Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
LEARNING TYPE FEEDFORWARD CONTROLLER
Document Type and Number:
Japanese Patent JP3188040
Kind Code:
B2
Abstract:

PURPOSE: To simply generate a part corresponding to a subsystem of the learning type feedforward controller in a multi-input/output system.
CONSTITUTION: The learning type feedforward controller 12 generates an input signal with respect to a feedback control system obtaining a deviation between an object value and an observed value of a controlled system 15 to control the controlled system and with respect to said controlled system. The learning type feedforward controller 12 executes the processing based on a linear state equation 'X'AX+Bv, y'CX' of the controlled system having an input v, state X and output y. Then the learning type feedforward controller 12 is made up of a matrix 16 obtaining the state X from an object value, a differentiation circuit 17 differentiating the state X to obtain the X', a multiplier 18 multiplying a product between the state X and a coefficient A, a subtractor 19 subtracting the AX from the differentiation value X' to obtain the Bv, and a neural network 20 multiplying a B-1 obtained through learning with the Bv to obtain the input v to the controlled system 15.


Inventors:
Tsuyoshi Kakiuchi
Ikuji Terada
Ikuo Yamamoto
Application Number:
JP11049893A
Publication Date:
July 16, 2001
Filing Date:
May 12, 1993
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
MITSUBISHI HEAVY INDUSTRIES,LTD.
International Classes:
G05B11/32; G05B13/02; G06F15/18; G06G7/60; G06N99/00; (IPC1-7): G05B13/02; G05B11/32
Domestic Patent References:
JP484303A
JP63214801A
Attorney, Agent or Firm:
Takehiko Suzue