Dynamic Categorization of Explorative Behaviors for Emergence of Stable Sensorimotor Configurations

A prerequisite for autonomous acquisition of novel behaviors would be the ability to evolve internal representation of embodied interaction structures, in other words, sensory-motor categorization. It requires a unified approach to categorical learning of spatio-temporal information. This paper proposes a novel artificial neural architecture for such processing. It is based on a ‘spiking neuron’ model and a temporal learning rule, and is capable of spatio-temporal association and categorization. The architecture has been applied to an example of visuo-ocular interaction with its surrounding by a binocular active vision system. Its sensory-motor interactions are triggered by a set of a priori reflexes and the neural architecture successfully categorized resulting structure of the interaction. Results from simulation experiments are presented.