The infant development timeline and its application to robot shaping

摘要

Early infancy is a time of remarkable sensorimotor learning and rapid cognitive growth. Such development offers a rich source of inspiration for models that might allow robotic systems to learn cumulatively and autonomously. This article consists of three parts. The first part introduces the key issues from a robotics perspective. In particular, we promote a constructivist approach in which synthesis provides complementary insights to analytical methods. The second part presents the findings of a detailed review of infant development, from conception to 12 months. This produced explicit timelines that display the relative ordering of emergent competencies and concomitant stages in behavior. The final part shows how this infant framework can be mapped into the robotic domain in order to construct and explore developmental algorithms for robots. This work explores the use of constraints on perception, cognition, and action in shaping emergent activity, and how constraints relate to staged behavior and competence learning. We draw on examples from our previous experiments and show how we are applying this developmental approach to an iCub humanoid robot.