .

[0033] The process of least squares fitting is as follows: First, regarding the model Linearization is performed to obtain ,in, , The length of the sampled data, This is the difference between the actual torque and the theoretical torque of the motor. The theoretical torque is obtained by calculating the sum of the gravitational torque and the frictional torque, which will not be elaborated here. Let be the regression matrix, and ; Let be the parameter vector to be identified, and Substitute the sampled data into the formula The parameter vector to be identified can then be obtained.

[0034] Then, according to the formula and Calculate amplitude and phase angle .in, and They are parameter vectors The first and second elements in the array.

[0035] Step S102c: Establish the amplitude-load model. Based on n sets of amplitudes The arithmetic mean of the motor feedback torque The linear relationship is fitted using the least squares method: The amplitude parameters of the amplitude-load model are obtained. and This will not be elaborated upon further here.

[0036] Step S102d: Determine the initial value of the phase angle. For n sets of phase angles... The arithmetic mean is taken to obtain the initial phase angle value for online RLS identification. .

[0037] Step S103: Obtain the dynamic parameters of the harmonic reducer vibration compensation model through online identification, and the dynamic parameters include the phase angle.

[0038] Specifically, during the operation of the robot joint drive system, the phase angle is updated in real time using the recursive least squares (RLS) method. That is, first establish the following discretized observation model: .

[0039] in, , This indicates the k-th control cycle. This is the feedback torque of the motor. The torque calculated for the full dynamic model includes inertial force, centrifugal force & Coriolis force and frictional force, which will not be elaborated here; This is the regression matrix; Let be the parameter vector to be identified, with initial values ​​obtained from offline identification. ; The noise of the motor is Gaussian white noise with variance of . .

[0040] In the above discretized observation model, , as well as As a known quantity The variable is unknown. The least squares recursive algorithm is used as follows: ,in, This represents the identification result from the previous control cycle. The gain matrix for the current control cycle is calculated as follows: ,in, The covariance matrix of the estimation error calculated for the previous control cycle is given below. The covariance matrix of the estimation error for the current control cycle is calculated as follows: ,in, For a unit matrix, the initial value is .

[0041] The parameter vector to be identified in the current control cycle is calculated based on the above. Then, you can use the formula The current phase angle is calculated.

[0042] Therefore, the specific process for this step is as follows: Step S103a: Initialize the parameters of the recursive least squares algorithm, including the initial values ​​of the parameter vector to be identified. Error covariance matrix and the noise variance of the Gaussian white noise of the motor .

[0043] Step S103b: Determine if the current period is the first control cycle. If so, then set... , ,and , If not, skip this step.

[0044] Step S103c: Obtain the motor position in the current control cycle. and motor feedback torque .

[0045] Step S103d: Calculate the motor torque using a full dynamics model. And calculate the difference between it and the motor feedback torque. , .

[0046] Step S103e: Calculate the amplitude at the current moment based on the amplitude-load model. .

[0047] Step S103f: Calculate the regression matrix ,and ; Step S103g: Calculate the gain matrix for the current control cycle. ,and ,in, R is the covariance matrix of the estimation error calculated in the previous control cycle, and R is the noise variance of the Gaussian white noise of the motor.

[0048] Step S103h: Calculate the parameter vector to be identified. ,and ,in, This represents the parameter vector identification result of the previous control cycle; Step S103i: Through the formula Calculate the current phase angle .

[0049] Step S104: Calculate the feedforward compensation torque of the motor based on the vibration compensation model of the harmonic reducer, and superimpose the feedforward compensation torque into the motor torque command to output the final motor control signal.

[0050] Specifically, based on the current phase angle Calculate feedforward compensation torque , Then, the feedforward compensation torque will be... Motor torque calculated with the full dynamics model Superimposed, the target feedforward torque is obtained. ,Right now It outputs the final motor control signal to counteract harmonic resistance and suppress vibration.

[0051] It should be noted that, after obtaining the target feedforward torque After sending the data to the servo motor, the process returns to step S103 c above and waits to jump to the next control cycle to achieve precise control in each control cycle and ensure the accuracy of the compensation phase.

[0052] Figure 4 This is a flowchart illustrating steps S103 and S104 above, specifically a flowchart illustrating online identification and feedforward compensation.

[0053] Figure 5 This is a schematic diagram illustrating the compensation effect of the amplitude-load model using experimental data. To visually demonstrate the effect, the motor torque data has been processed offline to remove bias components and then converted to the output. Figure 4 See the enlarged schematic diagram. Figure 6 .

[0054] Figure 7 This is a comparison chart showing the compensation effect of RLS online phase identification using experimental data. To visually demonstrate the effect, the motor torque data has been processed offline to remove bias components and then converted to the output. In the chart, the solid line represents the compensation result of RLS online phase identification, while the dashed line represents the compensation effect using only offline identification.

[0055] Figure 8 The image shows a comparison of vibration suppression effects using experimental data. In the image (the right image is a magnified view of a portion of the left image), the gray solid line represents the speed tracking error (calculated from the motor side to the output) when the collaborative robot's two-axis joints move at a speed of 40° / s. The black solid line represents the speed tracking error after compensation using the technical solution of this application under the same working conditions. It can be seen that the vibration amplitude is reduced by approximately 40%.

[0056] The above technical solution will be further explained below using a single joint of a collaborative robot as an example.

[0057] Implementation environment: The second joint drive system of the six-axis 5kg load collaborative robot has a built-in harmonic reducer with a reduction ratio of 1:121. The host computer is used to calculate and identify the feedforward model online, and the control frequency is 1kHz. The desired position, feedforward speed and feedforward torque calculated by the host computer are sent to the servo at a frequency of 1kHz, and the servo performs closed-loop control at a frequency of 4kHz.

[0058] Offline identification stage: Step 1: Under laboratory conditions, control the two joints to move at a constant speed of 60° / s from -90° to 90° and collect information on the motor position and motor torque (converted to the output end). During this period, the motor load is affected by gravity torque and friction torque, changing from a minimum of -137Nm to 0Nm and then to a maximum of +90Nm. The first half is used for identification, and the second half is used for verification.

[0059] Step 2: Select data on the motor torque changing from -137Nm to 0Nm, divide it into nine groups, and perform least-squares fitting on each group to obtain nine sets of parameters: amplitude and phase angle And calculate the arithmetic mean of the feedback torque for each group of motors. ,in .

[0060] Step 3: Based on nine sets of amplitude values and By fitting a linear relationship using the least squares method, the parameters of the amplitude-load model are obtained. , .

[0061] Step 4: Finally, for the nine sets of phase angles The arithmetic mean is taken to obtain the initial phase angle value for online RLS identification. The parameters are then permanently stored in the non-volatile memory of the control system.

[0062] Online identification and compensation stage: Step 1: Initialize the parameters of the Recursive Least Squares (RLS) method, i.e., use the initial phase angle value. Calculate the initial values ​​of the parameter vector to be identified. Error covariance matrix The noise variance of Gaussian white noise in motors .

[0063] Step 2: Determine if this is the first control cycle. If so, then... , If not, skip this assignment step.

[0064] Step 3: Obtain the motor position in the current control cycle and motor feedback torque .

[0065] Step 4: Calculate the motor torque using a full dynamics model. And calculate the difference between it and the motor feedback torque. , .

[0066] Step 5: Calculate the amplitude at the current moment based on the amplitude-load model. .

[0067] Step 6: Calculate the regression matrix .

[0068] Step 7: Calculate the gain matrix for the current control cycle. .

[0069] Step 8: Calculate the parameter vector to be identified ,and And calculate the phase angle of the current cycle in sequence. .

[0070] Step 9: Calculate the covariance matrix of the current control cycle estimation error. .

[0071] Step 10: Calculate the covariance matrix of the estimation error from the previous control cycle. Updated to The parameter vector identification results of the previous control cycle Updated to .

[0072] Step 11: Use phase angle to identify the results Calculate feedforward compensation torque and feedforward compensation torque Motor torque calculated with the full dynamics model Superimposed, the target feedforward torque is obtained. And send it to the servo motor.

[0073] Step 12: Return to Step 3 and wait to jump to the next control cycle.

[0074] This application also provides a harmonic reducer vibration suppression system based on hybrid parameter identification. See also Figure 9 The system includes: a model building module 201, an offline identification module 202, an online identification module 203, and a torque compensation module 204.

[0075] Specifically, the model building module 201 is used to construct a vibration compensation model for the harmonic reducer based on its physical characteristics. The constructed vibration compensation model for the harmonic reducer is as follows: In the formula, This corresponds to the second harmonic, and the unit is cycles per revolution. The motor position is in radians. This is the phase angle, in radians; Let be the amplitude as a function of the load, and , and For amplitude parameters; This is the feedback torque of the motor.

[0076] The offline identification module is used to obtain the static parameters of the vibration compensation model of the harmonic reducer through offline identification. Specifically, the offline identification module obtains vibration data under different loads through systematic experiments and uses the least squares method to fit and obtain the amplitude parameters. and .

[0077] In addition, the aforementioned harmonic reducer vibration suppression system based on hybrid parameter identification also includes a parameter storage module, which is used to store the amplitude parameters identified by the offline identification module 202. and And the initial parameters R and required by the online identification module .

[0078] The online identification module 203 is used to obtain the dynamic parameters of the harmonic reducer vibration compensation model through online identification. Specifically, the online identification module is used to update the phase angle in real time using the recursive least squares (RLS) method during the operation of the robot joint drive system. .

[0079] The torque compensation module 204 is used to calculate the feedforward compensation torque of the motor based on the vibration compensation model of the harmonic reducer, and to superimpose the feedforward compensation torque into the motor torque command to output the final motor control signal.

[0080] Therefore, the technical solution of this application establishes a mathematical model of the second harmonic resistance of the harmonic reducer, adds feedforward to the real-time control system to compensate for the second harmonic resistance, and determines the amplitude-related parameters through offline identification and updates the phase parameters online, thereby achieving the best balance between computational efficiency and adaptability. Without reducing the phase margin of the control system, it effectively suppresses the high-frequency vibration caused by the harmonic reducer.

[0081] In addition to the above embodiments, the present invention may have other implementation methods; all technical solutions formed by equivalent substitution or equivalent transformation fall within the protection scope claimed by the present invention.

Claims

1. A method for suppressing vibration of a harmonic reducer based on hybrid parameter identification, characterized in that: include: A vibration compensation model for harmonic reducers is constructed based on their physical characteristics. The static parameters of the vibration compensation model of the harmonic reducer are obtained by offline identification, and the static parameters include amplitude parameters. The dynamic parameters of the vibration compensation model of the harmonic reducer are obtained by online identification, and the dynamic parameters include the phase angle; The feedforward compensation torque of the motor is calculated based on the vibration compensation model of the harmonic reducer, and the feedforward compensation torque is superimposed on the motor torque command to output the final motor control signal.

2. The method for suppressing vibration of a harmonic reducer based on hybrid parameter identification according to claim 1, characterized in that: The vibration compensation model for the harmonic reducer is as follows: ; in, This corresponds to the second harmonic. Position of the motor; Phase angle; Let be the amplitude as a function of the load, and , and For amplitude parameters; This is the feedback torque of the motor.

3. The method for suppressing vibration of a harmonic reducer based on hybrid parameter identification according to claim 2, characterized in that: The process of obtaining the static parameters of the harmonic reducer vibration compensation model through offline identification includes: Control the target joint of the robot to move along a preset trajectory, and collect the feedback torque and position data of the motor; The collected data were divided into n groups according to the magnitude of the motor feedback torque. For each group of data, the model was... Perform least-squares fitting to obtain the amplitude under the corresponding load. and phase angle And calculate the arithmetic mean of the feedback torque for each group of motors. ; Based on n sets of amplitudes The arithmetic mean of the motor feedback torque By fitting a linear relationship using the least squares method, the amplitude parameters of the amplitude-load model are obtained. and ; For n sets of phase angles The arithmetic mean is taken to obtain the initial phase angle value used for online identification. .

4. The method for suppressing vibration of a harmonic reducer based on hybrid parameter identification according to claim 3, characterized in that: The model Perform least-squares fitting to obtain the amplitude under the corresponding load. and phase angle include: For the model Linearization is performed to obtain ,in, , The length of the sampled data, This is the difference between the actual torque and the theoretical torque of the motor. Let be the regression matrix, and , Let be the parameter vector to be identified, and ; According to the formula and Calculate amplitude and phase angle .

5. The method for suppressing vibration of a harmonic reducer based on hybrid parameter identification according to claim 4, characterized in that: The process of obtaining the dynamic parameters of the harmonic reducer vibration compensation model through online identification includes: Get the motor position in the current control cycle and motor feedback torque ; Motor torque was calculated using a full dynamics model. And calculate the difference between it and the motor feedback torque. , ; Calculate the amplitude at the current moment based on the amplitude-load model. ; Calculate the regression matrix ,and ; Calculate the gain matrix for the current control cycle. ,and ,in, R is the covariance matrix of the estimation error calculated in the previous control cycle, and R is the noise variance of the Gaussian white noise of the motor. Calculate the parameter vector to be identified ,and ,in, This represents the parameter vector identification result of the previous control cycle; Through formula Calculate the current phase angle .

6. The method for suppressing vibration of a harmonic reducer based on hybrid parameter identification according to claim 5, characterized in that: The motor position in the current control cycle is obtained. and motor feedback torque Previously, it also included: Determine if the current state is the first control cycle; if so, then... , ,and , If not, skip this step.

7. The method for suppressing vibration of a harmonic reducer based on hybrid parameter identification according to claim 6, characterized in that: The feedforward compensation torque of the motor is calculated based on the vibration compensation model of the harmonic reducer, and the feedforward compensation torque is superimposed on the motor torque command. The final motor control signal output includes: Based on the current phase angle Calculate feedforward compensation torque ,and ; feedforward compensation torque Motor torque calculated with the full dynamics model Superimposed, the target feedforward torque is obtained. ,Right now It then outputs the final motor control signal.

8. The method for suppressing vibration of a harmonic reducer based on hybrid parameter identification according to claim 7, characterized in that: The formula Calculate the current phase angle Following that, it also includes: Through formula Calculate the covariance matrix of the current control cycle estimation error, where, For an identity matrix, the initial value is ; The covariance matrix of the estimation error calculated in the previous control cycle. Updated to The parameter vector identification results of the previous control cycle Updated to .

9. The method for suppressing vibration of a harmonic reducer based on hybrid parameter identification according to claim 8, characterized in that: After calculating the feedforward compensation torque of the motor based on the vibration compensation model of the harmonic reducer, and superimposing the feedforward compensation torque into the motor torque command to output the final motor control signal, the process further includes: Wait for the next control cycle to start, and after jumping to the next control cycle, obtain the motor position of the current control cycle again. and motor feedback torque Then, proceed with the subsequent steps in sequence.

10. A harmonic reducer vibration suppression system based on hybrid parameter identification, characterized in that: include: The model building module is used to build a vibration compensation model for harmonic reducers based on their physical characteristics. The offline identification module is used to obtain the static parameters of the vibration compensation model of the harmonic reducer through offline identification, and the static parameters include amplitude parameters. An online identification module is used to obtain the dynamic parameters of the harmonic reducer vibration compensation model through online identification, and the dynamic parameters include the phase angle; The torque compensation module is used to calculate the feedforward compensation torque of the motor based on the vibration compensation model of the harmonic reducer, and to superimpose the feedforward compensation torque into the motor torque command to output the final motor control signal.

Images