Learning Object Functionalities in the Context of Behaviour Selection

In this paper, we propose a model that integrates behaviour selection and learning of affordances through the monitoring of internal motivations in the context of the interactions that an autonomous robot undertakes with objects in its environment while working towards its survival. To test the validity and effectiveness of this model, we have performed two different sets of experiments. The first one aimed to elucidate the question whether our robot could learn affordances by exploiting the outcome of interaction experiments through the monitoring of its internal motivations. The purpose of second one was to assess to what extent the use of previously learnt affordances in novel interactions could improve the survival of the robot.