Model-free state disturbance observer driven robotic manipulator control method and system

Abstract

This invention relates to a model-free state disturbance observer-driven robot manipulator control method and system, belonging to the field of robot manipulator control technology. The method includes: establishing the robot manipulator dynamic equations based on unmodeled dynamic interference values; establishing a state-disturbance observer model; designing a disturbance observer incorporating generalized errors and nonlinear compensation terms; establishing independent joint control laws based on the compensated state-disturbance observer; optimizing the closed-loop system dynamic equations in step four; and establishing a generalized error-driven human-machine interaction system dynamic model to achieve global compliant control of the robot manipulator. This invention solves the problems of decreased control accuracy and insufficient system robustness in robot manipulators caused by model uncertainty, unmodeled nonlinearity, and external disturbances, particularly addressing the existing control methods' strong reliance on explicit modeling of system dynamics, difficulty in accurately obtaining state variables, and lag in disturbance compensation.

Description

Technical Field

[0001] This invention relates to the field of robot manipulator control technology, and in particular to a robot manipulator control method and system driven by a model-free state disturbance observer, which is suitable for high-precision motion tracking and compliant human-machine interaction scenarios. Background Technology

[0002] With the development of intelligent manufacturing and human-machine collaboration technologies, robotic arms are increasingly being used in industrial automation, medical rehabilitation, and service interaction. For modern multibody systems, especially serial structures with multiple degrees of freedom such as industrial robotic arms, humanoid upper limb systems, or service operation platforms, motion control accuracy and system robustness have become core research issues.

[0003] Existing robotic arm control methods generally rely on complete system dynamics models, such as using joint-space inverse dynamics for feedforward compensation or constructing optimized control laws in the task space. However, due to complex factors in real-world systems, such as friction, flexible structures, unmodeled nonlinearities, and external disturbances, dynamic interference arises between the ideal model and the actual system, severely impacting control performance. To compensate for the aforementioned unmodeled interference, current mainstream methods often employ Disturbance Observer (DO) or Extended State Observer (ESO) structures for closed-loop compensation. However, these methods typically rely on known system mathematical models, especially explicit expressions for the inertia matrix, nonlinear terms, and friction terms. Therefore, when faced with self-developed structures, variable operating conditions, or systems where the model is difficult to accurately identify, traditional modeling methods struggle to meet the dual requirements of high-precision control and real-time performance.

[0004] In addition, although the neural network control method, which has gradually emerged in recent years, has shown some ability to handle system nonlinearity and uncertainty, it often requires a large number of training samples, global state feedback or computing resources, which limits its real-time application value.

[0005] In summary, how to effectively address the decrease in control accuracy caused by model uncertainty, structural nonlinearity, and external disturbances in robotic manipulators without relying on precise modeling, especially under challenges such as the difficulty in accurately observing state variables and the lag in disturbance compensation, is a crucial technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] To overcome the shortcomings of existing technologies, this invention provides a model-free state disturbance observer-driven robot manipulator control method and system. Based on limited local state information, this method achieves accurate observation and compensation of disturbances without relying on explicit models, and further promotes the extension of state-driven control strategies towards compliant interaction.

[0007] A model-free state-disturbance observer-driven robot manipulator control method, including

[0008] A model-free state disturbance observer-driven independent joint trajectory tracking control strategy is provided, which specifically includes steps S1-S4.

[0009] S1. Establish the dynamic equations of the robot manipulator based on unmodeled dynamic interference values;

[0010] S2. Establish a state-disturbance observer model;

[0011] S3. Design a disturbance observer that incorporates generalized error and nonlinear compensation terms;

[0012] S4. Based on the compensated state-disturbance observer, establish independent joint control laws and construct the closed-loop system dynamic equations.

[0013] A model-free state-disturbance observer-driven global compliant control strategy is proposed, specifically including steps S1-S6;

[0014] S1. Establish the dynamic equations of the robot manipulator based on unmodeled dynamic interference values;

[0015] S2. Establish a state-disturbance observer model;

[0016] S3. Design a disturbance observer that incorporates generalized error and nonlinear compensation terms;

[0017] S4. Based on the compensated state-disturbance observer, establish an independent joint controller, construct the closed-loop system dynamic equation, and realize independent joint trajectory tracking control;

[0018] S5. Optimize the closed-loop system dynamic equations from step S4;

[0019] S6. Establish a dynamic model of the human-computer interaction system driven by generalized error to achieve global compliant control of the robot arm.

[0020] A model-free state perturbation observer-driven robot arm control system is also provided, comprising:

[0021] The robot manipulator dynamics model building module is used to establish the robot manipulator dynamics equations based on unmodeled dynamic interference values;

[0022] The State-Perturbation Observer Building Module is used to build a state-perturbation observer model based on generalized error and nonlinear compensation terms;

[0023] The strategy setting module is used to establish independent joint trajectory tracking or compliant control laws based on the compensated state-disturbance observer, and then construct the closed-loop system dynamic equations.

[0024] The advantages of this invention compared to the prior art are:

[0025] First, it does not rely on explicit modeling information such as system inertia matrix, nonlinear terms and friction model. It can realize disturbance estimation and control output based solely on joint encoder feedback, which significantly improves the robustness to system parameter uncertainty and external disturbance.

[0026] Second, by regulating the transient response and accumulation process of the generalized error, indirect control of the compliant behavior of the robot manipulator is achieved, constructing a dynamic response system with "follow-drift-damping" capabilities. Through dynamic regulation of the response process of the generalized error, active guidance and adjustment of the system's compliant behavior are realized under model-free conditions.

[0027] Third, the observer structure supports motion response triggered by minute external forces (such as human guidance), adapting to human action intentions and improving the safety, compliance and controllability of the robot manipulator during the interaction process.

[0028] IV. The control strategy is suitable for applications where robot manipulators, such as industrial collaborative robotic arms, precision force-controlled assembly, and medical rehabilitation training, have high requirements for responsiveness and safe interaction.

[0029] Fifth, since the observer and controller do not depend on the specific structural parameters of the robot manipulator, they can be quickly adapted to different models and degrees of freedom of robot manipulators, supporting modular porting and rapid deployment.

[0030] The present invention will be further described below with reference to the accompanying drawings and embodiments: Attached Figure Description

[0031] Figure 1 This is a flowchart of the independent joint control implementation based on a model-free state-disturbance observer provided in this application;

[0032] Figure 2 This is a flowchart of the global compliant control implementation based on a model-free state-disturbance observer provided in this application;

[0033] Figure 3 This is a block diagram of the independent joint control logic based on a model-free state-disturbance observer provided in this application;

[0034] Figure 4 This is a block diagram of the global compliant control logic based on a model-free state-disturbance observer provided in this application;

[0035] Figure 5 This is a schematic diagram of a perturbation observer based on a model-free state-perturbation observer provided in this application;

[0036] Figure 6This is a schematic diagram of the state observer based on the model-free state-disturbance observer provided in this application;

[0037] Figure 7 This is a comparison chart of the observed and actual values ​​of the disturbances at joints 1 to 6 of the independent joint control method based on a model-free state-disturbance observer in this embodiment.

[0038] Figure 8 This is a perturbation observation error diagram of joints 1 to 6 of the independent joint control method based on a model-free state-perturbation observer in an embodiment.

[0039] Figure 9 This is a comparison chart of the position observations, actual values, and expected values ​​of joints 1 to 6 in the independent joint control method based on a model-free state-disturbance observer in this embodiment.

[0040] Figure 10 This is a comparison chart of the observed, actual, and expected velocity values ​​of joints 1 to 6 in the independent joint control method based on a model-free state-disturbance observer, as described in this embodiment.

[0041] Figure 11 This is a position observation error diagram of joints 1 to 6 of the independent joint control method based on a model-free state-disturbance observer in an embodiment.

[0042] Figure 12 This is a velocity observation error diagram of joints 1 to 6 of the independent joint control method based on a model-free state-disturbance observer in an embodiment.

[0043] Figure 13 This is a torque response curve of the 5th and 6th joints under compliance control in the embodiment;

[0044] Figure 14 This is a graph showing the position and velocity response of the 5th and 6th joints under compliant control in the embodiment. Detailed Implementation

[0045] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. Unless otherwise stated, the technical or scientific terms used in this application have the ordinary meaning as understood by those skilled in the art.

[0046] Example 1, Reference Figure 1 , Figure 2 , Figure 3 , Figure 5 and Figure 6 This embodiment provides a model-free state perturbation observer-driven independent joint trajectory tracking control method for a robot manipulator, comprising the following steps S1-S4:

[0047] S1. Establish the dynamic equations of the robot manipulator based on unmodeled dynamic interference values;

[0048] The general expression for the dynamics of a robot manipulator is as follows:

[0049]

[0050] In the formula, It is the inertia matrix. It is a nonlinear matrix. It is the friction torque vector. It is a torque-current constant matrix. It is a joint angle vector. It is the driving current vector. It is the vector of the applied torque. The equation can be rewritten as:

[0051]

[0052] In the formula, This is the unmodeled dynamic interference value, and its expression is as follows:

[0053]

[0054] in, It is an identity matrix.

[0055] S2. Establish a state-disturbance observer model;

[0056] Introducing state equations

[0057]

[0058] in, Indicates the angles of each joint of the robot arm, Indicates the angular velocity of each joint of the robot arm, This represents the proposed unmodeled dynamic interference value. This represents the drive current for each joint of the robot manipulator. Let... and They are respectively and The observation error of the observed position is and velocity observation error The computational structure of the state-disturbance observer is as follows:

[0059]

[0060] In the formula, >0, >0 is an empirical parameter. It is a smooth and monotonically increasing zero-crossing odd function.

[0061] S3. Design a disturbance observer that incorporates generalized error and nonlinear compensation terms;

[0062] Based on the mathematical modeling of the state-disturbance observer described above, the disturbance observation error is derived. With output observation error Nonlinear dynamic mapping relationship between them:

[0063]

[0064] In the formula:

[0065]

[0066] This is characterized by the short-term and long-term variations of the output observation error, taking into account both the cumulative and transient effects of the output observation error. Therefore, This is called generalized error.

[0067] This embodiment proposes a model-free state-disturbance observer. This observer can estimate system disturbances online based on joint encoder feedback (joint position) without relying on explicit system models (such as inertia matrix, Coriolis torque term, friction model, etc.), and construct a type of trajectory tracking or compliant control architecture that supports joint observation of state and disturbance.

[0068] Consider the following compensated linear dynamic disturbance observer:

[0069]

[0070] In the formula, It is a parameter. This is a small disturbance compensation term, and its calculation structure is as follows: Figure 5 and Figure 6 As shown;

[0071]

[0072] In the formula, These are control parameters. This is a type of neuron function called the gradient-increasing neuron function. It is a time-varying parameter that decays exponentially, and its definition is:

[0073]

[0074]

[0075] Figure 5 and Figure 6 The input and output are each other's input and output, forming a closed loop. Figure 5The observed values ​​(or estimates) are used as output. As Figure 6 The observed values ​​entered in the input, Figure 6 Generalized error as output As Figure 5 Generalized error of the input; Figure 6 In It is the actual location value transmitted back from the sensor.

[0076] S4. Based on the compensated state-disturbance observer, establish an independent joint controller, construct the closed-loop system dynamic equation, and realize independent joint trajectory tracking control.

[0077] Construct an independent joint control law based on a state-disturbance observer: such as Figure 3 The above;

[0078]

[0079] In the formula

[0080]

[0081] They are respectively Expected value , , and ;

[0082] The dynamic equations of the closed-loop system can be obtained as follows:

[0083] .

[0084] Example 2: Based on the above independent joint control strategy steps S1-S4, referring to... Figure 2 This embodiment provides a model-free state-disturbance observer-driven global compliant control scheme for a robot manipulator. The implementation steps of this scheme include S1-S6:

[0085] S5. Optimize the closed-loop system dynamic equations from step S4;

[0086] In the global compliance control problem, let Therefore, the above closed-loop system dynamics equations degenerate into:

[0087]

[0088] Asymptotic convergence to zero means the entire robotic arm is almost weightless. However, in real-world environments... and There must be a tiny deviation between them, and this is the only driving force propelling the robot arm's movement. Therefore, in a real-world environment, this manifests as "random drift" or even instability. control, Therefore, "random drifting" becomes controllable. This is because humans possess a certain strength and are also an intelligent dynamic system with a high degree of judgment and adaptability.

[0089] S6. Establish a dynamic model of the human-computer interaction system driven by generalized error to achieve global compliant control of the robot arm.

[0090] The global compliance control law based on generalized error is specifically expressed as follows: Figure 4 As shown;

[0091]

[0092] In the formula, the control parameters Based on the above equation, the dynamics of a human-computer interaction system driven by generalized error information can be obtained:

[0093]

[0094] because and These generate spring and damping effects respectively, thus effectively preventing system instability.

[0095] Global compliant control strategy is based on independent joint control. Global compliant control is a compliant control strategy that covers the entire motion process and the entire system dimension. Its core objective is to achieve global optimization of the dynamic responsiveness and behavioral adaptability of the robot manipulator to external force input under complex and ever-changing external disturbances or human-machine interaction conditions.

[0096] "Compliance" describes a robot's ability to move compliantly and dynamically adapt to external disturbances (such as human thrust). Traditional compliant control is mostly based on impedance control or force-position adjustment in Cartesian space. The key to global compliant control lies in:

[0097] (1) Full system variable coverage: It not only adjusts the response of position, velocity, etc., but also integrates the coordinated adjustment of all dynamic variables such as state observation, disturbance estimation, and control torque.

[0098] (2) Full-process time-domain control: It not only exhibits compliance in steady state or specific time periods, but also maintains good compliance throughout the entire process from disturbance triggering, dynamic response to error convergence.

[0099] (3) Enhanced robustness without model: external disturbance terms are estimated by using state-disturbance observers, and a "generalized error" driving control law is constructed to achieve distributed compliant regulation that does not depend on external force sensors.

[0100] Example 3: This example proposes an independent joint control method based on a model-free state-disturbance observer, applicable to robust control tasks of robot manipulators under unmodeled dynamics and external disturbances. This method has good versatility and can be widely applied to control scenarios of various types of robotic arms and multi-degree-of-freedom systems. Compliant control verification was performed using the PUMA560 series six-degree-of-freedom robotic arm as a representative example. The PUMA560 platform is only a typical application case and does not constitute a limitation on the scope of application of the method.

[0101] In this embodiment, each joint of the robotic arm is treated as an independent controlled object. A state observer and a disturbance estimator are constructed for each degree of freedom to achieve a coupled design of local state feedback and global compensation control. This method does not rely on complete dynamic modeling of the system; it constructs the following control structure solely based on joint sensor data:

[0102] Estimation variables for the structural state of each joint Velocity estimator And construct a generalized error based on error feedback. It is used to sense the non-ideal motion state of the system.

[0103] Synthetic construction of first-order dynamic estimator With nonlinear disturbance compensation terms, design the disturbance estimation structure:

[0104]

[0105] in, This is an adaptive error adjustment term used to enhance robustness at low speeds.

[0106] The final independent joint control law is as follows:

[0107]

[0108] in, For the desired trajectory term, Let be the control torque of the i-th joint.

[0109] like Figure 7 As shown in the figure (Disturbance estimation represents the observed disturbance value, and true value represents the actual disturbance value), Figure 7 (a) shows two different graphs representing the first and second joints. Figure 7 (b) shows two different graphs representing the 3rd and 4th joints. Figure 7(c) shows two curves for the 5th and 6th joints, with the vertical axis representing Disturbance. Even without using the actual dynamic parameters of the system, the observer proposed in this method can still achieve accurate estimation of the disturbance, demonstrating strong model independence and robustness. Figure 8 The figure shows the trend of perturbation observation error (Dof represents joints or degrees of freedom, and the vertical axis Disturbance error represents perturbation error, where perturbation observation error is the largest component). , Figure 8 (a) shows the curves of the first and second joints. Figure 8 (b) shows the curves for the 3rd and 4th joints. Figure 8 (c) shows the curves for joints 5 and 6, whose amplitudes remain consistently low, further validating the estimation accuracy. Furthermore, as... Figure 9 (In the figure, desired pos represents the expected position value, true pos represents the actual position value, estimated pos represents the observed position value, and the vertical axis position represents the location.) Figure 9 (a) Three curve diagrams representing the first and second joints; Figure 9 (b) Three curve diagrams representing the 3rd and 4th joints; Figure 9 (c) shows three different curve diagrams for the 5th and 6th joints. (As shown in the image.) Figure 10 (In the figure, desired vel represents the expected velocity value, true vel represents the actual velocity value, estimated vel represents the observed velocity value, and the vertical axis vel represents velocity.) Figure 10 (a) Three curve diagrams representing the first and second joints; Figure 10 (b) Three curve diagrams representing the 3rd and 4th joints; Figure 10 (c) shows three curves for the 5th and 6th joints. The expected, actual and observed states of each joint of the system are highly consistent in terms of position and velocity, indicating that the constructed observer has good dynamic tracking capability. Figure 11 (In the diagram, Dof represents a joint or degree of freedom, with a total of 6 joints) This illustrates the position observation error ( ); Figure 12 (In the diagram, Dof represents a joint or degree of freedom, with a total of 6 joints) This illustrates the velocity observation error. ),from Figure 11 and Figure 12 As can be seen, its amplitude is extremely small, which fully reflects the accuracy and stability of the state estimation.

[0110] Example 4 presents a global compliant control method for a robot manipulator based on a model-free state-disturbance observer, applicable to compliant interactive control scenarios in systems such as multi-DOF manipulators. This method does not rely on complete dynamic modeling; instead, it dynamically captures unmodeled disturbances within the system and external forces by constructing a state observer and a disturbance estimation structure, thereby achieving global control of generalized errors and real-time generation of compliant behavior. In the simulation verification shown in this example, a PUMA series six-DOF manipulator is selected as the verification platform, and each joint is considered as an independent control unit. Control is achieved through the following steps. It should be noted that the PUMA560 manipulator is only a specific application example; the proposed method has good versatility and portability, and is applicable to other multi-DOF multibody systems.

[0111] By utilizing joint angle position errors and constructing differential terms, a model-free state observer is designed to estimate actual position and velocity, and a generalized error is constructed. Sum of error integral terms ;

[0112] Introducing disturbance estimator Driven by the output of the state observer, and combined with momentum feedback, saturation function and exponential correction term, it realizes real-time estimation of unmodeled dynamics and external force disturbances.

[0113] Constructing a combination error compensation term:

[0114]

[0115] This is used as an independent control input for each joint to generate the desired driving torque. The proposed control structure reveals the internal generalized error response term. A direct proportional relationship between human-computer interaction and external forces can be formally expressed as:

[0116]

[0117] This expression indicates that the generalized error response term within the system... Including instantaneous error Its cumulative integral items Together, they constitute the torque applied to the outside. The dynamic response mechanism. Since this excitation term directly drives the control input, its change trend maintains an approximately linear relationship with external disturbances, thus it can be regarded as a "channel" for the external interaction force to regulate the behavior of the controller.

[0118] This control strategy can guide the system to achieve global compliant behavior at the joint-level response level. Without relying on known system dynamic parameters such as mass, inertia, and friction, it can estimate and compensate for unknown external forces and unmodeled dynamics in real time. Through error feedback, it can achieve continuous and smooth interaction between the robot arm and the environment, so that the robot arm can maintain stable compliant behavior throughout the entire task space without switching local control strategies.

[0119] Using the PUMA560 robotic arm platform (it should be noted that the torque-current ratio matrix in the platform's dynamics is assumed to be an identity matrix), Normalization was performed. Therefore, the units of all disturbance and control torque terms were unified to Nm, without the need for additional conversion. Simulation experiments verified that the proposed control method exhibits good tracking accuracy and compliance performance under different external disturbances and task trajectories, effectively enhancing the robustness and environmental adaptability of the system, and is suitable for complex human-computer interaction and collaborative task scenarios.

[0120] In the experiment, regarding the perturbation settings, the system in Internally, external disturbance torques of limited amplitude are applied to joints 5 and 6 respectively. The perturbation consists of two parts: a low-frequency main perturbation (simulating the force application process during human-robot manipulator interaction) and a high-frequency oscillation term (simulating muscle tremors or hand vibrations), to enhance the validation coverage of the compliant controller under complex disturbance conditions. Its specific expression is as follows:

[0121]

[0122] In the formula: Indicates the joint number of the functioning joint; For the first Peak value of principal component of joint external force ( ); The amplitude of high-frequency flutter ( ); This refers to the high-frequency vibration frequency. This represents the boundary of the effective time period.

[0123] The fifth and sixth joints' compliant control torque response is as follows: Figure 13 The diagram shows the interaction forces between the fifth and sixth joints. Control torque and generalized error response term The dynamic changes of the generalized error response term can be observed (the light green shaded area represents the time period during which human interaction perturbation was applied). The amplitude and phase characteristics of both the low-frequency main disturbance and the high-frequency jitter components maintain good consistency with the applied external force, demonstrating high amplitude-frequency response capability and disturbance tracking accuracy. Meanwhile, the control torque... The magnitude of the external disturbance torque is almost exactly the same as that of the external disturbance torque, but the direction is always opposite, showing a typical disturbance cancellation relationship.

[0124] In addition, the fifth and sixth joints compliantly control position and velocity response, such as Figure 14 As shown (solid lines represent joint positions (unit: rad), dashed lines represent joint velocities (unit: rad / s); red and blue correspond to joints 5 and 6 respectively. The light green shaded area represents the period of human interaction disturbance), it can be observed that: during the external disturbance, the displacement trajectories of both joints change significantly, with the main motion directions being negative (joint 5) and positive (joint 6), consistent with the direction of the disturbance; the joint velocities also exhibit a high degree of consistency with the external disturbance in amplitude, trend, and direction, displaying synchronous dynamic characteristics. This indicates that the global compliant control described in this invention can both generate the expected displacement in a timely manner under external force and ensure that the motion state and interaction direction are coordinated, demonstrating excellent dynamic following performance and compliant adjustment characteristics.

[0125] This application has been disclosed above with preferred embodiments, but it is not intended to limit this application. Any person skilled in the art can make some changes or modifications to the above-disclosed structure and technical content to create equivalent embodiments without departing from the scope of the technical solution of this application, and all such modifications and modifications are within the scope of the technical solution of this application.

Claims

1. A robot manipulator control method driven by a model-free state perturbation observer, characterized in that: Includes the following steps: S1. Establish the dynamic equations of the robot manipulator based on unmodeled dynamic interference values; S2. Establish a state-disturbance observer model; First, we introduce the state equation: in, This indicates the angles of each joint of the robot arm. This represents the angular velocity of each joint of the robot arm. This represents the unmodeled dynamic interference value. This indicates the driving current of each joint of the robot manipulator, let and They are respectively and Given the observed values, we have: the position observation error is and velocity observation error ; Secondly, establish a state-disturbance observer: In the formula, >0, >0 is an empirical parameter. It is a smooth and monotonically increasing zero-crossing odd function; S3. Design a disturbance observer that incorporates generalized error and nonlinear compensation terms; Based on the state-disturbance observer mathematical model, establish the disturbance observation error. With output observation error Nonlinear dynamic mapping relationship between them: In the formula in, It is the cumulative amount of disturbance observation error, called generalized error; Consider a compensated linear dynamic disturbance observer: In the formula, It is a parameter. This is a small disturbance compensation term, and its calculation structure is as follows: In the formula, These are control parameters. For gradient-increasing neuron functions, It is a time-varying parameter that decays exponentially, and its definition is: in, These are empirical parameters used for regulation. The rate of decay; S4. Based on the compensated state-disturbance observer, establish independent joint control laws and construct the closed-loop system dynamic equations.

2. The robot manipulator control method driven by a model-free state perturbation observer according to claim 1, characterized in that: It also includes the following steps: S5. Optimize the closed-loop system dynamic equations from step S4; S6. Establish a dynamic model of the human-computer interaction system driven by generalized error to achieve global compliant control of the robot arm.

3. The robot manipulator control method driven by a model-free state perturbation observer according to claim 1 or 2, characterized in that: The process of establishing the robot manipulator dynamic equations based on unmodeled dynamic interference values ​​in step S1 is as follows: In the formula, It is the inertia matrix. It is a nonlinear matrix. It is the friction torque vector. It is a torque-current constant matrix. It is a joint angle vector. It is the driving current vector. It is the vector of the applied torque, which can be further rewritten as: In the formula, This is an unmodeled dynamic interference, and its expression is as follows: in, It is an identity matrix.

4. The robot manipulator control method driven by a model-free state perturbation observer according to claim 1, characterized in that: Step S4 establishes the independent joint control law based on the state-disturbance observer. The specific process of constructing the closed-loop system dynamic equations is as follows: The independent joint control law based on the state-disturbance observer is expressed as: In the formula They are respectively Expected value , , and ; The dynamic equations of the closed-loop system can be obtained as follows: .

5. The robot manipulator control method driven by a model-free state perturbation observer according to claim 4, characterized in that: The specific process of optimizing the closed-loop system dynamic equations in step S4 of step S5 is as follows: make Then, the dynamic equations of the closed-loop system degenerate into: 。 6. The robot manipulator control method driven by a model-free state perturbation observer according to claim 5, characterized in that: The specific process of establishing the generalized error-driven dynamic model of the human-computer interaction system in step S6 is as follows: In the formula, First-order dynamic filter representing generalized error, control parameters Based on the above equation, the dynamic model of the generalized error-driven human-computer interaction system can be obtained as follows: because and These generate spring and damping effects respectively, thus effectively preventing system instability.

7. A system for implementing the model-free state perturbation observer-driven robot manipulator control method as described in claim 1, characterized in that: Include: The robot manipulator dynamics model building module is used to establish the robot manipulator dynamics equations based on unmodeled dynamic interference values; The State-Perturbation Observer Building Module is used to build a state-perturbation observer model based on generalized error and nonlinear compensation terms; The strategy setting module is used to establish independent joint control laws based on the compensated state-disturbance observer, and then construct the closed-loop system dynamic equations.

Images