Newsgroups: comp.robotics.research,sci.engr.control,comp.ai.fuzzy,comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!hookup!news.nstn.ca!newsflash.concordia.ca!sunqbc.risq.net!hobbit.ireq.hydro.qc.ca!NetNews.IREQ.CA!mboyer
From: berthouz@robocom.etl.go.jp (Luc Berthouz)
Subject: Re: Learning control of robots
X-Nntp-Posting-Host: amadeus.ireq.ca
Message-ID: <MBOYER.96Mar15072312@amadeus.robot.ireq.ca>
Lines: 105
Sender: news@ireq.hydro.qc.ca (Netnews Admin)
Organization: Electrotechnical Laboratory
References: <MBOYER.96Mar13073004@amadeus.robot.ireq.ca>
Date: Fri, 15 Mar 1996 12:23:12 GMT
Approved: mboyer@robot.ireq.ca
Xref: glinda.oz.cs.cmu.edu comp.robotics.research:504 sci.engr.control:9649 comp.ai.fuzzy:7012 comp.ai.neural-nets:30516

Wubbe Jan Velthuis <W.J.R.Velthuis@el.utwente.nl> writes:

 > I'm currently doing research on the application of learning
 > controllers to robots. At our laboratory the following controller
 > structure has been developed. [...]
 > 
 >                        _________
 >                       | neural  | 
 >             |-------->| network |-----
 >             |          ---------      |
 >             |             /\          |
 >             |             |           |
 >             |             ---------   |
 >      _____  |          __________  |  \/     _________
 >     | ref |---->o---->| feedback |--->o---->| process |-----> 
 >      -----      /\    | contr.   |           ---------    |
 >                 |      ----------                         |
 >                 |-----------------------------------------|
 > 

hi:

Unless I make a big mistake, this controller is called:
feedback-error-learning It has been first introduced by Kawato in 87.

@ARTICLE{kawatofs,
  AUTHOR      = {M. Kawato and K. Furukawa and R. Suzuki},
  TITLE       = {A Hierarchical Neural-Network Model for Control and
		  Learning of Voluntary Movement},
  JOURNAL     = {Biological Cybernetics},
  volume      = {57},
  pages       = {169-185},
  YEAR        = {1987},
}

Many articles have followed :

@ARTICLE{gomik92 ,
  AUTHOR      = {H. Gomi and M. Kawato},
  TITLE       = {Adaptive Feedback Control Models of the
		  Vestibulocerebellum and Spinocerebellum},
  JOURNAL     = {Biological Cybernetics},
  volume      = {68},
  pages       = {105-114},
  YEAR        = {1992},
}

@ARTICLE{gomik93,
  AUTHOR      = {H. Gomi and M. Kawato},
  TITLE       = {Neural Network Control for a Closed-loop System Using
		  Feedback-Error-Learning},
  JOURNAL     = {Neural Networks},
  volume      = {6},
  pages       = {933-946},
  YEAR        = {1993},
}

M.I. Jordan refers to it in a technical report you can download by ftp.
ftp://psyche.mit.edu/pub/jordan/control-intro.ps.Z
(not so sure for the file)

(this techreport has become a book i think: Handbook of Perception and
Action: Motor Skills, but i've not the reference)

@TECHREPORT{jordan ,
  AUTHOR      = {M.I. Jordan},
  INSTITUTION = {Department of Brain and Cognitive Sciences,
		 Massachusetts Institute of Technology},
  NUMBER      = {9206}
  MONTH       = {June},
  TITLE       = {Computational Aspects of Motor Control and Motor Learning},
  TYPE        = {TR},
  YEAR        = {1992}
}

Finally, I've used feedback-error-learning to make a gaze-platform
"learn" its control. So for further details, please contact me.

@ARTICLE{berthouzeetal1 ,
  AUTHOR      = {L. Berthouze and S. Rougeaux and F. Chavand and Y. Kuniyoshi},
  TITLE       = {A Learning Stereo-head Platform Control System},
  JOURNAL     = {World Automation Congress/International Symposium on
		  Robotics And Manufacturing, Montpellier, France},
  MONTH       = {May 27-29, },
  YEAR        = {1996},  
  note        = {to appear}, 
}

Hope it helps,
luc

--
Luc BERTHOUZE - ETL Visiting Researcher
Electrotechnical Laboratory, Autonomous Systems Section
1-1-4 Umezono, Tsukuba-shi, Ibaraki 305, JAPAN
Tel: +81(0)298-58-5980	Fax: +81(0)298-58-5971
Email: berthouz@etl.go.jp 
WWW:   http://www.etl.go.jp/People/berthouz/

--
*****   Crossposted with comp.robotics.research (moderated)  *****
  Summary: Academic, government & industry research in robotics.
    Charter: ftp://ftp.ireq-robot.hydro.qc.ca/pub/crr/Charter
	       Submissions: crr@ireq-robot.hydro.qc.ca
 Meta-discussions/information: crr-request@ireq-robot.hydro.qc.ca
