A robot compliant control method integrating multiple adaptive control mechanisms

Abstract

Robotic polishing and assisted rehabilitation therapy applications still face problems such as low motion accuracy, poor stability, and weak compliance. This paper discloses a robot compliant control method that integrates multiple adaptive control mechanisms. This method combines theoretical approaches such as globally stable nonlinear dynamic system learning, Lyapunov stability constraints, vector-valued sparse Gaussian process models, and passive power systems to innovatively construct a robot compliant control method that integrates multiple adaptive control mechanisms, enabling the robot to efficiently adapt to uncertain contact surfaces such as those with easily deformable and weak rigidity.

Description

Technical Field

[0001] The present invention relates to the field of robot control technology, and in particular to a robot compliant control method integrating multiple adaptive control mechanisms. Background Art

[0002] According to the "14th Five-Year Plan for the Development of the Robotics Industry," "skill learning and developmental evolution technology" is a cutting-edge technology in my country's core robotics research and development efforts, and is of great significance for strengthening the foundation for the development of the robotics industry. Furthermore, "how to achieve humanoid skill operations in robots" has been identified as one of the current cutting-edge scientific issues in my country's high-end manufacturing sector and a research focus at the intersection of artificial intelligence and robotics. This further validates the importance of robot skill learning technology as a key path to achieving humanoid skill operations in robots. Generally speaking, this technology typically involves three phases: "skill demonstration-skill learning-skill control." It can directly transfer human operational skills to robots, enabling them to possess humanoid operational capabilities, thereby achieving efficient programming, or even no programming required. It has broad application prospects in industries such as intelligent manufacturing and rehabilitation therapy. Skill control based on compliant control methods plays a decisive role in improving robots' task generalization and environmental adaptability.

[0003] However, based on previous research, the following problems were found:

[0004] 1) Existing robot compliance control methods mainly rely on fixed impedance parameters (such as stiffness and damping), which makes it difficult to strike a balance between motion accuracy and compliance. When the impedance parameters are small, the robot has better compliance but lower motion accuracy. When the impedance parameters are large, large impact forces are easily generated at the moment when the robot is in or out of contact with the environment, when transferring between different materials, and when the contact surface changes rapidly and dynamically, causing the robot to shake or even shut down safely. This is mainly due to the lack of an adaptive control mechanism; 2) Existing robot compliance control methods are mainly applicable to rigid contact surfaces and are difficult to apply to uncertain contact surfaces such as easily deformed soft tissues and weakly rigid thin-walled parts, resulting in problems such as low accuracy and poor adaptability.

[0005] Overall, traditional robot compliant control methods have not yet achieved ideal application results in handling complex force interaction tasks, and further research is urgently needed. Summary of the Invention

[0006] The purpose of the present invention is to overcome the above-mentioned deficiencies of the prior art and provide a robot compliant control method integrating multiple adaptive control mechanisms to achieve efficient adaptation of the robot to uncertain contact surfaces such as those that are easily deformed and have weak rigidity.

[0007] To achieve the above object, the technical solution of the present invention is:

[0008] The present invention provides a robot compliance control method integrating multiple adaptive control mechanisms, including:

[0009] Obtain four demonstration data sets of the robot end position and velocity, interaction force and velocity, position and interaction force, and interaction force and impedance;

[0010] For the position and velocity demonstration dataset, a Gaussian mixture model and a Gaussian mixture regression model are used to model and learn multiple nonlinear demonstration trajectories, and a parameterized quadratic Lyapunov function is introduced to perform globally stable motion estimation and parameter optimization to obtain a globally stable robot motion model.

[0011] For the three demonstration data sets of interaction force and velocity, position and interaction force, and interaction force and impedance, a vector-valued sparse Gaussian process model is used to establish an adaptive control mechanism between robot motion, interaction force, and impedance through nonlinear regression.

[0012] The globally stable robot motion model is integrated with the adaptive control mechanism among robot motion, interaction force and impedance.

[0013] Furthermore, four demonstration data sets of the robot end position and velocity, interaction force and velocity, position and interaction force, and interaction force and impedance are obtained as follows:

[0014] The robot was demonstrated multiple tasks using human-machine physical interaction, and the robot's end position, velocity, acceleration, and interaction force sensing information data were collected simultaneously to obtain four demonstration data sets of position and velocity, interaction force and velocity, position and interaction force, and interaction force and impedance.

[0015] Furthermore, the robot end position, velocity, acceleration, and interaction force sensing information data are subjected to noise reduction and singular point elimination through filters, and the dynamic time warping algorithm is used to time-align the multi-source demonstration trajectories to obtain effective demonstration data, which is expressed as where x t,m 、 They represent the position, velocity and acceleration of the mth demonstration trajectory at time t respectively; f e t,m represents the external interaction force of the mth demonstration trajectory at time t; T is the length of the mth demonstration trajectory; M is the number of demonstration trajectories.

[0016] Furthermore, assuming that the robot end is subject to the control force f c and external interaction force f e Interacting unit masses I M , then its dynamic model is simplified to in, represents acceleration; the control force f at the end of the robot c It is assumed to be described by a virtual spring-damper system to derive the dynamic model of the robot end at each moment: where K P is the stiffness matrix, is the expected position, x is the observed position, K V is the damping matrix, is the velocity; through equivalent transformation we get: Simplify both sides of the equation and express it as: K P X=Y, where X is the position error and Y is the intermediate variable; combining the sliding window and the least squares method, a rough estimate of the stiffness matrix is ​​achieved, expressed as K P =YX T [XX T ] -1 ; And through precise approximation, obtain the stiffness matrix of the robot end at each moment;

[0017] Through data annotation and processing, we obtain: position and velocity data sets where x n is the nth position quantity, v n is the nth velocity quantity, N is the number of demonstration data points; interactive force and velocity data set where f n is the nth interaction force, v n is the nth velocity; position and interaction force data set where x n is the nth position quantity, f n is the nth interaction force; interaction force and impedance data set where f n is the nth interaction force, d n is the nth stiffness measure.

[0018] Furthermore, the position and velocity demonstration dataset is modeled and learned using a Gaussian mixture model and a Gaussian mixture regression model for multiple nonlinear demonstration trajectories, and a parameterized quadratic Lyapunov function is introduced for global stable motion estimation and parameter optimization to obtain a globally stable robot motion model, including:

[0019] Demonstration dataset with position and velocity For reference, a Gaussian mixture model with K components is used to model the joint probability distribution of position and velocity. Where N(·) represents a high-dimensional Gaussian distribution model, is the overall parameter space of the probability distribution, Θ k is the parameter space of the kth component, i.e. Θk ={π k ,μ k ,Σ k},π k 、μ k and Σ k represent the weight, mean and covariance of the kth component respectively;

[0020] The Gaussian mixture regression model is used to obtain the conditional distribution of the velocity at a given position. Where 0<λ k (x) < 1 is a proportionality factor, and A k represents the transformation matrix composed of covariance, b k represents a linear term consisting of the mean;

[0021] Equivalently convert the function f(x) into a linear variable parameter system L(lx|A k ,b k ), and the parameterized quadratic Lyapunov function is used to verify the global stability condition of the linear variable parameter system, and the unknown parameter A of the linear variable parameter system is calculated by setting the error between the reference speed and the dynamic system speed as the objective function. k and b k Perform learning and optimization to obtain a globally stable robot motion model under any given target point.

[0022] Furthermore, the adaptive control mechanism among the robot motion, interaction force and impedance includes:

[0023] Control mechanism of interaction force and speed V d =φ(F r ), the regulatory mechanism of position and interaction force F d =φ(X r ), the regulation mechanism of interaction force and impedance D d =φ(F r );

[0024] Among them, V d ,F d ,D d are the estimated expected velocity, expected interaction force and expected impedance respectively, X r ,F r are the measured reference position and interaction force, respectively, and φ(·) is the mapping function.

[0025] Furthermore, the integration of the globally stable robot motion model and the adaptive control mechanism among robot motion, interaction force and impedance includes:

[0026] Assuming that the dynamic model of the robot in Cartesian space with torque control under gravity compensation is expressed as: Among them, x r , M(x r ), Represents the robot end reference position, reference velocity, reference acceleration, mass matrix and Coriolis matrix, F e and F c They represent the external force and control force at the end of the robot respectively;

[0027] External force F e It is directly measured by the force sensor at the end of the robot or the externally installed force sensor, while the control force is generated by the torque of each joint of the robot.

[0028] Furthermore, the passive power system controller is introduced to control the control force F c Characterize it so that the robot can move at the desired speed, that is, Among them, x r , Denote the robot reference position, reference velocity and desired velocity respectively, and D(x r ) is the damping matrix, expressed as D(x r )=Q(x r )ΛQ(x r ) T , Q(x r ) and Λ represent the orthogonal matrix and diagonal matrix of position drive, respectively, (·) T represents the transpose of the matrix; where the diagonal matrix Λ is the position-impedance control mechanism D d =φ(F r ) is obtained; the desired speed is obtained by the motion speed V obtained by the interactive force-speed control mechanism d =φ(F r ) and compensation speed f n (x r ), that is: The expected interaction force is controlled by the position-interaction force mechanism F d =φ(X r )get

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] Through an in-depth analysis of the current research status of existing robot compliance control methods, it was found that there is a lack of application research on uncertain contact surfaces such as deformable soft tissue and weak rigid thin-walled parts. In applications such as robot polishing and assisted rehabilitation therapy, problems such as low motion accuracy, poor stability, and weak compliance still exist. In response to this, the present invention combines the theories of global stable nonlinear dynamic system learning, Lyapunov stability constraints, vector-valued sparse Gaussian process models, and passive power systems to innovatively construct a robot compliance control method that integrates multiple adaptive control mechanisms to achieve efficient adaptation of the robot to uncertain contact surfaces such as deformable and weak rigidity. The main innovations include:

[0031] (1) A low-computational-complexity vector-valued sparse Gaussian process is proposed, which introduces sparse coding and variational inference to achieve modeling and learning of large-scale high-dimensional demonstration data, and obtain multiple adaptive control mechanisms with efficient responses;

[0032] (2) A robot compliant control framework integrating multiple adaptive control mechanisms is proposed to achieve efficient adaptation to uncertain contact surfaces such as those with easy deformation and weak rigidity. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a flow chart of a robot compliant control method integrating multiple adaptive control mechanisms provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0034] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0035] Example:

[0036] This paper innovatively constructs a robot compliant control method that integrates multiple adaptive control mechanisms, aiming to address the need for compliant operation on deformable and weakly rigid surfaces. This method combines theoretical approaches such as globally stable nonlinear dynamic system learning, Lyapunov stability constraints, vector-valued sparse Gaussian process models, and passive dynamic systems to achieve efficient adaptation to tasks and environments.

[0037] See Figure 1 As shown, the robot compliance control method integrating multiple adaptive control mechanisms provided in this embodiment includes the following steps:

[0038] 101. Obtain four demonstration data sets of the robot end position and velocity, interaction force and velocity, position and interaction force, and interaction force and impedance;

[0039] 102. For the position and velocity demonstration dataset, a Gaussian mixture model and a Gaussian mixture regression model are used to model and learn multiple nonlinear demonstration trajectories, and a parameterized quadratic Lyapunov function is introduced to perform globally stable motion estimation and parameter optimization to obtain a globally stable robot motion model;

[0040] 103. For the three demonstration data sets of interaction force and velocity, position and interaction force, and interaction force and impedance, a vector-valued sparse Gaussian process model is used to establish an adaptive control mechanism between robot motion, interaction force, and impedance through nonlinear regression;

[0041] 104. The robot's compliant control method is obtained by integrating the globally stable robot motion model with the adaptive control mechanism among robot motion, interaction force and impedance.

[0042] It can be seen that this method combines theoretical methods such as globally stable nonlinear dynamic system learning, Lyapunov stability constraints, vector-valued sparse Gaussian process models and passive power systems, and innovatively constructs a robot compliant control method that integrates multiple adaptive control mechanisms to achieve efficient adaptation of the robot to uncertain contact surfaces such as easily deformed and weakly rigid surfaces.

[0043] In one specific implementation, step 101 includes:

[0044] In order to meet the application requirements of robots for complex force interaction tasks, a six-dimensional force sensor is installed at the end of the robot to measure the interaction force information between the robot and the environment in real time. Under the condition of robot gravity compensation, the robot is demonstrated through human-machine physical interaction or remote operation, that is, humans guide the robot to complete the task multiple times, and simultaneously collect sensor information such as the position, velocity, acceleration, and interaction force of the robot end during the demonstration. In order to address the uncertainty and differences in the human-machine physical interaction process, a filter is designed to perform noise reduction and singular point removal on the demonstration data, and a dynamic time warping algorithm is used to time align the multi-source demonstration trajectories to obtain valid demonstration data, which is expressed as where x t,m 、 They represent the position, velocity and acceleration of the mth demonstration trajectory at time t respectively; f e t,m represents the external interaction force of the mth demonstration trajectory at time t; T is the length of the mth demonstration trajectory; M is the number of demonstration trajectories.

[0045] On this basis, in order to unify the dynamic characteristics of the interaction between the robot and the environment, it is assumed that the end of the robot is subject to the control force f c and external interaction force f e Interacting unit masses I M, then its dynamic model can be simplified to in, represents the acceleration. By changing the control force f at the end of the robot c It is set to be described by a virtual spring-damper system, so the dynamic model of the robot end at each moment is derived as follows: where K P is the stiffness matrix, is the expected position, x is the observed position, K V is the damping matrix, is the velocity. Through equivalent transformation, we can get:

[0046] According to the known variables and unknown variables, both ends of the equation are further simplified and expressed as: K P X=Y, where X is the position error and Y is the intermediate variable. In this regard, the sliding window and the least squares method are combined to achieve a rough estimate of the stiffness matrix, which is expressed as K P =YX T [XX T ] -1 In order to ensure the symmetric positive definite property of the stiffness matrix, the stiffness matrix of the robot end at each moment is obtained through accurate approximation.

[0047] Thus, through data annotation and processing, we can obtain: position and velocity data sets where x n is the nth position quantity, v n is the nth velocity quantity, N is the number of demonstration data points; interactive force and velocity data set where f n is the nth interaction force, v n is the nth velocity; position and interaction force data set where x n is the nth position quantity, f n is the nth interaction force; interaction force and impedance data set where f n is the nth interaction force, d n is the nth stiffness measure.

[0048] In one specific implementation, step 102 includes:

[0049] Demonstration dataset with position and velocity For reference, a Gaussian mixture model (GMM) with K components is used for modeling, and the joint probability distribution of position and velocity is obtained as Where N(·) represents a high-dimensional Gaussian distribution model, is the overall parameter space of the probability distribution, Θ kis the parameter space of the kth component, i.e. Θ k ={π k ,μ k ,Σ k},π k 、μ k and Σ k The Gaussian mixture regression model (GMR) is used to obtain the conditional distribution of the solution velocity at a given position. Where 0<λ k (x) < 1 is a proportionality factor, and A k represents the transformation matrix composed of covariance, b k represents a linear term composed of mean values. Then the function f(x) is equivalent to a linear variable parameter system L(lx|A k ,b k ), and the parameterized quadratic Lyapunov function is used to verify the global stability condition of the linear variable parameter system, and the unknown parameter A of the linear variable parameter system is calculated by setting the error between the reference speed (demonstration speed) and the dynamic system speed (speed estimated by Gaussian mixture regression model) as the objective function. k and b k Perform learning and optimization to obtain a globally stable robot motion model under any given target point.

[0050] In a specific embodiment, the step 103 includes:

[0051] The interaction force and velocity data sets are respectively Position and interaction force dataset Interaction Force and Impedance Datasets As a reference, an adaptive control mechanism between robot motion, interaction force and impedance is established. Generally, the input of the mapping mechanism is set to (position or interaction force), the output is (speed or impedance), since both input and output belong to high-dimensional data and there is a certain correlation between the dimensions, it is necessary to propose a mapping mechanism modeling method that satisfies high-dimensional input and output. This paper assumes that there is a functional relationship between input and output with Gaussian noise ∈, which is expressed as in is the functional relationship, σ s is the noise variance. In order to quickly respond to different input conditions and obtain the dynamic characteristics of the robot's interaction with the environment, the random variable is set Obey the Gaussian process distribution, that is Where GP(·) is a Gaussian process, μ(ξ I ) is the mean function, k(ξ iI ,ξ j I ) is the covariance function. However, the conventional Gaussian process needs to model all N demonstration data points, and its solution involves the inversion and determinant operation of the N×N dimensional similarity matrix, and the computational time cost is O(N 3 ), which is only applicable to simple tasks with small data volume, low dimension and low precision requirement. For the case of cubic time complexity, referring to Bayesian theory, this paper introduces a finite number of inducing variables and constructs a vector-valued sparse Gaussian process. M},M<<N approximates the original distribution and greatly reduces the time complexity of mapping mechanism modeling, where M is the number of induced variables.

[0052] Therefore, the low computational complexity vector-valued sparse Gaussian process is used to model and learn the mapping mechanism, and the interaction force and speed control mechanism V is obtained. d =φ(F r ), the regulatory mechanism of position and interaction force F d =φ(X r ), the regulation mechanism of interaction force and impedance D d =φ(F r ). Where V d ,F d ,D d are the estimated expected velocity, expected interaction force and expected impedance respectively, X r ,F r are the measured reference position and interaction force, respectively, and φ(·) is the mapping function.

[0053] In a specific embodiment, the step 104 includes:

[0054] Based on the robot's global stable motion model and motion-interaction force-impedance control mechanism, a stable variable impedance controller for the robot is constructed, which can effectively improve the robot's adaptability and safety. Assume that the dynamic model of the robot in Cartesian space under torque control under gravity compensation is expressed as: Among them, x r , M(x r ), Represents the robot end reference position, reference velocity, reference acceleration, mass matrix and Coriolis matrix, F e and F c They represent the external force and control force on the robot end. In particular, the external force F eThe control force can be directly measured by the force sensor at the end of the robot or the force sensor installed externally, and the control force is generated by the torque of each joint of the robot. In order to obtain a stable control law, the present invention introduces a passive power system controller to control the control force F. c Characterize it so that the robot can move at the desired speed, that is, Among them, x r , Denote the robot reference position, reference velocity and desired velocity respectively, and D(x r ) is the damping matrix, expressed as D(x r )=Q(x r )ΛQ(x r ) T , Q(x r ) and Λ represent the orthogonal matrix and diagonal matrix of position drive, respectively, (·) T Denotes the transpose of the matrix. Wherein, the diagonal matrix Λ is the position-impedance control mechanism D d =φ(F r ) is obtained; the desired speed is obtained by the motion speed V obtained by the interactive force-speed control mechanism d =φ(F r ) and compensation speed f n (x r ), that is: The expected interaction force is controlled by the position-interaction force mechanism F d =φ(X r )get.

[0055] In this way, a robot compliant control framework integrating multiple adaptive control mechanisms is obtained, which can achieve efficient adaptation to uncertain contact surfaces such as those that are easily deformed and have weak rigidity.

[0056] The above embodiments are intended only to illustrate the technical concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. They are not intended to limit the scope of protection of the present invention. Any equivalent changes or modifications made based on the essence of the present invention are intended to be covered by the scope of protection of the present invention.

Claims

1. A robot compliant control method integrating multiple adaptive control mechanisms, characterized in that: include: Obtain four demonstration data sets of the robot end position and velocity, interaction force and velocity, position and interaction force, and interaction force and impedance; For the position and velocity demonstration dataset, a Gaussian mixture model and a Gaussian mixture regression model are used to model and learn multiple nonlinear demonstration trajectories, and a parameterized quadratic Lyapunov function is introduced to perform globally stable motion estimation and parameter optimization to obtain a globally stable robot motion model. For the three demonstration data sets of interaction force and velocity, position and interaction force, and interaction force and impedance, a vector-valued sparse Gaussian process model is used to establish an adaptive control mechanism between robot motion, interaction force, and impedance through nonlinear regression. The globally stable robot motion model is integrated with the adaptive control mechanism between robot motion, interaction force and impedance, including: Assuming that the dynamic model of the robot in Cartesian space with torque control under gravity compensation is expressed as: Among them, x r 、 M(x r )and They represent the robot end reference position, reference velocity, reference acceleration, mass matrix and Coriolis matrix respectively, e and F c They represent the external force and control force at the end of the robot respectively; External force F e It is directly measured by the force sensor at the end of the robot or the externally mounted force sensor, while the control force is generated by the torque of each joint of the robot; The adaptive control mechanism among robot motion, interaction force and impedance includes: Control mechanism of interaction force and speed V d =φ(F r ), the regulatory mechanism of position and interaction force F d =φ(X r ), the regulation mechanism of interaction force and impedance D d =φ(F r ); Among them, V d ,F d ,D d are the estimated expected velocity, expected interaction force and expected impedance respectively, X r ,F r are the measured reference position and interaction force, respectively, and φ(·) is the mapping function.

2. The robot compliant control method integrating multiple adaptive control mechanisms according to claim 1, characterized in that: The four demonstration data sets of the robot end position and velocity, interaction force and velocity, position and interaction force, and interaction force and impedance are obtained as follows: The robot was demonstrated multiple tasks using human-machine physical interaction, and the robot's end position, velocity, acceleration, and interaction force sensing information data were collected simultaneously to obtain four demonstration data sets of position and velocity, interaction force and velocity, position and interaction force, and interaction force and impedance.

3. The robot compliant control method integrating multiple adaptive control mechanisms as claimed in claim 2, characterized in that: The robot’s terminal position, velocity, acceleration, and interaction force sensing information data are subjected to noise reduction and singular point elimination through filters, and the multi-source demonstration trajectory is time-aligned using the dynamic time warping algorithm to obtain effective demonstration data, which is expressed as where x t,m 、 They represent the position, velocity and acceleration of the mth demonstration trajectory at time t respectively; f e t,m represents the external interaction force of the mth demonstration trajectory at time t; T is the length of the mth demonstration trajectory; M is the number of demonstration trajectories.

4. The robot compliance control method integrating multiple adaptive control mechanisms as claimed in claim 3 is characterized in that: Assume that the robot end is subject to the control force f c and external interaction force f e Interacting unit masses I M , then its dynamic model is simplified to in, represents acceleration; the control force f at the end of the robot c It is assumed to be described by a virtual spring-damper system to derive the dynamic model of the robot end at each moment: where K P is the stiffness matrix, is the expected position, x is the observed position, K V is the damping matrix, is the velocity; through equivalent transformation we get: Simplify both sides of the equation and express it as: K P X=Y, where X is the position error and Y is the intermediate variable; combining the sliding window and the least squares method, a rough estimate of the stiffness matrix is ​​achieved, expressed as K P =YX T [XX T ] -1 ; And through precise approximation, obtain the stiffness matrix of the robot end at each moment; Through data annotation and processing, we obtain: position and velocity data sets where x n is the nth position quantity, v n is the nth velocity quantity, N is the number of demonstration data points; interactive force and velocity data set where f n is the nth interaction force, v n is the nth velocity; position and interaction force data set where x n is the nth position quantity, f n is the nth interaction force; interaction force and impedance data set where f n is the nth interaction force, d n is the nth stiffness measure.

5. The robot compliant control method integrating multiple adaptive control mechanisms as claimed in claim 1, characterized in that: The position and velocity demonstration dataset is modeled and learned using a Gaussian mixture model and a Gaussian mixture regression model for multiple nonlinear demonstration trajectories, and a parameterized quadratic Lyapunov function is introduced for global stable motion estimation and parameter optimization to obtain a globally stable robot motion model, including: Demonstration dataset with position and velocity For reference, a Gaussian mixture model with K components is used to model the joint probability distribution of position and velocity. Where N(·) represents a high-dimensional Gaussian distribution model, is the overall parameter space of the probability distribution, Θ k is the parameter space of the kth component, i.e. Θ k ={π k ,μ k ,Σ k },π k 、μ k and Σ k represent the weight, mean and covariance of the kth component respectively; The Gaussian mixture regression model is used to obtain the conditional distribution of the velocity at a given position. Where 0<λ k (x) < 1 is a proportionality factor, and A k represents the transformation matrix composed of covariance, b k represents a linear term consisting of the mean; Equivalently convert the function f(x) into a linear variable parameter system L(lx|A k , b k ), and the parameterized quadratic Lyapunov function is used to calculate the linear variable parameter system L(lx|A k , b k ) is verified, and the unknown parameter A of the linear variable parameter system is calculated by setting the error between the reference speed and the dynamic system speed as the objective function. k and b k Perform learning and optimization to obtain a globally stable robot motion model under any given target point.

6. The robot compliant control method integrating multiple adaptive control mechanisms as claimed in claim 1, characterized in that: Introducing the passive power system controller to control the control force F c Characterize it so that the robot can move at the desired speed, that is, Among them, x r , Denote the robot reference position, reference velocity and desired velocity respectively, and D(x r ) is the damping matrix, expressed as D(x r )=Q(x r )ΛQ(x r ) T , Q(x r ) and Λ represent the orthogonal matrix and diagonal matrix of position drive, respectively, (·) T represents the transpose of the matrix; where the diagonal matrix Λ is the position-impedance control mechanism D d =φ(F r ) is obtained; the expected speed is the motion speed V obtained by the interactive force-speed control mechanism d =φ(F r ) and compensation speed f n (x r ), that is: The expected interaction force is controlled by the position-interaction force mechanism F d =φ(X r )get.

Images