Learning control of robots

Wubbe Jan Velthuis (W.J.R.Velthuis@el.utwente.nl)
Fri, 15 Mar 1996 17:51:58 +0100


I'm currently doing research on the application of learning
controllers to robots. At our laboratory the following controller
structure has been developed. During operation the neural network is
trained by the feedback controller to improve the performance.This
type of controller has shown to be able to learn to accurately control
a robot and compensate for reproducable distrubances (for example
place dependant friction). For applications see:
http://www.rt.el.utwente.nl/icontrol

_________
| neural |
|-------->| network |-----
| --------- |
| /\ |
| | |
| --------- |
_____ | __________ | \/ _________
| ref |---->o---->| feedback |--->o---->| process |----->
----- /\ | contr. | --------- |
| ---------- |
|-----------------------------------------|

In literature little can be found on this controller structure.

Therefore I would like to ask if there are other groups also
working on this type of controller or if anybody knows some
literature about it.

Please mail me any information about this controller.

THANK YOU VERY MUCH FOR YOUR ATTENTION AND YOUR RESPONSES

Wubbe Jan Velthuis
Control Laboratory
Department of Electrical Engineering
University of Twente
P.O.Box 217
7500 AE Enschede
THe Netherlands

W.J.R.Velthuis@el.utwente.nl

--
*****   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