Developmental robotics architecture for active vision and reaching

A robotic architecture is presented which is inspired by the process of developmental learning of human infants in their twelve months after birth. The architecture integrates active vision and simple object manipulation (reaching and grasping). Robotic experiments demonstrate how visual and non-visual features determine visual attention and reaching. However, more important and main objective of this paper is the organisation of the architecture with respect to developmental learning of firstly, the behavioural competence of hand-eye coordination and secondly, the cognitive competence of multimodal visual attention. These topics of staged competence learning of behavioural and cognitive skills are discussed. In this discussion we also outline the value of this architecture as reference model for the investigation of mechanisms of staged competence learning in humanoid robots or even biological systems.