Adaptive kinematic control apparatus

• US 5,579,442 A
• Filed: 04/28/1994
• Issued: 11/26/1996
• Est. Priority Date: 04/30/1993
• Status: Expired due to Term

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Masumoto, Daiki, Nagata, Shigemi, Yamakawa, Hiroshi, Kimoto, Takashi
• Primary Examiner(s): Downs, Robert W.
• Assistant Examiner(s): KATBAB, ABDOLLAH
• Application Number:
  US08/233,423
• Time in Patent Office: 943 Days
• Field of Search: 395/80, 395/81, 395/86, 395/96, 901/9, 901/15, 364/165, 318/568.11
• US Class Current: 700/246
• CPC Class Codes: B25J 9/161 Hardware, e.g. neural networks, fuzzy logic, interfaces, processor
• CPC Class Codes: B25J 9/1697 Vision controlled systems
• CPC Class Codes: G05B 2219/33036 Error back propagation
• CPC Class Codes: G05B 2219/39268 Layer perceptron, drive torque from state variables
• CPC Class Codes: G05B 2219/39295 Learn position correction values to be added to reference values
• CPC Class Codes: G05B 2219/40564 Recognize shape, contour of object, extract position and orientation
• CPC Class Codes: G05B 2219/40617 Agile eye, control position of camera, active vision, pan-tilt camera, follow object

Abstract

A network is provided in which the direct model of the controlled target is learned, and an actuation amount for realizing a target control amount is generated using this network in accordance with the relaxation algorithm. A recognizing network and a prediction network are provided, wherein the recognition result of the recognizing network is input to the prediction network, and the prediction result of the predicting network is fed back to the recognizing network. Moreover, the predicting network is utilized so as to generate a command relating to the motion in accordance with the relaxation algorithm.

Claims

1. A control apparatus by which an actuation amount required in controlling a control amount of a controlled object to match a target control amount is determined, said control apparatus comprising:

a converting part which is constructed to map the correspondence between an actuation amount indicated by the controlled target and a control amount, and which receives an input of the actuation amount so as to output a control amount corresponding to said input, wherein the converting part comprises a layer neural network including an input layer and output layer, wherein an error occurring between a target joint position and a current joint position is computed in the output layer and the error is propagated to the input layer so as to output the control amount;

a calculating part which, when an initial actuation amount is input to said converting part, calculates a corrective actuation amount corresponding to said initial actuation amount and allowing the target control amount to match the control amount output from said converting part, on the basis of a difference between the target control amount and the control amount output from said converting part; and

an outputting part which outputs a sum of the initial actuation amount and the corrective actuation amount calculated by said calculating part, as the actuation amount to be fed to the controlled target, wherein the actuation amount is used to control the controlled object.

2. The control apparatus as claimed in claim 1, wherein the calculating part calculates the corrective actuation amount while applying a restriction to at least one of a plurality of actuation amounts input to the converting part, each actuation amount being associated with an actuated parameter of the control target.

3. The control apparatus as claimed in claim 1, wherein the calculating part uses the corrective actuation amount permitted according to said restriction as the calculated corrective actuation amount, when the corrective actuation amount that allows the control amount to match the target control amount cannot be obtained.

4. The control apparatus as claimed in claim 3, wherein the actuation amount previously output from the outputting part is used as the initial actuation amount input to the converting part.

5. The control apparatus as claimed in claim 1, wherein the calculating part calculates the corrective actuation amount in accordance with the relaxation algorithm.

6. The control apparatus as claimed in claim 1, wherein the calculating part performs calculations of the corrective actuation amount using some of a plurality of actuation amounts input to the converting part, each actuation amount being associated with an actuated parameter of the controlled target.

7. A control apparatus by which an actuation amount required in controlling a control amount of a controlled object to match a target control amount is obtained, said control apparatus comprising:

a converting part which is constructed to map the correspondence between an actuation amount/a state amount indicated by the controlled target and a control amount, and which receives an input of the actuation amount/the state amount so as to output a control amount corresponding to said input, wherein the converting part comprises a neural network including an input layer and output layer, wherein an error occurring between a target joint position and a current joint position is computed in the output layer and the error is propagated to the input layer so as to output the control amount;

a calculating part which, when an initial actuation amount and the state amount of the controlled target is input to said converting part, calculates a corrective actuation amount corresponding to said initial actuation amount and allowing the target control amount to match the control amount output from said converting part, on the basis of a difference between the target control amount and the control amount output from said converting part; and

an outputting part which outputs a sum of the initial actuation amount and the corrective actuation amount calculated by said calculating part, as the actuation amount to be fed to the controlled target, wherein the actuation amount is used to control the controlled object.