Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

A Method for Robotic Multi-Axis Hole Assembly Using Deep Reinforcement Learning

A technology of reinforcement learning and robotics, applied in manipulators, program-controlled manipulators, manufacturing tools, etc., can solve problems such as difficult stable control, poor generalization ability, and poor versatility

Active Publication Date: 2020-06-09
TSINGHUA UNIV
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, with this method, as the assembly model becomes more and more complex, such as when the assembly task of multi-axis holes needs to be completed at the same time, the contact state will increase exponentially and the jamming model cannot be accurately analyzed, which leads to the current situation. The traditional fuzzy control method is difficult to achieve stable control, and its versatility to different assembly environments is poor and its generalization ability is poor

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A Method for Robotic Multi-Axis Hole Assembly Using Deep Reinforcement Learning
  • A Method for Robotic Multi-Axis Hole Assembly Using Deep Reinforcement Learning
  • A Method for Robotic Multi-Axis Hole Assembly Using Deep Reinforcement Learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0066] The method proposed by the present invention utilizes deep reinforcement learning to realize robot multi-axis hole assembly, and its flow chart is as follows figure 1 As shown, including establishing a simulation model, using the established simulation model to generate an empirical data set, and using the data in the empirical data set to train the deep reinforcement learning network to complete the assembly task of the robot,

[0067] The establishment of the simulation model includes the following steps:

[0068] (1) Establish the axis three-dimensional coordinate system X-Y-Z on the shaft component to be assembled, the coordinate origin O of the three-dimensional coordinate system X-Y-Z is located at the midpoint of the line connecting the centers of the two axes on the upper surface of the axis to be assembled, and the positive direction of the Z axis is downward along the axis of the axis, The positive direction of the X-axis is along the center of the left axis ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention relates to a method for achieving robot multi-axis-hole assembling through deep reinforcement learning, and belongs to the technical field of robot assembling. The method comprises the steps that in the training process, common experience data and expert experience data which are produced through a traditional fuzzy force control method and a deep reinforcement learning network on the basis of a simulation model are added into an experience data set, experience data are randomly drawn from the experience data set to train the deep reinforcement learning network, and therefore theassembling action of the network can rapidly achieve the assembling level of the traditional fuzzy control method and can exceed the assembling effect of the traditional fuzzy control method if training continues to be carried out. The deep reinforcement learning network trained by the simulation model is directly used for a practical robot multi-axis-hole assembling task. The experience data produced by the simulation model are used for training, the problem that practical assembling environment cannot provide enough training data is solved, and meanwhile the training cost is reduced.

Description

technical field [0001] The invention relates to a method for realizing multi-axis hole assembly of a robot by using deep reinforcement learning, and belongs to the technical field of robot assembly. Background technique [0002] Under the trend of rapid development of intelligent manufacturing, robotic automatic assembly technology has a huge market demand, and has been more and more used in various assembly fields. Aiming at the large number of multi-axis hole assembly tasks that exist in the industry at present, the most used one is the one proposed by Zhang Kuangen of the Robotics and Automation Research Office of the Manufacturing Institute of the Department of Mechanical Engineering, Tsinghua University in his paper "Forcecontrol for a rigid dual peg-in-hole assembly" Based on the method of fuzzy force control, this method is based on the detailed analysis of the contact state of the shaft hole, and the contact force model under different states can be established, and ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): B25J9/16
CPCB25J9/163B25J9/1633B25J9/1687
Inventor 徐静侯志民王国磊吴丹陈恳宋立滨
Owner TSINGHUA UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Eureka Blog
Learn More
PatSnap group products