Title:
【発明の名称】マニピュレータ学習制御方法
Document Type and Number:
Japanese Patent JP2629291
Kind Code:
B2
Abstract:
PURPOSE:To find a mainpulator terminal position/attitude and a joint azimuth without the need for a troublesome analysis, large capacity memory, high performance computer, etc. by performing the repeating operation of a manipulator by using a neural network with the command value of the manipulator terminal position/attitude located on a working coordinate system as the input. CONSTITUTION:In the control system of the manipulator 2 of (m) degrees of freedom, which operates with the joint azimuth theta=(theta1, theta2,...,thetam) as the command input, learning is executed by performing the repeated operations of the manipulator 2 by using the neural network 1 composed of the three layer structure which makes the command value xd=(xd1, xd2,...,xdn) of the manipulators terminal position/attitude mentioned on a work coordinate system and outputs a manipulator joint azimuth theta. Consequently, the manipulators terminal position/ attitude and azimuth can easily be found without the need for a large capacity memory and high performance computer by the learning function of the neural network 1.
Inventors:
Ken Tsujimura
Tetsuro Yabuta
Shinichi Aoshima
Tetsuro Yabuta
Shinichi Aoshima
Application Number:
JP20577388A
Publication Date:
July 09, 1997
Filing Date:
August 19, 1988
Export Citation:
Assignee:
Nippon Telegraph and Telephone Corporation
International Classes:
B25J9/16; B25J9/10; G06F15/18; G06N3/00; G06N99/00; (IPC1-7): B25J9/10; B25J9/16; G06F15/18
Domestic Patent References:
| JP6272004A | ||||
| JP61253505A |
Attorney, Agent or Firm:
Masatoshi Isomura