Self-organisation and self-learning robot control

摘要

It is argued that self-organisation offers a means of data processing highly suitable for robot control. Self-organising structures are able to detect salient features of the environment and use them for tasks of classification, abstraction and association. Using self-organising mechanisms avoids the potential pitfalls of knowledge installation (rigidity and brittleness) and supervised learning (dependency upon the training information). (2 pages)