A Multi-Degree-of-Freedom Cooperative Control Method for Industrial Robots

By extracting three-phase current and encoder signals from the servo driver and reconstructing the dynamic model, the problems of interference and delay of external sensors are solved, enabling real-time degradation monitoring and compensation of multi-degree-of-freedom industrial robots, and improving trajectory accuracy and robustness.

CN123103676APending Publication Date: 2026-11-17NANJING HANGLEI MECHANICAL & ELECTRICAL EQUIPMENT CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611536049.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-30
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies for multi-degree-of-freedom industrial robots, external vision or six-dimensional torque sensors are easily affected by industrial site interference, resulting in unstable detection accuracy, large response delay, and a disconnect between the monitoring system and the control system, making it impossible to compensate for end-effector trajectory drift caused by joint degeneration in real time.

Method used

By acquiring three-phase current and encoder signals inside the servo driver, extracting joint degradation characteristics, reconstructing the dynamic model, and calculating joint compensation, collaborative torque control within the underlying control cycle is achieved, eliminating dependence on and delay of external sensors.

Benefits of technology

It improves the accuracy and robustness of the robot's end effector trajectory without increasing hardware costs, eliminates signal processing delays, and forms a closed loop of degradation monitoring and torque control, enabling targeted compensation for joint degradation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123103676A_ABST
    Figure CN123103676A_ABST
Patent Text Reader

Abstract

This invention discloses a multi-degree-of-freedom collaborative control method for industrial robots. The method, based on the three-phase current and encoder signals within the servo drivers of each joint, extracts a degradation state feature matrix containing abrupt changes in friction, increased transmission backlash, and decreased joint stiffness. This matrix is ​​then mapped to a time-varying degradation factor. The transient dynamic model is reconstructed online within the underlying control cycle. Degraded joints are identified based on the degradation factor. The asymmetric angle compensation amount of healthy joints spatially coupled with the degraded joints is obtained through Jacobian matrix redundancy calculation. Combined with the transient dynamic model, torque commands for each joint are generated, achieving online cancellation of robot end-effector trajectory drift without the need for external vision and force sensors.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of robot control technology, specifically to a multi-degree-of-freedom collaborative control method for industrial robots. Background Technology

[0002] Multi-degree-of-freedom industrial robots typically consist of n rotary joints connected in series. Each joint is equipped with a servo motor, reducer, and high-precision encoder. Motion control is achieved by receiving torque or position commands through the underlying servo driver. The overall dynamic characteristics of the robot can be described by the Newton-Euler equation or the Lagrange equation, that is, there is a definite nonlinear mapping relationship between joint torque and joint position, velocity, and acceleration. The parameters such as inertia matrix, Coriolis force matrix, gravity term, and friction term in this mapping relationship are determined offline when the robot leaves the factory and are fixed in the controller, forming the model basis for calculating torque control. In addition, the Jacobian matrix in robot kinematics describes the linear mapping relationship between joint space velocity and end effector Cartesian space velocity, which is the basic tool for realizing joint redundancy allocation and coordinated control.

[0003] After long-term high-load operation, the joints will experience hidden physical degradation due to wear, fatigue, and other factors, such as sudden changes in friction, increased transmission backlash, and decreased joint stiffness. This leads to a deviation between the factory-fixed dynamic model and the robot's current actual physical state, which in turn causes the end effector trajectory to drift. To compensate for this problem, the industry usually adds a machine vision system or a six-dimensional torque sensor to the robot's end effector to collect the actual pose or force state of the end effector in real time and compare it with the expected trajectory. The overall trajectory command of the robot is corrected through external closed-loop feedback. The specific implementation steps of this solution are as follows: the external sensor acquires the pose or force information of the end effector at a fixed sampling frequency.

[0004] The vision or force processing unit performs image processing or signal decoding on the acquired data to obtain the end-point deviation.

[0005] This deviation is transmitted to the robot's main controller via the communication bus;

[0006] The main controller corrects the overall motion trajectory command based on the deviation and then reissues it. Structurally, this scheme is a combination of the robot body, external sensing devices, independent signal processing unit and external closed-loop control loop.

[0007] The above-mentioned existing technical solutions have the following drawbacks: Since the vision or force sensors are installed on the outside of the robot end effector, they are directly exposed to the interference environment of dust, oil, vibration and light changes in the industrial site. Their collected signals are easily distorted or blocked, and the detection accuracy is unstable.

[0008] Because external sensors and their signal processing units require additional image acquisition, feature matching, or filtering processes, the signal processing link has an inherent delay of tens of milliseconds, making it difficult to use for high-frequency torque closed-loop control of the robot's underlying motors.

[0009] External attachments increase the load mass and size of the end effector, reducing the robot's dynamic response performance and increasing system costs.

[0010] Because the existing closed loop only corrects the overall trajectory command and does not delve into the physical level of joint degradation, it cannot distinguish which joint and which physical degradation mechanism caused the error, resulting in a lack of targeted compensation. Furthermore, the existing offline predictive maintenance system can only provide degradation warnings and cannot convert this information into torque coordination scheduling commands within the control cycle in real time, resulting in a disconnect between the monitoring system and the control system. Summary of the Invention

[0011] The technical problem this invention aims to solve is: how to perceive the latent physical degradation of each joint of a multi-degree-of-freedom industrial robot caused by long-term high-load operation in real time without adding external vision, six-dimensional torque, or other physical sensors, and to convert this degradation information into a collaborative torque compensation command for healthy joints within the underlying control cycle, thereby offsetting the end effector trajectory drift caused by degraded joints. This overcomes the problems of high cost, susceptibility to industrial site interference, large response delay, and disconnect between the monitoring and control systems in existing external sensor solutions.

[0012] To address the aforementioned technical problems, this invention provides a multi-degree-of-freedom cooperative control method for an industrial robot. The industrial robot comprises n serial joints, each joint equipped with a servo motor, a reducer, an encoder, and a servo driver, where n is a positive integer greater than 1. The method includes:

[0013] The three-phase current signal inside the servo driver of each joint and the joint position signal collected by the encoder are synchronously acquired at a fixed control cycle. The three-phase current signal is transformed by coordinate to obtain the measured equivalent torque current. The theoretical current is calculated based on the rigid body dynamic model parameters calibrated by the robot at the factory and the joint position signal. The measured equivalent torque current is subtracted from the theoretical current to obtain the current residual. The current residual is filtered. Based on the filtered current residual, the friction force change characteristics, transmission backlash characteristics and joint stiffness decay characteristics of each joint are extracted and arranged according to the joint dimension to form a degradation state feature matrix.

[0014] The estimated value of the time-varying degradation factor is obtained from the degradation state feature matrix. The estimated value of the time-varying degradation factor is substituted into the Newton-Euler recursive formula to reconstruct a transient dynamic model containing transient inertia matrix, transient Coriolis force matrix, transient gravity term and transient friction term.

[0015] The time-varying degradation factor estimate is compared with a preset degradation threshold. Joints with degradation factor estimates exceeding the corresponding preset degradation threshold are identified as degenerate joints. Based on the transmission backlash characteristic component and joint stiffness attenuation characteristic component in the degradation factor estimate of the degenerate joint, the joint position hysteresis of the degenerate joint is estimated. The global Jacobian matrix of the robot in the current joint pose is calculated, and the column vector corresponding to the degenerate joint is extracted. The spatial pose deviation of the end effector caused by the degenerate joint is calculated accordingly. The set of healthy joints that have a spatial coupling relationship with the degenerate joint is determined, and the sub-Jacobi matrix corresponding to this set is extracted. The joint angle compensation of the healthy joint set is calculated accordingly.

[0016] The joint angle compensation is superimposed on the desired trajectory command of the healthy joint set. Combined with the transient dynamic model, the torque command of each joint is generated according to the adaptive torque control law, and the torque command is sent to the servo driver of the corresponding joint for execution.

[0017] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows:

[0018] 1. This invention can extract multi-degree-of-freedom physical degradation features by relying only on the existing current and encoder data inside the servo driver, without the need to install external vision or force sensors, thus reducing hardware costs and improving robustness as it is not affected by factors such as dust and obstruction in industrial environments.

[0019] 2. This invention completes degradation feature extraction, dynamic model reconstruction, and torque compensation within the underlying control cycle, eliminating the inherent signal processing delay of the external sensing link. This allows degradation monitoring results to directly participate in high-frequency torque closed-loop control. At the same time, the new current and position feedback generated after the torque command is executed re-enter the degradation feature extraction stage, forming a closed-loop operation mode of degradation monitoring and torque control.

[0020] 3. This invention uses Jacobi matrix redundancy calculation to perform asymmetric spatial pre-compensation on a set of healthy joints for specific degenerate joints. It can distinguish three different physical degradation mechanisms—friction, backlash, and stiffness—according to the joint and provide targeted compensation amounts, which is beneficial to improving the robot's ability to maintain end-effector trajectory accuracy under long-term high-load operation. Attached Figure Description

[0021] Figure 1 This is a closed-loop flowchart of the overall system in an embodiment of the present invention;

[0022] Figure 2 This is a flowchart of the degradation state feature extraction process in an embodiment of the present invention;

[0023] Figure 3 This is a flowchart of the online adaptive reconstruction process of the dynamic model in an embodiment of the present invention. Detailed Implementation

[0024] The following is in conjunction with the appendix Figure 1-3 The specific embodiments of the present invention will be further described below. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0025] The present invention will be further described below with reference to specific embodiments, but the scope of protection of the present invention is not limited to the specific embodiments.

[0026] In this embodiment, the robot body of the industrial robot consists of n serial joints, where n is a positive integer greater than 1. Each joint is sequentially denoted as joint 11, joint 12, up to joint 1n. Each joint is equipped with a servo motor, a reducer, and a high-precision encoder, and corresponds to a servo driver. The servo driver integrates a three-phase current sampling unit and an encoder signal acquisition unit. The three-phase current sampling unit is used to acquire the three-phase current signal of the corresponding joint, and the encoder signal acquisition unit is used to acquire the joint position signal of the corresponding joint. The servo drivers of each joint are connected to the main controller through a fieldbus. The fieldbus can be an EtherCAT bus to achieve high-frequency data synchronous transmission in the frequency range of 1kHz to 4kHz. The main controller is sequentially divided into a degradation state feature extraction module, a dynamic model online adaptive reconstruction module, an asymmetric multi-degree-of-freedom cooperative torque distribution module, and an adaptive torque calculation controller. The modules are sequentially connected by a data stream to form a complete processing link from servo feedback data input to torque command output.

[0027] After the adaptive torque controller sends torque commands to each joint servo driver for execution, the resulting new three-phase current signals and new joint position signals re-enter the degradation state feature extraction module, forming a closed-loop operation structure of degradation monitoring and torque control.

[0028] I. Processing procedure of the degradation state feature extraction module

[0029] The degradation state feature extraction module synchronously acquires the three-phase current signal inside each joint servo driver and the joint rotor position signal acquired by the encoder signal acquisition unit at a fixed control cycle, and estimates the joint speed through differential method or observer method.

[0030] The Clarke and Park transforms are performed on the three-phase current signal to obtain the measured equivalent torque current, which is linearly related to the joint output torque, denoted as . ,in For joint serial numbers, For a moment.

[0031] Based on the rigid body dynamics model parameters calibrated at the robot's factory, including the inertia matrix, Coriolis force matrix, and gravity term, and combined with the currently acquired or estimated joint positions, velocities, and accelerations, the first... The theoretical joint torque is calculated, and the theoretical current is obtained by converting it using the motor torque constant:

[0032] ;

[0033] In the formula, For the first joints in The theoretical current at time t;

[0034] The first calculation based on the factory rigid body dynamics model Joint theory: joint torque;

[0035] For the first The torque constant of the joint servo motor is a fixed parameter obtained from factory calibration.

[0036] Subtracting the measured equivalent torque current from the theoretical current yields the current residual:

[0037] ;

[0038] In the formula, For the first joints in The current residual at time;

[0039] This is the measured equivalent torque current;

[0040] Given the theoretical current, the current residual is subjected to low-pass filtering to remove high-frequency noise and sampling interference, resulting in the filtered current residual.

[0041] Based on the filtered current residual, three types of degradation features are extracted: First, frictional abrupt change features are extracted based on the overall amplitude increase of the filtered current residual in the preset low-speed operating range, which is used to reflect the abnormal increase of joint friction.

[0042] Secondly, the transmission backlash characteristics are estimated based on the hysteresis width formed between the current signal and the joint position signal at the moment when the joint movement direction reverses, which is used to reflect the degree of increase in the transmission mechanism backlash.

[0043] Third, joint stiffness attenuation characteristics are estimated based on the slope between the load torque on the joint and the joint deformation angle, reflecting the degree of joint stiffness attenuation. The three types of features extracted from each joint are arranged according to the joint dimension to form an n-row, three-column multi-degree-of-freedom degradation state feature matrix. The rows correspond to the first The characteristics of joint frictional abrupt change, transmission backlash, and joint stiffness decay are analyzed, and this degradation state feature matrix is ​​output to the online adaptive reconstruction module of the dynamic model.

[0044] II. Processing Procedure of the Online Adaptive Reconstruction Module for the Dynamic Model

[0045] Define a time-varying degradation factor vector, where each component corresponds to the first... The linear combination coefficient of the deviations of physical parameters such as the joint equivalent rotational inertia deviation and the flexible deformation coefficient deviation is set to zero for each component, indicating that it is completely consistent with the factory dynamic model.

[0046] The online adaptive reconstruction module of the dynamic model maps the degradation state feature matrix into degradation factor increments using a mapping function obtained from pre-identified offline methods. This mapping function can be implemented using a multinomial regression model. Furthermore, an adaptive update law designed based on the Lyapunov stability criterion is used, with the current residual as the driving signal, to perform online iterative updates of the time-varying degradation factor estimates for each joint.

[0047] ;

[0048] In the formula, For the first Rate of change of estimated joint time-varying degeneration factors;

[0049] The positive definite adaptive gain matrix is ​​used to adjust the convergence speed. This matrix is ​​preset by those skilled in the art according to the actual convergence requirements.

[0050] For the first The regression matrix obtained by linearly parameterizing the degradation factor in the joint dynamics model. , , These are joint position, joint velocity, and joint acceleration, respectively.

[0051] The current residual obtained above.

[0052] The updated time-varying degradation factor estimates for each joint are substituted into the Newton-Euler recursive formula to reconstruct the transient inertia matrix, transient Coriolis force matrix, transient gravity term, and transient friction term. These four components together constitute a transient dynamic model that fits the robot's current real physical state. This transient dynamic model is fed back to the adaptive computational torque controller in real time according to the underlying control sub-cycle (250 microseconds in this embodiment), serving as the model basis for the next round of torque command calculation. Together with the aforementioned feature extraction and degradation factor update steps, it forms a continuously iterative closed loop.

[0053] In another alternative implementation, instead of using the adaptive update law based on the Lyapunov stability criterion, the time-varying degradation factor can be used as a state variable and the current residual as an observation. The time-varying degradation factor can be estimated and updated online through Kalman filtering recursive operation, which can also achieve online reconstruction of the transient dynamic model.

[0054] III. Processing Procedure of the Asymmetric Multi-Degree-of-Freedom Cooperative Torque Distribution Module

[0055] The asymmetric multi-degree-of-freedom collaborative torque distribution module compares the estimated time-varying degradation factor of each joint with the corresponding preset degradation threshold one by one. If the estimated time-varying degradation factor of a certain joint exceeds its corresponding preset degradation threshold, the joint is determined to be a degenerate joint. Based on the transmission backlash characteristic component and the joint stiffness attenuation characteristic component in the estimated time-varying degradation factor of the degenerate joint, the joint position hysteresis of the degenerate joint is estimated.

[0056] Calculate the robot's global Jacobian matrix at the current joint pose, extract the column vectors corresponding to the degenerate joints from the global Jacobian matrix, and calculate the spatial pose deviation of the end effector caused by the degenerate joints based on the product of these column vectors and the joint position hysteresis.

[0057] ;

[0058] In the formula, Let be the pose deviation vector of the end effector in Cartesian space;

[0059] For the corresponding degenerate joint in the global Jacobian matrix Column vectors;

[0060] For degenerated joints The amount of joint position lag.

[0061] Based on the robot's kinematic topology, joints adjacent to degenerate joints and jointly influencing the spatial pose of the end effector are identified as the set of healthy joints with spatial coupling to the degenerate joints. A sub-Jacobi matrix is ​​extracted from the overall Jacobian matrix, consisting of the columns corresponding to this healthy joint set. The Moore-Ponros pseudoinverse of this sub-Jacobi matrix is ​​calculated, and the end effector spatial pose deviation is multiplied by this pseudoinverse to obtain the required asymmetric spatial pose pre-compensation for the healthy joint set.

[0062] ;

[0063] In the formula, For the set of healthy joints, there is a vector of joint angle compensation.

[0064] For the sub-Jacobi matrix corresponding to the set of healthy joints;

[0065] Transpose it;

[0066] This refers to the inverse term in pseudo-inverse operations;

[0067] Spatial pose deviation of the end effector caused by degenerated joints.

[0068] In another alternative implementation, the compensation calculation method based on the pseudo-inverse of the sub-Jacobi matrix can be abandoned. Instead, a multi-joint collaborative optimization method based on model predictive control can be used. In each control cycle, the optimization objective is to reduce the trajectory tracking error of the end effector. The torque increment and angle increment of the healthy joint set are used as optimization variables. Constraints are constructed by combining the transient dynamics model and the joint position hysteresis of the degenerated joints. The joint angle compensation amount of the healthy joint set is obtained by solving the rolling time domain optimization. This can also achieve collaborative cancellation of trajectory drift caused by degenerated joints.

[0069] IV. Processing Procedure of Adaptive Calculation Torque Controller

[0070] The adaptive computational torque controller superimposes the joint angle compensation calculated by the asymmetric multi-degree-of-freedom collaborative torque distribution module onto the desired trajectory command of the healthy joint set. Combined with the transient dynamic model output by the online adaptive reconstruction module of the dynamic model, it generates torque commands for each joint according to the adaptive computational torque control law. Specifically, the torque command is obtained by multiplying the transient inertia matrix with the desired joint acceleration, the compensation acceleration obtained by second differentiation of the compensation angle, the velocity loop feedback error, and the position loop feedback error as a feedforward and feedback synthesis term. Then, it is superimposed with the product of the transient Coriolis force matrix and the joint velocity, the transient gravity term, and the transient friction term. The velocity loop gain matrix and the position loop gain matrix are preset by those skilled in the art according to the dynamic response requirements of the robot. The torque command is sent to the servo driver corresponding to each joint via the fieldbus for execution, driving the servo motor of each joint to move and realize the online cancellation of the trajectory drift of the end effector.

[0071] After each joint servo driver executes the torque command, its internal three-phase current sampling unit and encoder signal acquisition unit acquire new three-phase current signals and new joint position signals. These new signals re-enter the degradation state feature extraction module, and together with the dynamic model online adaptive reconstruction module, the asymmetric multi-degree-of-freedom collaborative torque distribution module, and the adaptive calculation torque controller, they continuously cycle according to the aforementioned control cycle, forming an overall control process in which degradation monitoring and torque control are mutually closed.

[0072] The above embodiments show that the method of the present invention can extract multi-degree-of-freedom physical degradation features and cooperative torque compensation by relying only on the existing current and encoder data inside the servo driver, without the need to install external vision or force sensors, and all related processing is completed within the underlying control cycle, which can meet the requirements of industrial robots for real-time performance and robustness.

[0073] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. A multi-degree-of-freedom cooperative control method for an industrial robot, the industrial robot comprising n serial joints, each joint being equipped with a servo motor, a reducer, an encoder, and a servo driver, where n is a positive integer greater than 1, characterized in that, The method includes: The three-phase current signal inside the servo driver of each joint and the joint position signal collected by the encoder are synchronously acquired at a fixed control cycle. The three-phase current signal is transformed by coordinate to obtain the measured equivalent torque current. The theoretical current is calculated based on the rigid body dynamic model parameters calibrated by the robot at the factory and the joint position signal. The measured equivalent torque current is subtracted from the theoretical current to obtain the current residual. The current residual is filtered. Based on the filtered current residual, the frictional force change characteristics, transmission backlash characteristics and joint stiffness decay characteristics of each joint are extracted and arranged according to the joint dimension to form a degradation state feature matrix. The estimated value of the time-varying degradation factor is obtained based on the degradation state feature matrix. The estimated value of the time-varying degradation factor is substituted into the Newton-Euler recursive formula to reconstruct a transient dynamic model containing transient inertia matrix, transient Coriolis force matrix, transient gravity term and transient friction term. The estimated value of the time-varying degradation factor is compared with a preset degradation threshold. Joints with degradation factor estimates exceeding the corresponding preset degradation threshold are identified as degraded joints. Based on the transmission backlash characteristic component and joint stiffness attenuation characteristic component in the degradation factor estimate of the degraded joint, the joint position hysteresis of the degraded joint is estimated. Calculate the overall Jacobian matrix of the robot in the current joint pose, extract the column vectors corresponding to the degenerate joints in the overall Jacobian matrix, and calculate the spatial pose deviation of the end effector caused by the degenerate joints based on the column vectors and the joint position hysteresis. A set of healthy joints that have a spatial coupling relationship with the degenerated joint is determined. A sub-Jacobi matrix corresponding to the set of healthy joints is extracted from the overall Jacobian matrix. Based on the sub-Jacobi matrix and the spatial pose deviation of the end effector, the joint angle compensation amount of the set of healthy joints is calculated. The joint angle compensation is superimposed on the desired trajectory command of the healthy joint set. Combined with the transient dynamic model, torque commands for each joint are generated according to the adaptive torque control law, and the torque commands are sent to the servo drivers of the corresponding joints for execution.

2. The method according to claim 1, characterized in that, The extraction of frictional force abrupt change characteristics, transmission backlash characteristics, and joint stiffness attenuation characteristics of each joint includes: extracting the frictional force abrupt change characteristics based on the degree of amplitude increase of the filtered current residual within a preset low-speed operating range; The transmission backlash feature is extracted based on the hysteresis width formed between the current signal and the joint position signal at the moment when the joint movement direction reverses. The joint stiffness attenuation feature is extracted based on the slope between the load torque and the joint deformation angle.

3. The method according to claim 1, characterized in that, The step of obtaining the time-varying degradation factor estimate based on the degradation state feature matrix includes: mapping the degradation state feature matrix into degradation factor increments using a mapping function obtained through pre-offline identification; An adaptive update law based on the Lyapunov stability criterion is adopted to perform online iterative updates on the estimated value of the time-varying degradation factor using the current residual as the driving signal. The adaptive update law includes a positive definite adaptive gain matrix to adjust the convergence speed of the estimated value of the time-varying degradation factor.

4. The method according to claim 1, characterized in that, The step of obtaining the estimated value of the time-varying degradation factor based on the degradation state feature matrix includes: using the estimated value of the time-varying degradation factor as a state variable, using the current residual as an observation, and performing online estimation and updating of the estimated value of the time-varying degradation factor through Kalman filtering recursive operation.

5. The method according to claim 1, characterized in that, The determination of the set of healthy joints that have a spatial coupling relationship with the degenerate joint includes: determining the set of healthy joints as the joints that are adjacent to the degenerate joint and jointly affect the spatial pose of the end effector, based on the robot's kinematic topology. The step of calculating the joint angle compensation amount of the healthy joint set based on the sub-Jacobi matrix and the end effector spatial pose deviation includes: calculating the Moore-Penrose pseudoinverse of the sub-Jacobi matrix, multiplying the end effector spatial pose deviation by the Moore-Penrose pseudoinverse to obtain the joint angle compensation amount of the healthy joint set.

6. The method according to claim 1, characterized in that, The step of calculating the joint angle compensation amount of the healthy joint set based on the sub-Jacobi matrix and the spatial pose deviation of the end effector includes: taking the reduction of the end effector trajectory tracking error as the optimization objective, using the torque increment and angle increment of the healthy joint set as optimization variables, constructing constraints by combining the transient dynamics model and the joint position hysteresis, and obtaining the joint angle compensation amount of the healthy joint set through rolling time-domain optimization.

7. The method according to claim 1, characterized in that, The method further includes: sending the torque command to the servo driver of the corresponding joint to generate a new three-phase current signal and a new joint position signal, and re-entering the step of synchronously acquiring the three-phase current signal inside the servo driver of each joint and the joint position signal acquired by the encoder at a fixed control cycle, forming a closed-loop control process in which degradation state feature extraction, dynamic model reconstruction and cooperative torque distribution are continuously cyclically executed according to the control cycle.

8. A multi-degree-of-freedom cooperative control system for an industrial robot, comprising a robot body consisting of n serial joints, each joint being equipped with a servo motor, a reducer, an encoder, and a servo driver, where n is a positive integer greater than 1, and each joint's servo driver is connected to a main controller via a fieldbus, characterized in that... The main controller includes: The degradation state feature extraction module is used to synchronously collect the three-phase current signals inside the servo driver of each joint and the joint position signals collected by the encoder at a fixed control cycle. The three-phase current signals are transformed by coordinate to obtain the measured equivalent torque current. The theoretical current is calculated based on the rigid body dynamics model parameters calibrated by the robot at the factory and the joint position signals. The measured equivalent torque current is subtracted from the theoretical current to obtain the current residual. The current residual is filtered. Based on the filtered current residual, the friction force change feature, transmission backlash feature and joint stiffness decay feature of each joint are extracted and arranged by joint dimension to form a degradation state feature matrix. The dynamic model online adaptive reconstruction module is connected to the degradation state feature extraction module. It is used to obtain the estimated value of the time-varying degradation factor based on the degradation state feature matrix, and substitute the estimated value of the time-varying degradation factor into the Newton-Euler recursive formula to reconstruct a transient dynamic model containing transient inertia matrix, transient Coriolis force matrix, transient gravity term and transient friction term. An asymmetric multi-degree-of-freedom collaborative torque allocation module, connected to the online adaptive reconstruction module of the dynamic model, is used to compare the estimated value of the time-varying degradation factor with a preset degradation threshold, determine that joints with degradation factor estimates exceeding the corresponding preset degradation threshold are degenerate joints, estimate the joint position hysteresis of the degenerate joint based on the degradation factor estimate of the degenerate joint, calculate the overall Jacobian matrix of the robot in the current joint pose, calculate the spatial pose deviation of the end effector based on the column vectors in the overall Jacobian matrix corresponding to the degenerate joint and the joint position hysteresis, determine the set of healthy joints that have a spatial coupling relationship with the degenerate joint, and calculate the joint angle compensation of the healthy joint set based on the sub-Jacobi matrix corresponding to the healthy joint set and the spatial pose deviation of the end effector. An adaptive computational torque controller, connected to the asymmetric multi-degree-of-freedom collaborative torque distribution module, is used to superimpose the joint angle compensation amount onto the desired trajectory command of the healthy joint set, combine the transient dynamic model, generate torque commands for each joint according to the adaptive computational torque control law, and send the torque commands to the servo drivers of the corresponding joints for execution.

9. The system according to claim 8, characterized in that, The fieldbus is an EtherCAT bus, used to realize synchronous data transmission between each joint servo driver and the main controller in the frequency range of 1kHz to 4kHz. The new three-phase current signal and new joint position signal generated after the adaptive torque controller issues the torque command are fed back to the degradation state feature extraction module, forming a closed-loop control structure in which the degradation state feature extraction module, the dynamic model online adaptive reconstruction module, the asymmetric multi-degree-of-freedom cooperative torque distribution module, and the adaptive torque controller continuously cycle according to the control period.

10. The system according to claim 8, characterized in that, Each joint's servo driver integrates a three-phase current sampling unit and an encoder signal acquisition unit. The three-phase current sampling unit is used to acquire the three-phase current signal of the corresponding joint and send it to the degradation state feature extraction module. The encoder signal acquisition unit is used to acquire the joint position signal of the corresponding joint and send it to the degradation state feature extraction module.