Proximo-Distal Competence Based Curiosity-Driven Exploration

Learning constraints and guiding mechanisms are involved since the very beginning of the infant development. Allowing a progressive and open-ended scaffolding of new skills, these mechanisms have been described as crucial, by psychologists and neuroscientists. Developmental heuristics presented here are directly inspired by the ability to control the growth of complexity of both exploration and learning in human children. More precisely, we focus on intrinsic motivations guiding mechanisms, responsible of spontaneous exploration, and on maturational evolution of the neural and muscular systems, that progressively allow the organism to control novel muscles, and thus, to increase its number of degrees of freedom (Lungarella and Berthouze, 2002). Therefore, we present a system using both self-motivation, and neuro-physiological maturation in an integrated computational mechanism, that aim to guide a robot, to gradually explore and learn its sensorimotor space.