An industrial robot control system
By constructing a multi-source error online decoupling module and an adaptive sliding mode compensation control module, and utilizing existing servo motor and encoder signals for error observation and real-time compensation, the problem of time-varying attenuation of absolute positioning accuracy of industrial robots is solved. This achieves online and real-time error decoupling and compensation, avoiding losses in hardware costs and production efficiency.
Patent Information
- Application Number
- CN202611250204.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies for improving the absolute positioning accuracy of industrial robots suffer from problems such as long calibration cycles, high hardware modification costs, end-effector load encroachment, and a lack of provable closed-loop stability in compensation algorithms. These make it difficult to achieve online, real-time, and non-stop decoupling identification and compensation of multi-source time-varying errors without adding external measurement hardware.
An industrial robot control system is adopted, including a host motion planning module, a multi-source error online decoupling module, and an adaptive sliding mode compensation control module. By reusing servo motor and encoder signals, a third-order extended state observer is constructed to observe errors, and the adaptive sliding mode control module is used for real-time compensation, thereby realizing online decoupling and compensation of multi-source errors.
It achieves online maintenance of robot absolute positioning accuracy without adding external hardware, reduces hardware modification costs, avoids the occupation of load space by external sensors, and has proven closed-loop stability, thereby improving production efficiency.
Smart Images

Figure CN123103632A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent manufacturing equipment control technology, specifically to an industrial robot control system. Background Technology
[0002] Industrial robots typically consist of joint servo motors, precision reducers, high-precision encoders, servo drivers, and a host motion controller. The controller drives the joint motors to move according to the planned joint trajectory through a cascaded servo control architecture of position loop, velocity loop, and current loop, thereby driving the end effector to reach the target pose. The positioning accuracy of a robot is usually divided into two categories: repeatability and absolute positioning accuracy. Repeatability reflects the deviation caused by multiple positioning at the same position, while absolute positioning accuracy measures the accuracy of the robot in reaching the target point. Together, they constitute the core indicators for evaluating the robot's motion performance. During long-term high-load service, various physical factors such as reducer wear backlash, motor thermal deformation, and arm elastic deformation are coupled together, which will cause the robot's absolute positioning accuracy to continuously decay over time, while repeatability is less affected by these factors.
[0003] Existing technical solutions for improving the absolute positioning accuracy of industrial robots can be mainly divided into the following categories;
[0004] The first type is the offline laser tracking calibration scheme. This type of scheme usually first plans the measurement path and determines the theoretical pose of the end effector based on the robot's kinematic model. After the robot completes the localization, the actual coordinates of the target point are measured using a laser tracker. The pose deviation of the end effector is calculated by matching it with the theoretical coordinates. Then, the kinematic model parameters inside the controller are corrected using identification algorithms such as the least squares method to achieve accuracy compensation. This type of scheme requires the robot to be stopped and a dedicated person to carry the laser tracker and other measuring equipment for on-site calibration. The calibration cycle is long and it is difficult to adapt to the continuous time-varying characteristics of errors during continuous robot service. In addition, frequent shutdowns for calibration will reduce the production efficiency of the production line.
[0005] The second type is the external sensor online feedback compensation scheme, which involves adding external sensing devices such as multi-dimensional force / torque sensors and laser tracking measurement systems to the robot end or joints. The controller is corrected in a closed loop by real-time feedback of end pose deviation data. Although this type of scheme can achieve a certain degree of online compensation, the hardware cost is high, the system integration is complex, the external sensors will occupy the effective load space of the end, and they are easily interfered with in the industrial environment with a lot of dust and strong electromagnetic interference, which will reduce the reliability of measurement.
[0006] The third category is schemes that estimate and compensate for non-geometric errors of robots based on artificial intelligence algorithms such as neural networks. These schemes have certain self-learning capabilities and real-time performance, but because their internal structure is an uninterpretable black box mapping structure and lacks clear mathematical boundaries for stability, they are difficult to meet the requirements of industrial functional safety certification for predictable error convergence, thus limiting the large-scale application of such schemes.
[0007] In summary, existing technologies for maintaining absolute positioning accuracy of industrial robots online suffer from problems such as long calibration cycles, high hardware modification costs, end-effector load encroachment, and a lack of provable closed-loop stability in compensation algorithms. It is difficult to achieve online, real-time, and non-stop decoupling identification of multi-source time-varying errors and adaptive compensation with theoretical stability guarantees without adding external measurement hardware. A new industrial robot control system is urgently needed to solve these problems. Summary of the Invention
[0008] The technical problem this invention aims to solve is: how to achieve online, real-time, and non-stop decoupling identification and compensation for time-varying absolute accuracy attenuation errors caused by multi-source coupling factors such as reducer wear, motor thermal deformation, and arm elastic deformation during long-term high-load service of industrial robots, without adding external force / torque sensors, laser trackers, or other hardware devices. At the same time, it addresses the problem that existing black-box intelligent compensation schemes based on deep learning lack closed-loop stability theoretical guarantees and are difficult to pass industrial functional safety certification.
[0009] To achieve the above objectives, the present invention provides the following technical solution: an industrial robot control system, comprising a host motion planning module, a multi-source error online decoupling module, an adaptive sliding mode compensation control module, and a servo driver, wherein the host motion planning module is used to generate the desired position, desired velocity, and desired acceleration of the robot joints;
[0010] The multi-source error online decoupling module is used to receive the joint position signal fed back by the encoder and the q-axis current signal fed back by the servo driver current loop. Based on the above signals, a third-order extended state observer is constructed, which includes the joint position estimation state, the joint velocity estimation state, and the lumped disturbance estimation state. Through the recursive operation of this observer, the time-varying errors of the robot joint caused by reducer wear, motor thermal deformation, and arm elastic deformation are uniformly classified into the lumped disturbance estimation state for online observation, and the joint position estimate, joint velocity estimate, and lumped disturbance estimate are output.
[0011] The adaptive sliding mode compensation control module is connected to the multi-source error online decoupling module. It receives the lumped disturbance estimate as the feedforward compensation input. Internally, it is equipped with a sliding surface calculation unit and a gain adaptive update unit. The sliding surface calculation unit constructs a sliding surface function based on the tracking error between the expected position and the joint position estimate and its derivative. The gain adaptive update unit adjusts the sliding proportional gain and the sliding switching gain online based on the real-time value of the sliding surface function. The rate of change of the sliding proportional gain is positively correlated with the square of the sliding surface function value, and the rate of change of the sliding switching gain is positively correlated with the absolute value of the sliding surface function value.
[0012] The adaptive sliding mode compensation control module synthesizes the feedforward compensation input with the feedback compensation term consisting of sliding mode proportional gain and sliding mode switching gain to generate a compensated joint torque command and output it to the servo driver, which then drives the joint motor to move.
[0013] The system provided by this invention completely reuses the joint servo motors, high-precision encoders, and servo drivers of existing industrial robots in terms of hardware architecture, without the need to add any external measurement hardware. Its single-joint dynamic behavior can be modeled as a second-order system affected by multi-source time-varying disturbances:
[0014] ;
[0015] In the formula, This indicates the joint angle position, which is obtained in real time from the encoder feedback.
[0016] Indicates the nominal equivalent moment of inertia of the joint;
[0017] Indicates the viscous damping coefficient;
[0018] This represents the nonlinear frictional torque related to velocity;
[0019] This represents the lumped time-varying uncertainty error torque generated by the coupling of reducer wear backlash, motor thermal deformation, and arm elastic deformation.
[0020] This represents the motor torque constant;
[0021] This invention represents the q-axis current value fed back by the servo driver current loop. The unknown nonlinear terms such as inertia, damping, and friction are treated as an extended state variable for online observation and identification. The identified lumped disturbance estimate is then introduced into the adaptive sliding mode control law in a feedforward manner, thereby achieving compensation without relying on external measurement hardware.
[0022] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows:
[0023] 1. This invention can complete the online decoupling and identification of multi-source errors by simply reusing the robot's original servo motor current loop data and encoder feedback signal, without adding any external force / torque sensors or optical tracking equipment, thus reducing hardware modification costs and avoiding the occupation of the effective load space of the end effector by external sensors.
[0024] 2. The present invention has a provable closed-loop stability theoretical basis through the control law and its gain adaptive update law adopted by the adaptive sliding mode compensation control module, which enables the compensation control loop to have a clear and calculable error convergence boundary, thus overcoming the defects of black box intelligent compensation schemes that lack stability mathematical boundaries and are difficult to pass industrial functional safety certification.
[0025] 3. This invention enables the signal interaction and calculation between the multi-source error online decoupling module and the adaptive sliding mode compensation control module to be completed within one communication cycle of the industrial Ethernet real-time communication bus, thereby achieving non-stop online accuracy maintenance of the robot during long-term continuous service and avoiding the production efficiency loss caused by traditional offline calibration methods. Attached Figure Description
[0026] Figure 1 This is a schematic diagram of the overall hardware hierarchy of the system in an embodiment of the present invention;
[0027] Figure 2 This is a flowchart illustrating the vertical calculation process within the multi-source error online decoupling module in this embodiment of the invention.
[0028] Figure 3 This is a complete single-cycle control execution flowchart in an embodiment of the present invention. Detailed Implementation
[0029] 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.
[0030] The industrial robot control system provided by this invention completely reuses the joint servo motors, high-precision encoders and servo drives of existing industrial robots in terms of hardware architecture, without the need to add any external measurement hardware.
[0031] In terms of software and firmware architecture, two core functional modules have been built: a multi-source error online decoupling module and an adaptive sliding mode compensation control module. The two work together through an industrial Ethernet real-time communication bus to form a complete closed-loop control loop. In the overall system architecture, the top layer is the upper-level motion planning module, which is responsible for generating the desired position, desired velocity and desired acceleration trajectory of each joint of the robot.
[0032] The middle layer is the industrial robot controller, which is internally divided into two functional sub-modules: a multi-source error online decoupling module and an adaptive sliding mode compensation control module.
[0033] The lower layer is a servo driver, which receives the compensated torque command output by the controller and drives the joint motor;
[0034] The bottom layer consists of the robot body joints, including servo motors, precision reducers, high-precision encoders, and load linkages. The joint position signal fed back by the encoder and the q-axis current signal fed back by the servo driver current loop flow along the signal link to the multi-source error online decoupling module. The lumped disturbance estimate output by the decoupling module flows to the adaptive sliding mode compensation control module. The compensation torque command output by the compensation control module flows to the servo driver via the industrial Ethernet real-time communication bus, forming a complete closed-loop path. The communication cycle of the industrial Ethernet real-time communication bus is less than 1 millisecond.
[0035] I. Online Decoupling Module for Multi-Source Errors
[0036] This invention models the dynamic behavior of a single joint of a robot as a second-order system affected by multi-source time-varying perturbations:
[0037] ;
[0038] In the formula, The joint angle position is indicated by real-time feedback from a high-precision encoder.
[0039] Indicates the nominal equivalent moment of inertia of the joint;
[0040] Indicates the viscous damping coefficient;
[0041] This represents the nonlinear frictional torque related to velocity;
[0042] This represents the lumped time-varying uncertainty error torque caused by multi-physical field coupling factors such as reducer wear backlash, motor thermal deformation, and arm elastic deformation.
[0043] This represents the motor torque constant;
[0044] This represents the q-axis current value fed back from the servo driver's current loop, used to indirectly estimate the actual output torque of the motor. In other words, the actual output torque of the motor is determined by... and The product is obtained.
[0045] To achieve lumped disturbance In the online identification process, this invention treats it together with unknown nonlinear terms such as inertia, damping, and friction as an extended state variable, constructs a third-order extended state observer, and defines the state variable. , , For the lumped perturbation expansion state that includes all the above uncertainties, the discretized observer recurrence equation is:
[0046] ;
[0047] ;
[0048] ;
[0049] ;
[0050] In the formula, Indicates the first One control cycle;
[0051] This indicates the duration of the control cycle, corresponding to the communication cycle of the industrial Ethernet real-time communication bus;
[0052] , , These represent the observers at the th... Estimates of joint position, joint velocity, and lumped disturbance for each cycle;
[0053] Indicates the encoder at the 1st The joint position actually measured in each cycle;
[0054] This represents the estimation error between the estimated joint position and the actual position.
[0055] , , These are the observer gain coefficients, tuned according to the joint inertia characteristics;
[0056] Here is a nonlinear error feedback function, where The power coefficient is... The width parameter of the linear interval, when hour, ;
[0057] when hour, ,in The sign function is designed to enable the observer to have smooth linear convergence characteristics when the estimation error is small to avoid high-frequency chattering, and to have power-law accelerated convergence characteristics when the estimation error is large to improve the observation response speed.
[0058] The internal signal processing flow of the multi-source error online decoupling module is as follows: It uses q-axis current feedback provided by the servo driver. and the position feedback provided by the encoder As input, first calculate the estimation error. Then, through a nonlinear function operation, we obtain... Then, substitute the results into the three integral recursion steps to complete the process. , , The update involves communication between different integral stages via gain coefficients. , , The feedback branches are connected to form a closed-loop correction structure;
[0059] Final output joint position estimate Joint velocity estimates and lumped disturbance estimate ,in The core output signal is transmitted to the downstream adaptive sliding mode compensation control module.
[0060] II. Adaptive Sliding Mode Compensation Control Module
[0061] Obtaining real-time estimates of lumped disturbances Subsequently, this invention designs an adaptive sliding mode control law guaranteed by closed-loop stability theory, which dynamically and in real-time compensates and updates the servo feedforward and feedback gains. First, the joint tracking error and sliding surface are defined:
[0062] ;
[0063] ;
[0064] In the formula, This indicates the desired joint position given by the higher-level motion planning module;
[0065] This represents the tracking error between the desired position and the observed position.
[0066] Positive sliding surface design parameters;
[0067] For sliding surface functions.
[0068] Based on this, a torque control law integrating feedforward compensation and adaptive feedback gain is constructed as the final torque command issued to the current loop of the servo driver:
[0069] ;
[0070] In the formula, For the desired acceleration;
[0071] This is a perturbation feedforward compensation term constructed based on the real-time identification results of the extended state observer, used to actively offset the impact of multi-source time-varying errors on the system;
[0072] , The sliding mode proportional gain and switching gain are respectively updated online adaptively over time. To enable the gain to automatically adjust according to the real-time tracking error and avoid reliance on offline manual tuning, the following adaptive update law is designed:
[0073] ;
[0074] ;
[0075] In the formula, , A positive adaptive learning rate constant determines the adjustment speed of the gain as the error changes. The aforementioned adaptive update law makes the sliding mode proportional gain... The rate of change is positively correlated with the square of the sliding surface function value, and the sliding mode switching gain The rate of change is positively correlated with the absolute value of the sliding surface function value, thereby achieving automatic matching of gain with the amplitude of tracking error.
[0076] Closed-loop stability theory guarantees: To verify the closed-loop stability of the above adaptive sliding mode control law, the following energy-type scalar function is constructed:
[0077] ;
[0078] In the formula, , Let $\mathbf$ and $\mathbf$ represent the estimation errors between the real-time adaptive gain and the ideal proportional gain and ideal switching gain required to overcome observation residuals and unmodeled dynamics, respectively. The ideal proportional gain and ideal switching gain are unknown but bounded constant parameters. Substituting the control law and adaptive update law into the equation... The derivative expression with respect to time can be derived and proven to be valid with appropriate parameter selection. , , Under the conditions, The semi-negative definiteness holds, and based on this, the convergence criterion for energy-type scalar functions can be rigorously proven for the sliding surface. and tracking error The system converges to a computable bounded neighborhood centered at zero within a finite time, thus providing a clear mathematical boundary for the error convergence of the compensation control loop of the present invention.
[0079] Compared to the black-box compensation schemes such as neural networks in the background technology, which cannot provide convergence boundaries due to the uninterpretable internal mapping, and the external sensor schemes that rely on the accuracy of measurement hardware and cannot avoid the accumulation of installation errors, the present invention can achieve a compensation effect with a calculable convergence boundary simply by reusing the encoder and current loop feedback signals. This overcomes the technical problem of black-box schemes and hardware-dependent schemes that are difficult to balance between stability provability and hardware cost.
[0080] III. Specific Implementation Steps
[0081] At the start of each industrial Ethernet real-time communication cycle, the robot controller synchronously acquires the angular position data fed back by the joint encoder. and q-axis current data from the servo driver current loop feedback As the input signal of the multi-source error online decoupling module, the multi-source error online decoupling module, based on the aforementioned third-order extended state observer recursive equation, utilizes the acquired... , Joint position estimation Speed estimates and lumped time-varying disturbance estimates Recursive updates are performed to achieve online decoupling and identification of multi-source errors such as reducer wear backlash and motor thermal deformation. Subsequently, the adaptive sliding mode compensation control module follows the desired trajectory given by the upper-level motion planning module. Calculate the tracking error using the aforementioned position and velocity estimates. and sliding surface function And adjust the sliding mode proportional gain according to the adaptive update law. With switching gain Online real-time adjustments are performed to automatically match the control gain based on the current error amplitude, eliminating the need for manual offline tuning. Then, the lumped disturbance estimate is... As a feedforward compensation term, it is combined with the sliding mode feedback compensation term to generate the final compensated joint torque control command. Finally, the compensated torque control command is sent to the servo driver current loop via the industrial Ethernet real-time communication bus to drive the joint motor to output the corrected actual torque, thus completing the real-time active compensation for multi-source time-varying errors.
[0082] Then, the process repeats in the next control cycle, forming a continuous closed-loop online error decoupling and compensation process without stopping the machine.
[0083] In one specific implementation, the industrial robot is a six-axis serial industrial robot. Each joint is independently equipped with the aforementioned multi-source error online decoupling module and adaptive sliding mode compensation control module. The joint controllers are synchronized with the upper-level motion planning module via an industrial Ethernet real-time communication bus, thereby achieving online maintenance of the robot's overall end-effector absolute positioning accuracy during long-term continuous service. The observer gain coefficient... , , Power coefficient and linear interval width parameter The sliding surface design parameters can be adjusted according to the specific joint's moment of inertia, load range, and encoder resolution. With adaptive learning rate constant , It can be tuned according to the dynamic characteristics of the desired joint trajectory to achieve an engineering balance between convergence speed and chatter suppression.
[0084] In another specific implementation, the industrial robot control system is also equipped with a vision measurement sensor access interface, which is connected to the multi-source error online decoupling module. This interface is used to access the initial pose deviation data collected by the vision measurement sensor during the initial power-on or maintenance phase of the system, so as to assist in calibrating the initial state of the third-order extended state observer, thereby further improving the convergence speed of the lumped disturbance estimate in the early stage of service.
[0085] During normal continuous service of the robot, the online decoupling module for multi-source errors and the adaptive sliding mode compensation control module can still independently complete online decoupling identification and compensation by relying entirely on encoder feedback signals and q-axis current feedback signals. The visual measurement sensor access interface is not necessary to realize the online decoupling identification and compensation function.
[0086] This invention also provides other alternative implementations that can achieve the same inventive objective. In the online decoupling and identification stage of multi-source errors, an unscented Kalman filter or a strong tracking Kalman filter can be used instead of the aforementioned third-order extended state observer. Similarly, only the motor current loop feedback and encoder feedback data are used. By constructing a nonlinear state-space model that includes the lumped disturbance state, the time-varying error is recursively estimated to achieve an online decoupling and identification effect similar to the extended state observer scheme. In the adaptive compensation control stage, adaptive backstepping control or fuzzy adaptive sliding mode control can be used instead of the aforementioned adaptive sliding mode control law. By designing virtual control quantities step by step and combining them with energy-type scalar functions, the stability of each subsystem is proven separately. Alternatively, fuzzy rules can be introduced to smoothly adjust the sliding mode switching gain to further suppress chattering. Both methods can achieve dynamic real-time compensation of the lumped disturbance estimate while ensuring that the closed-loop stability theory is provable.
[0087] The above-described specific embodiments are only some examples of the present invention. Any equivalent structural changes made using the content of this specification within the scope of the technical concept of the present invention, or direct or indirect applications in other related technical fields, should be included within the protection scope of the present invention.
[0088] Although the present invention has been disclosed above with reference to the specific embodiments described above, 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. An industrial robot control system, characterized in that, include: The upper-level motion planning module is used to generate the desired position, desired velocity, and desired acceleration of the robot joints; The multi-source error online decoupling module is used to receive the joint position signal fed back by the encoder and the q-axis current signal fed back by the servo driver current loop, and construct a third-order extended state observer that includes the joint position estimation state, the joint velocity estimation state and the lumped disturbance estimation state based on the joint position signal and the q-axis current signal. Through the recursive operation of the third-order extended state observer, the time-varying errors of the robot joint caused by reducer wear, motor thermal deformation and arm elastic deformation are merged into the lumped disturbance estimation state for online observation, and the joint position estimate, joint velocity estimate and lumped disturbance estimate are output. The adaptive sliding mode compensation control module is connected to the multi-source error online decoupling module via signal transmission, and is used to receive the lumped disturbance estimate output by the multi-source error online decoupling module as the feedforward compensation input; The adaptive sliding mode compensation control module includes a sliding mode surface calculation unit and a gain adaptive update unit. The sliding mode surface calculation unit is used to construct a sliding mode surface function based on the tracking error and its derivative between the expected position given by the upper-level motion planning module and the estimated joint position. The gain adaptive update unit is used to adjust the sliding mode proportional gain and the sliding mode switching gain online according to the real-time value of the sliding mode surface function. The rate of change of the sliding mode proportional gain is positively correlated with the square of the value of the sliding mode surface function, and the rate of change of the sliding mode switching gain is positively correlated with the absolute value of the value of the sliding mode surface function. The adaptive sliding mode compensation control module performs a synthesis operation based on the feedforward compensation input, the product of the sliding mode proportional gain and the sliding mode surface function, and the product of the sliding mode switching gain and the sign of the sliding mode surface function, to generate and output the compensated joint torque command. A servo driver, connected to the adaptive sliding mode compensation control module, is used to receive the compensated joint torque command and drive the joint motor to move.
2. The industrial robot control system according to claim 1, characterized in that, When performing the recursive calculation of the third-order extended state observer, the multi-source error online decoupling module corrects the recursive update amount of the joint velocity estimation state and the lumped disturbance estimation state through a nonlinear error feedback function based on the estimation error between the joint position estimate and the joint position signal fed back by the encoder. The nonlinear error feedback function exhibits linear characteristics when the absolute value of the estimated error is not greater than the preset linear interval width parameter, and exhibits power-law characteristics when the absolute value of the estimated error is greater than the preset linear interval width parameter.
3. The industrial robot control system according to claim 1, characterized in that, The multi-source error online decoupling module estimates the actual output torque of the motor based on the product of the motor torque constant and the q-axis current signal, and substitutes the joint position signal, the estimated joint speed state, and the actual output torque into the recursive calculation of the third-order extended state observer to obtain the lumped disturbance estimate.
4. An industrial robot control system according to claim 1, characterized in that, The estimation error between the sliding mode proportional gain and the ideal proportional gain, the estimation error between the sliding mode switching gain and the ideal switching gain, and the sliding surface function together constitute an energy-type scalar function. The derivative of the energy-type scalar function along the system running time is semi-negative definite, which makes the sliding surface function and the tracking error converge to a preset bounded neighborhood centered at zero within a finite time.
5. An industrial robot control system according to claim 1, characterized in that, It also includes an industrial Ethernet real-time communication bus, through which the multi-source error online decoupling module, the adaptive sliding mode compensation control module and the servo driver interact with each other, and the communication cycle of the industrial Ethernet real-time communication bus is 0.1 milliseconds to 1 millisecond.
6. An industrial robot control system according to claim 1, characterized in that, It also includes a visual measurement sensor access interface, which is connected to the multi-source error online decoupling module and is used to access the initial pose deviation data collected by the visual measurement sensor to assist the third-order extended state observer in performing initial state calibration.
7. An industrial robot control method, applied to the industrial robot control system according to any one of claims 1 to 6, characterized in that, include: Acquire the joint position signal fed back by the encoder and the q-axis current signal fed back by the servo driver current loop; Based on the joint position signal and the q-axis current signal, a recursive calculation is performed using a third-order extended state observer to obtain the joint position estimate, joint velocity estimate, and lumped disturbance estimate. The lumped disturbance estimate characterizes the time-varying error of the robot joint caused by reducer wear, motor thermal deformation, and arm elastic deformation. A sliding surface function is constructed based on the tracking error between the desired position and the estimated joint position and its derivative. The sliding proportional gain and sliding switching gain are adjusted online based on the real-time value of the sliding surface function. The rate of change of the sliding proportional gain is positively correlated with the square of the value of the sliding surface function, and the rate of change of the sliding switching gain is positively correlated with the absolute value of the value of the sliding surface function. The lumped disturbance estimate is used as the feedforward compensation input and combined with the feedback compensation term consisting of the sliding mode proportional gain and the sliding mode switching gain to generate the compensated joint torque command. The compensated joint torque command is sent to the servo driver to drive the joint motor to move.
8. The industrial robot control method according to claim 7, characterized in that, When performing the recursive calculation of the third-order extended state observer, the recursive update amount of the joint velocity estimate and the lumped disturbance estimate is corrected by a nonlinear error feedback function based on the estimation error between the joint position estimate and the joint position signal. The nonlinear error feedback function exhibits linear characteristics when the absolute value of the estimated error is not greater than the preset linear interval width parameter, and exhibits power-law characteristics when the absolute value of the estimated error is greater than the preset linear interval width parameter.
9. The industrial robot control method according to claim 7, characterized in that, The steps of acquiring the joint position signal and the q-axis current signal, performing the recursive calculation, constructing the sliding surface function and adjusting the sliding proportional gain and the sliding switching gain, and generating and issuing the compensated joint torque command are all completed within one communication cycle of the industrial Ethernet real-time communication bus, and the communication cycle is less than 1 millisecond.
10. The industrial robot control method according to claim 7, characterized in that, The estimation error between the sliding mode proportional gain and the ideal proportional gain, the estimation error between the sliding mode switching gain and the ideal switching gain, and the sliding mode surface function together constitute an energy-type scalar function. The derivative of the energy-type scalar function along time is semi-negative definite, which makes the sliding mode surface function and the tracking error converge to a preset bounded neighborhood centered at zero within a finite time.
