In this paper, a method is presented for fast learning of visually guided movements. The presented algorithms have been tested with a manipulator tracking manoeuvering target. Three parameters critical for the visuo-motor co-ordination have been identified, and are learned in less than one hour with repeated movements. The conditions for fast and precise vision have been investigated analytically, and the results of this analysis have been used for improving the image processing during the motions. After learning, the robot performs smooth and fast reaching movements and can easily drop small objects into the waggon of a moving model train. Finally, the method is generalised as a methodology for the representation of artificial systems, which provides them with the ability of adapting themselves to many different tasks. It is also explained how biological models can be used in this scheme.
Visuo-motor Co-ordination, Learning Robots, Human Motor Control, Vision Processing, Motion Planning
All articles published by Studies in Informatics and Control (SIC) are published in accordance with the terms and conditions of the Creative Commons Attribution License (CC BY-NC-SA 4.0).