Cooperation of neural networks applied to a robotic hand-eye coordination task

Abstract

A basic positioning task is addressed on a hand-eye robotic platform using competitive neural networks. The approach is based on the cooperation of simple self-organizing maps acting together in sequential or simultaneous manner to achieve the global task. To use this concept efficiently a new neuronal structure is introduced, called the recruiting activation. The results are promising: fast learning and good precision are obtained through the cooperation of simple networks.