Active Compliant Interaction Control Method and System for a Collaborative Robot with Force Sensors

Abstract

The present invention relates to the field of collaborative robot control, and specifically to an active compliant interaction control method and system for a collaborative robot without a force sensor. This method estimates the human-robot interaction force in real time through an interaction force estimation system, without the need for additional sensors, reducing the hardware cost; the impedance model optimization system solves the optimal impedance model through policy iteration to adapt to the uncertain interaction environment; and the active compliant interaction control system corrects the robot reference trajectory to achieve precise and reliable compliant control. The main solutions include: interaction force estimation based on generalized momentum, target impedance equation modeling, performance index function optimization, and barrier Lyapunov function control input design. This method solves the problems of high hardware cost, poor interaction adaptability, and unstable control of traditional collaborative robots, and is applicable to human-robot interaction scenarios in fields such as intelligent manufacturing and medical health.

Description

Technical Field

[0001] The present invention relates to the field of collaborative robot control, and particularly to an active compliant interaction control method and system for a collaborative robot without a force sensor. Background Art

[0002] With the development of fields such as intelligent manufacturing and healthcare, and the increasing demand for human-robot interaction, collaborative robots with characteristics such as good flexibility, lightweight, and high safety have received extensive attention, and interaction control has also become a research hotspot. In order to ensure the successful completion of interaction tasks, the robot needs to obtain the interaction force with humans in real time. The traditional method is to install a six-dimensional end force sensor on the robot wrist, but this not only increases the weight and size of the hardware system, but also raises the system cost. Moreover, when the robot comes into contact with a person at a location other than the end effector, the sensor cannot sense it. Collaborative robots generally consist of rigid links and joints containing motors and reducers. The application of harmonic reducers in the joint structure increases the joint torque and reduces the joint stiffness, which improves safety to a certain extent, but also severely limits the robot's ability and stability for high-precision operation due to the low stiffness. In addition, when the robot physically interacts with the environment, the human is an unknown time-varying dynamic environment for the robot, and traditional interaction control methods based on known interaction environment information are difficult to apply. Summary of the Invention

[0003] The purpose of the present invention is to solve how to achieve precise and reliable active compliant interaction control by estimating the interaction force in real time, optimizing the impedance model, and adjusting the control input for a collaborative robot without an external force sensor, thereby reducing the hardware cost and improving the adaptability and stability of human-robot interaction.

[0004] To achieve the above purpose, the present invention adopts the following technical solutions:

[0005] The present invention provides an active compliant interaction control method for a collaborative robot without a force sensor, including the following steps:

[0006] Step 1: Based on the robot state, estimate the interaction force during the interaction process of physical contact between a human and the collaborative robot through an interaction force estimation system to obtain interaction force information;

[0007] Step 2: Transmit the robot state and the interaction force information obtained in Step 1 to an impedance model optimization system, and return the optimal impedance model through a policy iteration solving module to adjust the dynamic characteristics of the robot to adapt to the uncertain interaction environment;

[0008] Step 3: Substitute the initial reference trajectory of the robot and the interaction force obtained in Step 1 into the optimal impedance model obtained in Step 2 to obtain the actual reference trajectory, and solve the control input of the robot through the active compliant interaction control system, so that the collaborative robot can track the actual reference trajectory to achieve accurate and reliable active compliant interaction control.

[0009] Among them, the reference trajectory includes:

[0010] Task space reference trajectory: The expected movement path of its end effector (such as the end of the robotic arm) in three-dimensional space;

[0011] Joint space reference trajectory: Defines how each joint moves over time to achieve the expected path in the task space.

[0012] In the above solution, Step 1 includes:

[0013] Step 1.1: Establish an interaction force observation system model through the dynamic equation of the flexible joint collaborative robot and the generalized momentum of the robot. The dynamic equation of the flexible joint collaborative robot is as follows:

[0014]

[0015]

[0016] represents the inertia matrix of the flexible joint robotic arm, represents the Coriolis force matrix of the flexible joint robotic arm, is the gravity force acting on the robot, is the human-robot interaction torque acting on the robot, is the control torque, represents the inertia matrix of the motor, represents the stiffness matrix of the joint, represents the joint angle vector, represents the joint angular velocity vector, represents the joint angular acceleration vector, represents the motor angle vector, represents the motor angular acceleration vector, where “·” represents the first derivative and “··” represents the second derivative;

[0017] The generalized momentum of the robot is:

[0018]

[0019] Thus, the interaction force observation system model of the flexible joint collaborative robot can be written as:

[0020]

[0021] Among them denotes the transpose of;

[0022] Step 1.2: Based on the interaction force observation system model, design an improved robust estimation method for the interaction force, input the robot state, and estimate the interaction torque in real time;

[0023] Define the interaction force estimation error as: where is the estimated value of the interaction torque, and the estimated value of the interaction torque is calculated by the following formula:

[0024]

[0025] where denotes the output of the generalized momentum observer, denotes the momentum of the robot system at time , and are both positive definite diagonal matrices, and , denotes the robust term.

[0026] In the above solution, the said step 2 includes:

[0027] Step 2.1: Establish a target impedance equation to describe the relationship between the position and velocity changes of the robot and the state changes of the interaction force with people. Specifically, establish the following target impedance equation to model the human-robot interaction:

[0028]

[0029] where are the unknown mass, damping, and stiffness matrices respectively, where denotes the change in the position of the robot, denotes the change in velocity, denotes the change in acceleration, denotes the interaction force;

[0030] Step 2.2: Based on the target impedance equation, combine the position change of the robot and the interaction force with people to determine the performance index function, form a quadratic programming problem, and the performance index function is set as follows:

[0031]

[0032]

[0033] where the state variable satisfies the following state equation:

[0034]

[0035]

[0036]

[0037] where denotes the inverse matrix of denotes the identity matrix of

[0038] where are all symmetric weight matrices, and are known matrices, is an auxiliary variable;

[0039] Step 2.3: Solve the quadratic programming problem by iteratively updating the calculation factor and impedance parameter iteration process of the policy iteration solution module to obtain the optimal impedance model:

[0040] Step 2.3.1: Based on the state equation in Step 2.2 , design the state feedback control input to minimize the performance index function in Step 2.2 :

[0041]

[0042] where denotes the optimal gain.

[0043] In the above solution, the said Step 3 includes:

[0044] Step 3.1: Substitute the initial reference trajectory of the robot and the interaction force information obtained in Step 1 into the optimal impedance model obtained in Step 2, solve the trajectory correction value of the robot, and correct the initial reference trajectory of the robot to obtain the actual reference trajectory;

[0045] The trajectory correction value is a sub-vector of the optimization variable:

[0046]

[0047] Thus, the initial reference trajectory can be corrected to obtain the actual reference trajectory :

[0048] ;

[0049] Step 3.2: Convert the actual reference trajectory of the robot into a numerical solution in the joint space through the joint space motion solution module to obtain the joint space reference trajectory ;

[0050]

[0051] Integrating gives the reference trajectory in the joint space where ; denotes the inverse matrix of the Jacobian matrix;

[0052] Step 3.3: Based on the barrier Lyapunov function, use the backstepping method to derive the control input of the flexible joint collaborative robot from the reference trajectory in the joint space, enabling the robot to track the actual reference trajectory;

[0053] Step 3.3.1: Define the following errors on the link side , sliding mode signal and auxiliary signal :

[0054]

[0055]

[0056]

[0057] denotes the desired joint angle vector, and in the following are all positive definite diagonal constant coefficient matrices;

[0058] Step 3.3.2: Construct the Lyapunov candidate function as:

[0059] where 𝜀>0 is a constant, is the inertia matrix of the flexible joint manipulator;

[0060] Step 3.3.3: Differentiate the Lyapunov function and design the virtual control input. Differentiate and substitute the sliding mode signal to obtain:

[0061]

[0062] Design a virtual control input , that is, the expected value of the motor angle :

[0063]

[0064] is a positive definite diagonal constant coefficient matrix used to adjust the gain of the control input;

[0065] Step 3.3.4: Define the error of the motor rotation angle and the sliding mode signal as:

[0066]

[0067]

[0068] Step 3.3.5: Construct the Lyapunov function :

[0069]

[0070] Step 3.3.6: Differentiate the Lyapunov function and design the control input:

[0071] Differentiate and substitute the signal and the virtual control input, then we can get:

[0072]

[0073] Design the control input as:

[0074]

[0075] is a positive definite diagonal constant coefficient matrix, and satisfies that its eigenvalues are greater than the maximum eigenvalue of the joint stiffness matrix ;

[0076] Step 3.3.7: Verify the stability of the system

[0077] Through the above control input, ensure that:

[0078] Thus, the stability of the system is guaranteed.

[0079] The present invention also provides a force - sensor - less cooperative robot active compliance interaction control system, including an interaction force estimation system, an impedance model optimization system, and an active compliance interaction control system;

[0080] Interaction force estimation system: Estimate the interaction force during the physical contact between humans and cooperative robots based on the robot state, and obtain the interaction force information;

[0081] Impedance model optimization system: Return the optimal impedance model according to the robot state and the interaction force information through the policy iteration solution module;

[0082] Active Compliant Interaction Control System: The initial reference trajectory and interaction force of the robot are substituted into the optimal impedance model to obtain the actual reference trajectory, and the control input of the robot is solved to enable the collaborative robot to track the actual reference trajectory, so as to achieve accurate and reliable active compliant interaction control.

[0083] Furthermore, in the above system, the interaction force estimation system includes:

[0084] An interaction force observation system modeling module, which is used to establish an interaction force observation system model based on the generalized momentum. The dynamic equation of the flexible-joint collaborative robot is as follows:

[0085]

[0086]

[0087] represents the inertia matrix of the flexible-joint manipulator, represents the Coriolis force matrix of the flexible-joint manipulator, is the gravity force acting on the robot, is the human-robot interaction torque acting on the robot, is the control torque, represents the inertia matrix of the motor, represents the stiffness matrix of the joint, represents the joint angle vector, represents the joint angular velocity vector, represents the joint angular acceleration vector, represents the motor angle vector, represents the motor angular acceleration vector;

[0088] The generalized momentum of the robot is:

[0089]

[0090] Therefore, the interaction force observation system model of the flexible-joint collaborative robot can be written as:

[0091]

[0092] where represents the transpose of;

[0093] An interaction force observation module, which is used to estimate the interaction torque in real time based on the interaction force observation system model and the improved interaction force robust estimation method;

[0094] Define the interaction force estimation error as: , where is the estimated value of the interaction torque, and the estimated value of the interaction torque is calculated by the following formula:

[0095]

[0096] where represents the output of the generalized momentum observer, represents the momentum of the robot system at time , and are both positive definite diagonal matrices, and , represents the robust term.

[0097] Furthermore, in the above system, the impedance model optimization system includes:

[0098] A target impedance equation establishment module, which is used to establish a target impedance equation that describes the relationship between the position, velocity and interaction force of the robot, and specifically establishes the following target impedance equation to model human-robot interaction:

[0099]

[0100] where are the unknown mass, damping and stiffness matrices respectively, where represents the change in the position of the robot, represents the change in velocity, represents the change in acceleration, represents the interaction force;

[0101] A performance index function determination module, which is used to determine a performance index function based on the target impedance equation, combined with the position change and interaction force of the robot, to form a quadratic programming problem. The performance index function is set as follows:

[0102]

[0103]

[0104] where the state variable satisfies the following state equation:

[0105]

[0106]

[0107]

[0108] where represents the inverse matrix of, represents The identity matrix;

[0109] wherein are all symmetric weight matrices, and are known matrices, is an auxiliary variable;

[0110] The policy iteration solving module is used to solve the quadratic programming problem by calculating the factor iteration update and the impedance parameter iteration process, and obtain the optimal impedance model. Specifically:

[0111] Based on the state equation design the state feedback control input to minimize the performance index function in step 2.2 :

[0112]

[0113] wherein represents the optimal gain.

[0114] Furthermore, in the above system, the active compliant interaction control system includes:

[0115] The reference trajectory acquisition module: used to obtain the actual reference trajectory of the robot based on the optimal impedance model and the initial reference trajectory of the robot. Specifically:

[0116] Substitute the initial reference trajectory of the robot and the interaction force information into the optimal impedance model, solve the trajectory correction value of the robot, and correct the initial reference trajectory of the robot to obtain the actual reference trajectory;

[0117] The trajectory correction value is a sub-vector of the optimization variable:

[0118]

[0119] Thus, the initial reference trajectory can be corrected to obtain the actual reference trajectory :

[0120] ;

[0121] The joint space motion solution module: used to convert the actual reference trajectory into a numerical solution in the joint space to obtain the joint space reference trajectory :

[0122]

[0123] Integrating can obtain the joint space reference trajectory wherein denotes the inverse matrix of the Jacobian matrix;

[0124] The control law update module, based on the barrier Lyapunov function, uses the backstepping method to derive the control input of the flexible joint collaborative robot from the joint space reference trajectory, enabling the robot to track the actual reference trajectory.

[0125] Furthermore, in the above system, the active compliant interaction control system includes: The specific implementation of the control law update module includes the following steps:

[0126] Step A: Define the following errors on the link side , sliding mode signal and auxiliary signal :

[0127]

[0128]

[0129]

[0130] denotes the desired joint angle vector, and the subsequent are all positive definite diagonal constant coefficient matrices;

[0131] Step B: Construct the Lyapunov candidate function as:

[0132] where 𝜀>0 is a constant, is the inertia matrix of the flexible joint manipulator;

[0133] Step C: Differentiate the Lyapunov function and design the virtual control input. Differentiate and substitute the sliding mode signal to obtain:

[0134]

[0135] Design a virtual control input , that is, the expected value of the motor angle :

[0136]

[0137] is a positive definite diagonal constant coefficient matrix used to adjust the gain of the control input;

[0138] Step D: Define the error of the motor rotation angle and the sliding mode signal as:

[0139]

[0140]

[0141] Step E: Construct the Lyapunov function :

[0142]

[0143] Step F: Differentiate the Lyapunov function and design the control input:

[0144] Differentiate , and substitute the semaphore and virtual control input, then we can get:

[0145]

[0146] Design the control input as:

[0147] 、 are positive definite diagonal constant coefficient matrices, and satisfies that its eigenvalues are greater than the maximum eigenvalue of the joint stiffness matrix ;

[0148] Step G: Verify the stability of the system

[0149] Through the above control input, ensure that:

[0150] Thus, the stability of the system is guaranteed.

[0151] Since the present invention adopts the above technical means, it has the following beneficial effects:

[0152] 1. Through the interaction force estimation system, the interaction force is observed in real time, and the interaction force information not limited to the end of the robotic arm can be accurately obtained in real time when the interaction force is time-varying, realizing safe and convenient human-machine interaction without external sensors, solving the problems of high integration difficulty and high cost existing in the traditional sensor-dependent method, and facilitating its large-scale application;

[0153] 2. Through the impedance model optimization system, it is not necessary to preset the expected impedance model according to expert experience or prior knowledge, overcoming the problem of difficult acquisition of real impedance parameters. At the same time, it can solve the problem that the interaction process is dynamic and changing, and it is difficult to maintain the optimal impedance control throughout the process;

[0154] 3. Through the active compliant interaction control system, the control problems brought about by the complex model and uncertain dynamic parameters of the flexible joint collaborative robot system can be solved, the state of the robot can be maintained within a feasible range, and the adaptability and stability of human-robot interaction can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0155] Figure 1 is a schematic diagram of the overall control system of the present invention;

[0156] Figure 2 is a schematic diagram of the interaction force estimation system of the present invention;

[0157] Figure 3 is a schematic diagram of the impedance model optimization system of the present invention;

[0158] Figure 4 is a schematic diagram of the active compliant interaction control system in the present invention;

[0159] Figure 5 is the working process of the system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0160] The following will give a detailed description of the embodiments of the present invention. Although the present invention will be described and explained in conjunction with some specific embodiments, it should be noted that the present invention is not limited to these embodiments only. On the contrary, any modifications or equivalent replacements made to the present invention should be covered within the scope of the claims of the present invention.

[0161] In addition, in order to better illustrate the present invention, numerous specific details are given in the following detailed description. Those skilled in the art will understand that the present invention can be implemented without these specific details.

[0162] The present invention is directed to a flexible joint collaborative robot with multiple degrees of freedom. Based on an improved robust estimation method for interaction force, the force during the interaction process is estimated, and the interaction force can be obtained in real time without adding additional sensors; an optimal impedance learning method based on an iterative strategy is used to learn the impedance model parameters during the interaction process, and the reference trajectory of the robot is corrected based on this impedance model; finally, a robust control method applicable to the flexible joint collaborative robot is used to track the reference trajectory, realizing accurate and reliable active compliant interaction control. The present invention well solves a series of problems such as high hardware cost, poor interaction adaptability, and unstable control of traditional collaborative robots.

[0163] Embodiment 1

[0164] The present invention provides a method for active compliant interaction control of a force sensorless collaborative robot, including the following steps:

[0165] Step 1: Based on the robot state, estimate the interaction force during the physical contact interaction between humans and collaborative robots through an interaction force estimation system, enabling the flexible joint collaborative robot to obtain interaction information with humans without relying on additional force sensors under time-varying interaction forces, thereby reducing hardware costs;

[0166] Step 2: Transmit the robot state and the interaction force information obtained in Step 1 to the impedance model optimization system, and return the optimal impedance model through the policy iteration solution module, thereby adjusting the dynamic characteristics of the robot to adapt to the uncertain interaction environment;

[0167] Step 3: Substitute the initial reference trajectory of the robot and the interaction force obtained in Step 1 into the optimal impedance model obtained in Step 2 to obtain the actual reference trajectory of the robot. Through the active compliant interaction control system, solve the control input of the robot so that it can track the reference trajectory, achieving the purpose of adjusting its response according to human actions and enhancing the fluency and compliance of the interaction.

[0168] As an implementation method, the sub-steps of Step 1 include:

[0169] Step 1.1: Establish an interaction force observation system model of the collaborative robot through the dynamic equation of the flexible joint collaborative robot and the generalized momentum of the robot;

[0170] The dynamic equation of the flexible joint collaborative robot is as follows:

[0171]

[0172]

[0173] where = ∈ and = ∈ are the joint angle vector and the motor angle vector respectively, ∈ is the control torque. ∈ and ∈ are the inertia matrix and the Coriolis force matrix of the flexible joint manipulator. ∈ is the gravity force acting on the robot, ∈ is the human-robot interaction torque acting on the robot. At the same time , is the human-robot interaction force, is the Jacobian matrix of the robot. ∈ and ∈ represents the inertia matrix of the motor and the stiffness matrix of the joint, represents the joint angular velocity vector, represents the joint angular acceleration vector, represents the motor angular acceleration vector.

[0174] The generalized momentum of the robot is:

[0175]

[0176] Thus, the interaction force observation system model of the flexible joint collaborative robot can be written as:

[0177]

[0178] where represents the transpose of.

[0179] Step 1.2, based on the interaction force observation system model, design an improved interaction force robust estimation method to input the robot state and estimate the interaction force in real time.

[0180] Define the interaction force estimation error as: , where is the estimated value of the interaction torque, then the estimated value of the interaction torque can be calculated by the following formula:

[0181]

[0182] where represents the output of the generalized momentum observer, represents the momentum of the robot system at time , and are both positive definite diagonal matrices, and , represents the sign function, that is:

[0183] As an implementation manner, the sub-steps of Step 1 include:

[0184] Step 2.1, establish a target impedance equation to describe the relationship between the changes in the position and velocity of the robot and the state changes of the interaction force with people;

[0185] Establish the following target impedance equation to model the human-robot interaction:

[0186]

[0187] where are the unknown mass, damping, and stiffness matrices, respectively, where represents the change in the position of the robot, represents the change in velocity, represents the change in acceleration, represents the interaction force.

[0188] Step 2.2: Based on the target impedance equation, combined with the position change of the robot and the interaction force with humans, determine the performance index function to form a quadratic programming problem;

[0189] The performance index function is set as follows:

[0190]

[0191] where are all symmetric weight matrices. Set the auxiliary variable to satisfy,[[]]END]] represents the derivative of with respect to,[[]]END]] and are known matrices. Thus, define the state variable . Then, the target impedance equation in Step 2.1 can be represented by the state equation:

[0192]

[0193] where , .

[0194] where represents the inverse matrix of,[[]]END]] represents the identity matrix of;

[0195] The performance index function can also be re-expressed as:

[0196]

[0197] where

[0198] Step 2.3: Solve the quadratic programming problem by iteratively updating the calculation factor and impedance parameter iteration process of the policy iteration solution module to obtain the optimal impedance model.

[0199] Based on the state equation in Step 2.2 , design the state feedback control input

[0200]

[0201] to make the performance index function in Step 2.2 Minimize and obtain the optimal gain 。

[0202] Obtain the optimal gain including the following steps:

[0203] 2.3.a. Select the feedback gain matrix that stabilizes the state equation in step 2.2 and let the control input be , where is the noise that satisfies persistent excitation. Calculate the factors , and until is satisfied:

[0204]

[0205] where represents the rank of matrix , that is, the number of linearly independent rows or columns in the matrix, represents the dimension of the state variable, represents the dimension of the control input, represents the th state of the system, represents squared, represents at time ;

[0206] 2.3.b. Solve and and where

[0207]

[0208]

[0209]

[0210] where represents the meaning of the th element of matrix , represents identity matrix, vec is an operator that "stretches" the matrix into a vector, ;

[0211] 2.3.c. Let and repeat 2.3.b until is satisfied, . Finally, obtain the optimal gain 。

[0212] As an implementation, the sub-steps of step 3 include:

[0213] Step 3.1: Substitute the initial reference trajectory of the robot and the interaction force information obtained in step 1 into the optimal impedance model obtained in step 2.3, solve for the trajectory correction value of the robot, and correct the initial reference trajectory of the robot to obtain the actual reference trajectory;

[0214] The trajectory correction value is a sub-vector of the optimization variable:

[0215]

[0216] Thus, the initial reference trajectory can be corrected to obtain the actual reference trajectory :

[0217]

[0218] Step 3.2: Convert the actual reference trajectory of the robot into a numerical solution in the joint space through the joint space motion solution module to obtain the joint space reference trajectory ;

[0219]

[0220] Integrating gives the joint space reference trajectory , where represents the inverse matrix of the Jacobian matrix.

[0221] Step 3.3: Based on the barrier Lyapunov function, use the backstepping method to derive the control input of the flexible joint collaborative robot from the joint space reference trajectory so that it can track the reference trajectory.

[0222] First, define the following link-side errors, sliding mode signal quantities, and auxiliary signal quantities:

[0223]

[0224]

[0225]

[0226] Here and in the following text are all positive definite diagonal constant coefficient matrices. Construct the Lyapunov candidate function

[0227]

[0228] where 𝜀 > 0 is a constant. For taking the derivative and substituting the semaphore, we can obtain:

[0229]

[0230] Thus, a virtual control input in the backstepping recurrence process is designed, that is, the expected value of is:

[0231]

[0232] Define the error of the motor rotation angle and the sliding mode semaphore as:

[0233]

[0234]

[0235] Construct the Lyapunov function :

[0236]

[0237] Similarly, taking the derivative of and substituting the semaphore and the virtual control input, we can get:

[0238]

[0239] Thus, the control input is designed as:

[0240]

[0241] where , are positive definite diagonal constant coefficient matrices, and satisfies that its eigenvalues are greater than the maximum eigenvalue of the joint stiffness matrix . Substituting into we can obtain

[0242]

[0243] Therefore, the human-computer interaction system is stable under the action of this control law.

[0244] Refer to Figures 1-4As shown in the figure, the system provided by the present invention includes an interaction force estimation system, an impedance model optimization system, and an active compliant interaction control system. The interaction force estimation system estimates the force of the physical interaction between humans and collaborative robots. The impedance model optimization system learns the impedance model parameters during the interaction process, and modifies the reference trajectory of the robot based on this impedance model. Finally, the active compliant interaction control system solves the control input of the robot to achieve accurate and reliable human-robot active compliant interaction control.

[0245] The interaction force estimation system first establishes an interaction force observation system model, and uses an improved interaction force estimation method based on a generalized momentum observer to estimate the interaction force. The improved interaction force estimation method based on a generalized momentum observer adds a robust control term compared with the traditional estimation method, solves the disadvantage that it cannot estimate time-varying interaction forces, and improves the accuracy and robustness of the interaction force estimation.

[0246] The impedance model optimization system first establishes a target impedance equation, then determines the performance index function, and finally returns the optimal impedance control gain through the policy iteration solution module.

[0247] The active compliant interaction control system first obtains the reference trajectory through the impedance model, then uses the joint space motion solution module to obtain the numerical solution in the joint space from the task space trajectory, and finally updates the control input of the flexible joint collaborative robot from the desired output through the control rate update module based on the barrier Lyapunov function.

[0248] Please refer to Figure 5 As shown in the figure, the working process of the system provided by the present invention is as follows: At the beginning of human-robot interaction, first obtain the current position and speed of the robot, and estimate the interaction force through the interaction force estimation system. Then, transfer the state and interaction force information of the robot to the target impedance equation and the performance index function, and obtain the optimal impedance model through policy iteration. Subsequently, correct the initial robot trajectory to obtain the reference trajectory of the robot's task space, and solve the corresponding reference trajectory in the joint space. Finally, generate the control input of the robot through the active compliant interaction control system so that it can track the reference trajectory. This process will be repeated continuously until the robot completes the specified task.

[0249] The above content is only an example and explanation of the present invention. Those skilled in the art of this technology make various modifications or supplements to the described specific embodiments or use similar methods to replace them. As long as they do not deviate from the invention or exceed the scope defined by this claim book, they should all belong to the protection scope of the present invention.

Claims

1. A method for active compliant interaction control of a collaborative robot without force sensors, characterized in that: The following steps are involved: Step 1: Based on the robot state, the interaction force in the interaction process of physical contact between humans and collaborative robots is estimated through the interaction force estimation system to obtain the interaction force information; Step 2: The robot state and the interaction force information obtained in step 1 are passed to the impedance model optimization system, and the optimal impedance model is returned through the strategy iteration solution module to adjust the dynamic characteristics of the robot to adapt to the uncertain interaction environment; Step 3: Substitute the initial reference trajectory of the robot and the interaction force obtained in step 1 into the optimal impedance model obtained in step 2 to obtain the actual reference trajectory, and solve the control input of the robot through the active compliant interactive control system so that the collaborative robot can track the actual reference trajectory to achieve active compliant interactive control; The reference trajectories include: Task space reference trajectory: the expected motion path of the robot end effector in three-dimensional space; Joint space reference trajectory: defines how each joint moves over time to achieve the desired path in task space; The step 1 comprises: Step 1.1: Establish the interactive force observation system model through the dynamic equation of the flexible joint collaborative robot and the generalized momentum of the robot. The dynamic equation of the flexible joint collaborative robot is as follows: represents the inertia matrix of the flexible joint manipulator, represents the Coriolis force matrix of the flexible joint manipulator, is the gravity acting on the robot, is the human-machine interaction torque on the robot, To control the torque, represents the inertia matrix of the motor, represents the stiffness matrix of the joint, represents the joint angle vector, represents the joint angular velocity vector, represents the joint angular acceleration vector, represents the motor angle vector, represents the motor angular acceleration vector, where " " represents the first-order derivative, " ” represents the second-order derivative; The generalized momentum of the robot is: Therefore, the interactive force observation system model of the flexible joint collaborative robot is written as: in express The transpose of Step 1.2: Based on the interaction force observation system model, an improved interaction force robust estimation method is designed. The robot state is input and the interaction torque is estimated in real time. The details are as follows: The interaction force estimation error is defined as: ,in is the estimated value of the interaction moment, which is calculated by the following formula: in represents the output of the generalized momentum observer, Indicates at time The robot system momentum at time , and are all positive definite diagonal matrices, and , represents the robust term; The step 2 comprises: Step 2.1: Establish the target impedance equation to describe the state change relationship between the robot's position, velocity change, and interaction force with the human. Specifically, establish the following target impedance equation to model human-machine interaction: in are the unknown mass, damping and stiffness matrices respectively, where represents the change in the robot's position, Indicates the speed change, represents the change in acceleration, represents the interaction force; Step 2.2: Based on the target impedance equation, combined with the robot's position change and the interaction force with the human, determine the performance index function to form a quadratic programming problem. The performance index function The setup is as follows: The state quantity The following state equation is satisfied: in express The inverse matrix of express The identity matrix of in are all symmetric weight matrices. and is a known matrix, is an auxiliary variable; Step 2.3: Solve the quadratic programming problem through the iterative update of the calculation factors and the iterative process of the impedance parameters of the strategy iteration solution module to obtain the optimal impedance model: Step 2.3.1: State equation based on step 2.2 , design state feedback control input Make the performance indicator function of step 2.2 Minimize: in represents the optimal gain; The step 3 comprises: Step 3.1: Substitute the initial reference trajectory of the robot and the interaction force information obtained in step 1 into the optimal impedance model obtained in step 2, solve the trajectory correction value of the robot, correct the initial reference trajectory of the robot, and obtain the actual reference trajectory; The trajectory correction value is a subvector of the optimization variables: So the initial reference trajectory Make corrections to obtain the actual reference trajectory : ； Step 3.2: Convert the actual reference trajectory of the robot into a numerical solution in the joint space through the joint space motion solution module to obtain the joint space reference trajectory : right The integration gives the joint space reference trajectory ,in represents the inverse matrix of the Jacobian matrix; Step 3.3: Based on the barrier Lyapunov function, the backstepping method is used to derive the control input of the flexible joint collaborative robot from the joint space reference trajectory, so that the robot can track the actual reference trajectory.

2. The method according to claim 1, characterized in that The step 3.3 comprises: Step 3.3.1: Define the error on the connecting rod side as follows , Sliding mode signal and auxiliary semaphores : represents the desired joint angle vector, And the following are all positive definite matrices with diagonal constant coefficients; Step 3.3.2: Construct Lyapunov candidate function for: where 𝜀 > 0 is a constant, is the inertia matrix of the flexible joint manipulator; Step 3.3.3: Derivative the Lyapunov function and design the virtual control input. Take the derivative and substitute the sliding mode signal into the equation to obtain: Design a virtual control input , that is, the motor angle Expected value: is a positive definite diagonal constant coefficient matrix used to adjust the gain of the control input; Step 3.3.4: Define the error of the motor rotation angle and sliding mode semaphore for: Step 3.3.5: Construct Lyapunov function : Step 3.3.6: Derivative the Lyapunov function and design the control input: right Taking the derivative and substituting the semaphore and virtual control input, we get: The design control input is: , is a positive definite matrix with diagonal constant coefficients, and Satisfies that its eigenvalue is greater than the joint stiffness matrix The maximum eigenvalue of ; Step 3.3.7: Verify the stability of the system by using the above control inputs to ensure: Thereby ensuring the stability of the system.

3. A sensorless collaborative robot active compliance interactive control system, characterized in that: Including interactive force estimation system, impedance model optimization system, active compliant interactive control system; Interaction force estimation system: estimates the interaction force during the physical contact between humans and collaborative robots based on the robot state to obtain interaction force information; Impedance model optimization system: returns the optimal impedance model through the strategy iteration solution module according to the robot state and interaction force information; Active compliant interactive control system: Substitute the robot's initial reference trajectory and interactive force into the optimal impedance model to obtain the actual reference trajectory, solve the robot's control input, and enable the collaborative robot to track the actual reference trajectory to achieve accurate and reliable active compliant interactive control; The reference trajectories include: Task space reference trajectory: the expected motion path of its end effector in three-dimensional space; Joint space reference trajectory: defines how each joint moves over time to achieve the desired path in task space; The interaction force estimation system includes: The interactive force observation system modeling module is used to establish an interactive force observation system model based on generalized momentum on the basis of the dynamics of the flexible joint collaborative robot. The dynamic equation of the flexible joint collaborative robot is as follows: represents the inertia matrix of the flexible joint manipulator, represents the Coriolis force matrix of the flexible joint manipulator, is the gravity acting on the robot, is the human-machine interaction torque on the robot, To control the torque, represents the inertia matrix of the motor, represents the stiffness matrix of the joint, represents the joint angle vector, represents the joint angular velocity vector, represents the joint angular acceleration vector, represents the motor angle vector, represents the motor angular acceleration vector, where " " represents the first-order derivative, " ” represents the second-order derivative; The generalized momentum of the robot is: Therefore, the interactive force observation system model of the flexible joint collaborative robot is written as: in express The transpose of An interaction force observation module, used for estimating the interaction torque in real time based on the interaction force observation system model and the improved interaction force robust estimation method; The interaction force estimation error is defined as: ,in is the estimated value of the interaction moment, which is calculated by the following formula: in represents the output of the generalized momentum observer, Indicates at time The robot system momentum at time , and are all positive definite diagonal matrices, and , represents the robust term; The impedance model optimization system comprises: The target impedance equation establishment module is used to establish the target impedance equation that describes the relationship between the robot's position, velocity, and interaction force. Specifically, the following target impedance equation is established to model human-machine interaction: in are the unknown mass, damping and stiffness matrices respectively, where represents the change in the robot's position, Indicates the speed change, represents the change in acceleration, represents the interaction force; The performance indicator function determination module is used to determine the performance indicator function based on the target impedance equation, combined with the position change and interaction force of the robot, to form a quadratic programming problem. The performance indicator function The settings are as follows: The state quantity The following state equation is satisfied: in express The inverse matrix of express The identity matrix of in are all symmetric weight matrices. and is a known matrix, is an auxiliary variable; The strategy iteration solution module is used to solve the quadratic programming problem and obtain the optimal impedance model by calculating the factor iteration update and the impedance parameter iteration process. Specifically: Based on the equation of state , design state feedback control input Make the performance indicator function Minimize: in represents the optimal gain; The active compliance interactive control system comprises: Reference trajectory acquisition module: used to acquire the actual reference trajectory of the robot based on the optimal impedance model and the initial reference trajectory of the robot, specifically: Substitute the robot's initial reference trajectory and interaction force information into the optimal impedance model, solve the robot's trajectory correction value, correct the robot's initial reference trajectory, and obtain the actual reference trajectory; The trajectory correction value is a subvector of the optimization variables: So the initial reference trajectory Make corrections to obtain the actual reference trajectory : ； Joint space motion solution module: used to convert the actual reference trajectory into a numerical solution in joint space to obtain the joint space reference trajectory : right The integration gives the joint space reference trajectory ,in represents the inverse matrix of the Jacobian matrix; The control rate update module, based on the barrier Lyapunov function, uses the backstepping recursion method to derive the control input of the flexible joint collaborative robot from the joint space reference trajectory, so that the robot can track the actual reference trajectory.

4. The system according to claim 3, characterized in that The active compliance interactive control system includes: the control rate update module is specifically implemented by the following steps: Step A: Define the error on the connecting rod side as follows , Sliding mode signal and auxiliary semaphores : represents the desired joint angle vector, And the following are all positive definite matrices with diagonal constant coefficients; Step B: Construct Lyapunov candidate function for: where 𝜀 > 0 is a constant, is the inertia matrix of the flexible joint manipulator; Step C: Derivate the Lyapunov function and design the virtual control input. Take the derivative and substitute the sliding mode signal into the equation to obtain: Design a virtual control input , that is, the motor angle Expected value: is a positive definite diagonal constant coefficient matrix used to adjust the gain of the control input; Step D: Define the error in the motor rotation angle and sliding mode semaphore for: Step E: Construct Lyapunov function : Step F: Derivative the Lyapunov function and design the control input: right Taking the derivative and substituting the semaphore and virtual control input, we get: The design control input is: , is a positive definite matrix with diagonal constant coefficients, and Satisfies that its eigenvalue is greater than the joint stiffness matrix The maximum eigenvalue of ; Step G: Verify the stability of the system, through the above control inputs, ensure that: Thereby ensuring the stability of the system.

Images