Functional concept acquisition using action schemata

This paper discusses unsupervised concept acquisition in autonomous agents. Autonomous agents build their knowledge from action and perception in their environment. A structure inspired in Piaget´s schema mechanism was used in order to represent functional concepts, that is, concepts related to conditions, actions and results. This kind of mechanism was first implemented by Dresher (1992). This paper presents a new approach that uses a kind of competitive neural network (the Schemata) to find the condition/action/result correlation when the concepts are presented as fuzzy signals