ABSTRACT
Interdisciplinary approaches to the study of cognitive development benefit from the integration of child psychology experiments with computational and robotics models. Via the computational modelling of specific cognitive functions, it is possible to test developmental theories and hypotheses and generate further predictions on child development. In particular, recent developments in artificial intelligence and cognitive robotics have led to the novel approach of developmental robotics. This specifically aims to design sensorimotor and cognitive capabilities in robots by taking direct inspiration from child psychology and via the modeling of incremental stages of the development of cognitive and sensorimotor skills. Developmental robotics typically use “baby” robot platforms, such as the iCub and NAO robots, with a cognitive architecture to control the robot’s behavior. These cognitive architectures can be realized with artificial neural networks or other machine learning methods. The robot is then trained to acquire specific sensorimotor and cognitive skills, including human–robot experiments where the child robot learns from a human tutor. This chapter briefly analyses a set of examples of developmental robotics models of the acquisition of words and of the learning of numbers via gesture and finger counting. This will also show how robotics models can be directly constrained on specific child psychology experiments and theories. It will then look at ongoing developments both for scientific modeling and for applications in education and assistive robotics.
