Curiosity-Driven Development of Tool Use Precursors: a Computational Model

Studies of child development of tool use precursors show successive but overlapping phases of qualitatively different types of behaviours. We hypothesize that two mechanisms in particular play a role in the structuring of these phases: the intrinsic motivation to explore and the representation used to encode sensorimotor experience. Previous models showed how curiosity-driven learning mechanisms could allow the emergence of developmental trajectories. We build upon those models and present the HACOB (Hierarchical Active Curiosity-driven mOdel Babbling) architecture that actively chooses which sensorimotor model to train in a hierarchy of models representing the environmental structure. We study this architecture using a simulated robotic arm interacting with objects in a 2D environment. We show that overlapping phases of behaviours are autonomously emerging in hierarchical models using active model babbling. To our knowledge, this is the first model of curiosity-driven development of simple tool use and of the self-organization of overlapping phases of behaviours. In particular, our model explains why and how in-trinsically motivated exploration of non-optimal methods to solve certain sensorimotor problems can be useful to discover how to solve other sensorimotor problems, in accordance with Siegler's overlapping waves theory, by scaffolding the learning of increasingly complex affordances in the environment.