Acquisition of Automatic Activity through Practice: Changes in Sensory Input

Jack Gelfand, Marshall Flax, Raymond Endres, Stephen Lane, David Handelman

This paper will present computer models of three robotic motion planning and learning systems which use a multi-sensory learning strategy for learning and control. In these systems machine vision input is used to plan and execute movements utilizing an algorithmic controller while at the same time neural networks learn the control of those motions using feedback provided by position and velocity sensors in the actuators. A specific advantage of this approach is that, in addition to the system learning a more automatic behavior, it employs a computationally less costly sensory system more tightly coupled from perception to action.


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.