Search

CN-117325177-B - Anthropomorphic motion planning method, device and medium based on motion primitives

CN117325177BCN 117325177 BCN117325177 BCN 117325177BCN-117325177-B

Abstract

The invention discloses a anthropomorphic motion planning method, a device and a medium based on motion primitives, which comprise the steps of dividing subtasks of a task to be executed, planning the execution sequence of the subtasks according to task requirements, capturing motion data of a human arm when each subtask is executed, constructing a motion primitive library of each subtask by extracting and fitting quantification of the motion primitives under each task, and transferring the motion skills of the human arm to the anthropomorphic arm by taking the motion primitive library of each subtask as a carrier to realize anthropomorphic motion planning of a redundant mechanical arm. The invention can greatly reduce the data volume and improve the adaptability of the robot to tasks.

Inventors

  • WAN MINHONG
  • DUAN JINJUN
  • HUANG QIULAN
  • SUN WEIDONG
  • Bin Yiming
  • WANG LINGYU

Assignees

  • 之江实验室
  • 南京航空航天大学

Dates

Publication Date
20260508
Application Date
20231108

Claims (9)

  1. 1. The anthropomorphic motion planning method based on the motion primitives is characterized by comprising the following steps of: (1) Sub-task division is carried out on tasks to be executed, and the execution sequence of the sub-tasks is planned according to task requirements; (2) Capturing motion data of a human arm when each subtask is executed, and constructing a motion primitive library of each subtask by extracting and fitting quantification of motion primitives under each subtask; The action primitive extraction specifically comprises the following steps: (2.1) calculating the position of each frame motion from the motion data of the human arm Posture and attitude Elbow rotation angle ; (2.2) Position of motion for each frame Posture and attitude Elbow rotation angle Smoothing by adopting a filtering algorithm; (2.3) screening out effective movement fragments by judging the value of the hand speed, wherein if the speed is non-zero, the effective movement fragments are effective movement; (2.4) identifying motion primitives for effective motion for each frame, the motion primitives including a position change rate, an attitude change rate, and an elbow rotation angle change rate; (2.5) merging motion primitives with adjacent frames and consistent types to obtain all motion primitives of the whole effective motion segment; The action primitives with consistent types are consistent with elements contained in the action primitives, wherein the elements comprise position change rate, attitude change rate and elbow rotation angle change rate; (3) The motion primitive library of each subtask is used as a carrier to transfer the motor skills of the human arm to the anthropomorphic arm.
  2. 2. The method according to claim 1, wherein capturing the motion data of the human arm when each subtask is executed specifically comprises the steps of obtaining the motion data of the human arm at successive moments when each subtask is executed by using a motion capturing device; The motion capture device employs a sensor to sense motion of a human body.
  3. 3. The method of claim 1, wherein the elbow rotation angle The reference plane is a plane formed by two axes when the axis of a joint 2 and the axis of a joint 4 of the anthropomorphic serial mechanical arm are parallel, and sequentially represents a base, a shoulder 1, a shoulder 2, an elbow 1, an elbow 2, an elbow 3 and a wrist according to the joint numerical sequence; The elbow rotation angle The calculation method of (1) is as follows: Wherein, the Representing the vector formed from shoulder to elbow; representing the vector formed from shoulder to wrist; the vector formed from elbow to wrist is shown.
  4. 4. The method according to claim 1, wherein for the task to be performed, BVH data, i.e. motion data, is obtained by means of sensors of the body part where the data is to be acquired, said BVH data comprising rotation data of bones and limb joints of the body, and wherein the position and posture of the right/left hand in the shoulder coordinate system, i.e. the homogeneous transformation matrix, is calculated by means of a logical chain of bone trees in the BVH data: Wherein, the And Representing the posture and position of each physical node in the shoulder coordinate system, namely the base coordinate system, respectively For the number of physical nodes between the hand coordinate system and the shoulder coordinate system, it goes through 4 physical nodes of the hand, forearm and shoulder, thus Has a value of 4; Represented is the current physical node.
  5. 5. The method of claim 1, wherein the action primitives in the effective motion segment are fit quantized using an optimized fourier series function and an action primitive library corresponding to the subtasks is built; The optimized Fourier series function is as follows: Wherein, the 、 And (3) with All are coefficients of fourier series; the number of times of fourier series expansion is shown; Indicating the rotation speed; For being fitted with a function Is a periodic one.
  6. 6. The method according to claim 1, wherein the step (3) comprises the sub-steps of: (3.1) converting each element of the human arm motion primitive into state quantities describing human arm motion, namely a human arm end position, a human arm end gesture and an elbow rotation angle; (3.2) solving the spatial position of each physical node of the human arm or the redundant mechanical arm relative to a shoulder coordinate system according to the state quantity of the motion, wherein the physical nodes comprise arm tail ends, wrist nodes and elbow nodes; and (3.3) analyzing the movement joint angle of the redundant mechanical arm based on the spatial position of the physical node as follows: Wherein the method comprises the steps of 、 、 The position change rate, the posture change rate and the elbow rotation angle change rate respectively, Representing joints in numerical order The base, shoulder 1, shoulder 2, elbow 1, elbow 2, elbow 3, wrist are shown in that order.
  7. 7. The method according to claim 6, wherein the converting the elements of the human arm motion primitive into state quantities describing human arm motion is specifically: Wherein, the 、 、 The position of the tail end of the human arm, the gesture of the tail end of the human arm and the elbow rotation angle are respectively; is the initial position of the tail end of the human arm, For the initial attitude angle of the end of the human arm, Is the initial elbow rotation angle.
  8. 8. An action primitive based anthropomorphic motion planning device comprising one or more processors configured to implement the action primitive based anthropomorphic motion planning method of any one of claims 1-7.
  9. 9. A computer readable storage medium having stored thereon a program, which when executed by a processor is adapted to carry out the motion primitive based anthropomorphic motion planning method according to any one of claims 1-7.

Description

Anthropomorphic motion planning method, device and medium based on motion primitives Technical Field The invention belongs to the technical field of redundant mechanical arm motion planning, and particularly relates to an anthropomorphic motion planning method, device and medium based on motion primitives. Background With the improvement of economic level and technological level, the interaction scene of robots and human beings is more and more. Personification (HRI) refers to imparting human characteristics to non-human individuals, and realizing that robotic personified motions can bring more realistic and comfortable interaction experience to humans. The anthropomorphic arm is used as a key part for the interactive robot to execute complex tasks, and has the most intimate interaction with human beings. In order to ensure the high efficiency and safety of the interaction process, the improvement of the intelligent level and the anthropomorphic motion capability is an important problem to be solved. Anthropomorphic motion planning is a core technology for robots to generate anthropomorphic motions. At present, anthropomorphic motion planning methods based on performance indexes and human arm motion reproduction have the defects of difficult solution, various indexes, poor universality aiming at different tasks and the like. Therefore, in order to make up for a plurality of defects of the traditional anthropomorphic motion planning method, reduce the data volume and improve the adaptability of the robot to tasks, the design of the redundant mechanical arm anthropomorphic motion planning method which is simple in calculation and can be well adapted to tasks with different difficulties has urgent demands and great significance. Disclosure of Invention The invention aims to solve the technical problem of providing a anthropomorphic motion planning method, a device and a medium based on motion primitives aiming at the defects of the prior art. In order to achieve the technical aim, the technical scheme adopted by the invention is that an anthropomorphic motion planning method based on motion primitives comprises the following steps: (1) Sub-task division is carried out on tasks to be executed, and the execution sequence of the sub-tasks is planned according to task requirements; (2) Capturing motion data of a human arm when each subtask is executed, and constructing a motion primitive library of each subtask by extracting and fitting quantification of motion primitives under each subtask; (3) The motion primitive library of each subtask is used as a carrier to transfer the motor skills of the human arm to the anthropomorphic arm. Further, capturing motion data of the human arm when each subtask is executed, wherein the motion data of the human arm at continuous moments when each subtask is executed is obtained by adopting motion capturing equipment; The motion capture device employs a sensor to sense motion of a human body. Further, the action primitive extraction specifically includes: (2.1) calculating the position of each frame motion from the motion data of the human arm Posture and attitudeElbow rotation angle; (2.2) Position of motion for each framePosture and attitudeElbow rotation angleSmoothing by adopting a filtering algorithm; (2.3) screening out effective movement fragments by judging the value of the hand speed, wherein if the speed is non-zero, the effective movement fragments are effective movement; (2.4) identifying motion primitives for effective motion for each frame, the motion primitives including a position change rate, an attitude change rate, and an elbow rotation angle change rate; (2.5) merging motion primitives with adjacent frames and consistent types to obtain all motion primitives of the whole effective motion segment; the action primitives with consistent types are consistent with elements contained in the action primitives, wherein the elements comprise position change rate, gesture change rate and elbow rotation angle change rate. Further, the elbow rotation angleThe included angle between the reference plane and the arm plane is formed by a big arm and a small arm, wherein the reference plane is a plane formed by two axes when the axis of the joint 2 and the axis of the joint 4 of the anthropomorphic serial mechanical arm are parallel; The elbow rotation angle The calculation method of (1) is as follows: Wherein, the Representing the vector formed from shoulder to elbow; representing the vector formed from shoulder to wrist; the vector formed from elbow to wrist is shown. Further, according to the task to be executed, BVH data, namely motion data, is obtained through a sensor of a human body part needing to collect data, wherein the BVH data comprises rotation data of bones and limb joints of a human body, and the position and the posture of a right hand/left hand under a shoulder coordinate system, namely a homogeneous transformation matrix, are calculated through a logic cha