[ont.events] Artificial Neural Networks for Modelling Human Prehension.

ylkingsbury@watdragon.waterloo.edu (Yvonne Kingsbury) (05/31/89)

             The Institute of Computer Research
                          and the
                 Department of Kinesiology

                  present a colloquium on


                 Artificial Neural Networks
               for Modelling Human Prehension



by   Dr. Thea Iberall

of   Department of Computer Science, University of Southern California


ABSTRACT

Conceptual models, such as Arbib's coordinated control program of human
prehension, capture observations from behavioral studies of Jeannerod.
In trying to implement such a model in simulations of multi-layered artificial
neural networks, important questions arise in terms of internal representations
and sensorimotor integration.  In this talk, we will discuss how to extend the
Arbib conceptual model and then present results of simulations of artificial
neural networks that plan grasp postures based on task and object characteristics.  The goals of this research are to develop better models of the central
nervous system, and to develop algorithms for the control of dextrous robot
hands, such as the five-fingered Belgrade/USC Robot hand.



DATE:     Wednesday, June 7, 1989
TIME:     3:30 p.m.
ROOM:     DC 1302

Everyone is welcome.  Refreshments served.