Developmental robot with ostensive cue sensitive learning for real-world interaction based on local contingency evaluation

Joint attention is one of the most important cognitive function for establishing successful communication in human interaction. It is even appeared during the interaction of the human infant with its caregiver, and plays essential roles for childhood development [1], [2], [3]. In the field of developmental psychology, cognitive science and even interactional robotics, studies on these behaviours have become one of the hot topics [4], [5], [6], [7]. On the other hand, some studies utilized synthetic approaches to understand the developmental process of joint attention related behaviours and social skills [8], [9], [10], [11]. In the previous work, Mahzoon et al. [12] proposed local contingency evaluation of the events for a learning robot model and showed that it can acquire simple and complex social skills during the interaction with the human caregiver model. However, it needs to gather lots of experiences to distinguish contingent events, and still non-accurate for implementing on a real world robot. In this study, we propose ostensive cue sensitive learning to solve the first problem, and cross-modal comparison of the contingency (CmC) for the second one.