Title:
ROBOT MODEL LEARNING DEVICE, ROBOT MODEL MACHINE LEARNING METHOD, ROBOT MODEL MACHINE LEARNING PROGRAM, ROBOT CONTROL DEVICE, ROBOT CONTROL METHOD, AND ROBOT CONTROL PROGRAM
Document Type and Number:
WIPO Patent Application WO/2022/172812
Kind Code:
A1
Abstract:
This robot control device acquires a past record value of a positional attitude of a robot and a past record value of external force to be applied to the robot, executes, on the basis of the past record value of the positional attitude at a certain time and an action command that can be given to the robot, a robot model including a state transition model for calculating a prediction value of the positional attitude of the robot and an external force model for calculating a prediction value of the external force to be applied to the robot, calculates a reward on the basis of an error of the positional attitude and the prediction value of the external force, generates a plurality of candidates for the action command so as to give the candidates to the robot model in every control cycle, determines an action command for maximizing the reward on the basis of rewards calculated corresponding to the plurality of candidates of the action command, and updates the external force model such that the difference between the prediction value of the external force calculated by the external force model on the basis of the determined action command and the past record value of the external force corresponding to the prediction value of the external force becomes small.
Inventors:
HAMAYA MASASHI (JP)
TANAKA KAZUTOSHI (JP)
TANAKA KAZUTOSHI (JP)
Application Number:
PCT/JP2022/003877
Publication Date:
August 18, 2022
Filing Date:
February 01, 2022
Export Citation:
Assignee:
OMRON TATEISI ELECTRONICS CO (JP)
International Classes:
B25J13/08
Foreign References:
| JP2018126802A | 2018-08-16 | |||
| JP2017030137A | 2017-02-09 | |||
| JP6458912B1 | 2019-01-30 | |||
| JP2020055095A | 2020-04-09 | |||
| JP2021020049A | 2021-02-18 |
Other References:
RUSU ET AL., PROGRESSIVE NEURAL NETWORKS, 2016
Attorney, Agent or Firm:
TAIYO, NAKAJIMA & KATO (JP)
Download PDF: