A developmental approach accelerates learning of joint attention

This paper argues how robot learning can be accelerated by a developmental approach that changes the capability of the learning robot (robot development) and/or the complexity of the environment (environmental development). "Robot development" means that the perceptual, cognitive, and behavioral capabilities of the learning robot change from immature to mature states in accordance with learning process. On the other hand, "environmental development" means that the complexity of the environmental structure changes from simple to complicated ones in accordance with learning process. Then, the developmental approaches can be categorized according to the relationships between what does develop and what does make the development. A case study of developmental robot learning for joint attention in which a caregiver can be regarded as a part of the environment is given to show how it can accelerate the learning. Finally, future work and discussion are given.