CN-122008213-A - Humanoid variable impedance control method based on man-machine stiffness mapping

CN122008213ACN 122008213 ACN122008213 ACN 122008213ACN-122008213-A

Abstract

A humanoid variable impedance control method based on man-machine stiffness mapping comprises the steps of constructing an extended DMP frame on a two-dimensional sphere manifold, uniformly modeling motion trail, tail end stiffness and contact force information in multi-mode demonstration of a person, building a kinematic model of the upper half body of the person based on a spin theory, adopting a layered task control strategy, taking tail end pose tracking as a main task and chassis movement and arm configuration coordination as a secondary task, designing a humanoid variable impedance control law by combining demonstration force and man-machine operation difference, realizing closed loop tracking of expected force/moment, and generating coordinated and compliant vehicle arm cooperative actions by a system learning motion and interaction strategy from human demonstration during working. The robot has the beneficial effects that the direction track has no singularity, the motion-rigidity-force integration is repeated, the strong coupling coordination of the vehicle arm is good, the interaction is smooth and self-adaptive, and the humanoid performance and the environment adaptability of the robot in the tasks such as carrying are obviously improved.

Inventors

XU JUN
MEI XUESONG
WANG KAILONG
WU KEYAN
LI BAOZHU
Mei Jiale

Assignees

西安交通大学

Dates

Publication Date: 20260512
Application Date: 20260212

Claims (8)

1. The humanoid variable impedance control method based on the man-machine rigidity mapping is used for controlling the vehicle arm to execute a compliant operation task in cooperation with a robot and is characterized by comprising the following steps of: Step S1, establishing a unified model of multi-mode demonstration information of a person under a two-dimensional ball manifold extension DMP framework, wherein a tail end position track of the person demonstration is modeled in European space, and a tail end gesture track is modeled on the two-dimensional ball manifold; Step S2, combining the two-dimensional spherical manifold extension DMP frame modeling and the rotation theory of the step S1 to establish a human upper body kinematic model, and adopting a layered task control strategy to perform a motion adjustment method of human-computer physical interaction so that the robot can adapt to the dynamic change of the environment; S3, under the motion adjustment method in the step S2, obtaining the rigidity of the tail end of the upper limb of the person through a disturbance method or an sEMG estimation method, and generating a robot expected rigidity matrix through a mapping relation Constructing a humanoid variable impedance model based on the rigidity of the extended DMP frame and the terminal of the upper limb of the human; and S4, obtaining multi-mode demonstration information of the robot based on the humanoid variable impedance model in the step S3, designing a humanoid variable impedance controller of the robot, and dynamically adjusting the expected pose to accurately track the expected contact force/moment.
2. The humanoid variable impedance control method based on the man-machine stiffness mapping according to claim 1, wherein the step S1 is characterized in that the European space demonstration information is reproduced or generalized through an algorithm 1, the tail end position track of the human demonstration is modeled in the European space, the two-dimensional ball manifold demonstration information is reproduced or generalized through an algorithm 2, and the tail end gesture track is modeled on the two-dimensional ball manifold; The algorithm 1 is as follows: wherein { is as follows Model parameters of European space dynamic system, Is a parameter of a regular system, Is a time scale factor { The initial position, speed and acceleration of the robot, For the current state of the robot, Is a new target and { -Center, width and weight of radial basis function; The inputs to the algorithm 2 include new start and target { s } , Demonstration of Unit mapping vectors between adjacent vectors in corresponding tangent space on two-dimensional sphere manifold The implementation flow is as follows: Where λ is the tangential space velocity, d g (t) is the manifold distance of the current pose and the target pose, d g (0) is the manifold distance of the initial pose and the target pose, ω j is the weight of the jth radial basis function, h j is the width parameter of the jth radial basis function, and c j is the center position of the jth radial basis function.
3. The humanoid variable impedance control method based on man-machine stiffness mapping of claim 1, wherein in the step S2, the motion adjustment method adopts a hierarchical task control strategy for adjustment, and specifically includes: the primary task is to adjust the movement of the tail end of the robot so that the angular velocity of the joint is as follows: wherein: E 9 , the joint angular velocity under the primary task, ∈ 6 , ∈ 6 Respectively representing the current and expected terminal pose of the robot; ( ) ∈ 6×9 Representing a robot complete jacobian matrix; 1 ∈ 6×6 , 2 ∈ 9×9 And ∈ 6 ×6 Are symmetrical positive definite matrices, in which 1 >0; The secondary task is to make the movable chassis conform to the expected motion law under the condition of keeping a certain configuration of the mechanical arm, and the expected angular velocity of the joints is as follows: In the formula, ∈ 9 Representing a desired configuration of the mobile base and the manipulator in the secondary task; = [ ∈ 3 , 0 ∈ 6 And is determined by the back movement of the presenter, wherein, Representing edges The positions of the axes and the y-axis and the rotation angle along the z-axis, 0 Is the default joint configuration of the robotic arm; 3 ∈ 9×9 Is a positive definite diagonal matrix: 3 = diag {k 2 1 1×3 ,k 3 1 1 ×6 }。
4. the humanoid variable impedance control method based on man-machine stiffness mapping according to claim 1, wherein in the step S3, a robot expected stiffness matrix is obtained by mapping the upper extremity stiffness of the human being obtained by a perturbation method or an sEMG estimation method And constructing a humanoid variable impedance model based on the extended DMP frame and the rigidity of the tail end of the upper limb of the human: Wherein, the ∈ 6 For external force/moment applied to the tail end of the robot, the position and posture error of the tail end of the robot = d ∈ 6 ; , And Respectively representing the pose, the speed and the acceleration errors of the robot, wherein Represents the current pose of the robot, d Then for the desired pose of the robot, ∈ 6×6 , ∈ 6×6 And ∈ 6×6 Respectively a desired inertia, damping and stiffness matrix, which is to be expected in order to avoid the need for external force observation ∈ 6×6 Set as the inertia item of the robot itself ( ) The method comprises the following steps: = ( )。
5. the humanoid variable impedance control method based on man-machine stiffness mapping according to claim 1, wherein, In step S4, a robot-simulated variable impedance controller is designed based on the multi-mode demonstration information of the robot, and the robot-simulated variable impedance controller is based on a position correction term delta Impedance control model, delta =K d 1 (F d F) F is the actual interaction force, F d is the expected contact force, and the expected contact force/moment is accurately tracked by dynamically adjusting the expected pose; Combining the controller of step S4 with steps S1 to S3 to generate complete control instruction logic: Input: Generating an original expected pose X d DMP by the extended DMP of the step S1; the desired stiffness as planned by step S3 ; External force F ext and desired force F d measured in real time; And (3) calculating: calculating a force error, namely e F =F d -F ext ; calculating position correction term delta =K d 1 e F ; Generating a corrected expected pose of X d new =X d DMP +delta ; Output, substituting X d new and its derivative into the impedance model to obtain the required acceleration instruction ; Finally, the required acceleration instruction is solved through the layered task control strategy in the step S2 And converting the motion information into a speed or moment instruction of a joint space to drive the vehicle arm system.
6. A humanoid variable impedance control system based on man-machine stiffness mapping, characterized in that it performs the control method of any one of claims 1 to 5, comprising: the multi-modal information unified modeling module is used for establishing a unified model of multi-modal demonstration information of a person under a two-dimensional spherical manifold extension DMP frame, wherein the multi-modal demonstration information of the person comprises demonstration motion trail, demonstration rigidity information and demonstration force information; The kinematic modeling and adjusting module is used for establishing a human upper body kinematic model based on a rotation theory in step S2, and performing motion adjustment of human-computer physical interaction by adopting a layered task control strategy, wherein a primary task is end effector pose tracking, and a secondary task is coordination of movement of a mobile chassis and mechanical arm configuration; the humanoid variable impedance model module is used for acquiring the rigidity of the tail end of the upper limb of the human by a disturbance method or an sEMG estimation method in the step S3 and generating a robot expected rigidity matrix by a mapping relation Constructing a humanoid variable impedance model based on the rigidity of the extended DMP frame and the terminal of the upper limb of the human; And the humanoid variable impedance tracking control module is used for designing and outputting a robot humanoid variable impedance control instruction considering the demonstration force according to the demonstration force information and the man-machine operation difference in step S4 so as to finely adjust the expected pose of the tail end of the robot to realize accurate tracking of expected force/moment.
7. An arm co-robot device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, which when executed by the processor, performs the steps of the humanoid variable impedance control method based on man-machine stiffness mapping as claimed in any one of claims 1 to 7.
8. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the humanoid variable impedance control method based on man-machine stiffness mapping as claimed in any one of claims 1 to 5.

Description

Humanoid variable impedance control method based on man-machine stiffness mapping Technical Field The invention relates to the technical field of man-machine cooperation, in particular to a man-machine stiffness mapping-based humanoid variable impedance control method. Background As robots continue to expand from traditional industrial fields to fields of service, medical rehabilitation, home, education, etc., the mission of robots is evolving from traditional automated equipment to "partners" with the ability to autonomously complete tasks. The human-machine co-fusion is a trend of the development of the next generation robot, and the co-existence with people, the sharing of an operation space and the co-perception of information are important characteristics of the co-fusion robot, so that the universality of the robot is improved, and the use threshold of the robot is reduced. However, the conventional robot programming method generally has the defects of high upper threshold, complex task planning, low programming efficiency, poor flexible operation capability and the like, and greatly limits the application range of the robot. In the future, the operation environment of the robot is full of uncertainty, and it is difficult for people to directly establish the relation among multi-mode information such as operation motion, force and rigidity of the robot through a physical rule. The robots must also have the ability to be quickly programmed and adaptively adjusted in response to the needs of personalized designs and flexible manufacturing. The method gives the operation skill to the robot, enables the robot to quickly master the operation skill through demonstration and learning, realizes the quick planning of multi-mode information, adapts to the needs of personalized tasks, reduces the use threshold of the robot, and is one of important ways for improving the co-fusion level of the robot. Paper [Liu et al., "Learning Physical Human–Robot Interaction Skills via DMP-Based Variable Impedance Control,"IEEE Transactions on Industrial Electronics, 2020.] attempts to combine DMP with variable impedance control to adjust robot tip stiffness to achieve compliant interactions by learning position and force data in human operation. However, this method has problems: 1) The direction and the position are still decoupled, and the direction part does not adopt manifold constraint, so that the reproduction accuracy is low when complex gesture adjustment (such as overturning and oblique grabbing) is involved; 2) The rigidity parameter adjustment depends on simplifying assumptions (such as a diagonal rigidity matrix and a fixed adjustment law), and the time-varying characteristics and task correlation of the rigidity of the upper limb of the human body are not fully excavated; 3) The motion of the mobile platform is not integrated, and the mobile platform is only suitable for a fixed base mechanical arm, and can not solve the coordination problem caused by the double degrees of freedom of movement and operation in a vehicle arm cooperative system; 4) The force tracking is an open loop or weak feedback mechanism, a closed loop self-adaptive law of force deviation and pose correction is not established, and robustness is insufficient when the environment uncertainty is faced. Disclosure of Invention In order to overcome the problems of the prior art, the invention aims to provide a humanoid variable impedance control method based on man-machine stiffness mapping, which systematically solves the core problems of motion distortion, force control dislocation, poor coordination, weak adaptability and the like in the existing vehicle-arm cooperative technology through geometrically consistent motion representation, multi-modal information fusion, strong coupling system coordinated control and data-driven force interaction strategies, provides a novel efficient, safe and humanoid intelligent operation mode for scenes such as service robots, logistics transportation, rehabilitation assistance and the like, and solves the problems that the vehicle-arm cooperative robots have limited information modeling capability in the man-machine cooperation process, are difficult to simulate human flexible operation and are difficult to simultaneously process multi-modal information, so that the flexibility and control precision of the robots are improved. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: a humanoid variable impedance control method based on man-machine rigidity mapping is used for controlling a vehicle arm to cooperate with a robot to execute a compliant operation task, and specifically comprises the following steps: Step S1, establishing a unified model of multi-mode demonstration information of a person under a two-dimensional ball manifold extension DMP framework, wherein a tail end position track of the person demonstration is modeled in European space, and a tail end g