Skip navigation
PDF 357 KB
DOI: http://dx.doi.org/10.7551/978-0-262-32621-6-ch142
Pages 864-871
First published 30 July 2014

A Robot that Uses Arousal to Detect Learning Challenges and Seek Help

Antoine Hiolle, Matthew Lewis, Lola Cañamero

Abstract (Excerpt)

In the context of our work on dyadic robot-human (caregiver) interaction from a developmental robotics perspective, in this paper we investigate how an autonomous robot that explores and learns novel environments can make use of its arousal system to detect situations that constitute learning challenges, and request help from a human at points where this help is most needed and can be most beneficial. In a set of experiments, our robot learns to classify and recognize the perceptual properties of various objects placed on a table. We show that the arousal system of the robot permits it to identify and react to incongruent and novel features in the environment. More specifically, our results show that the robot identifies perceived outliers and episodic perceptual anomalies. As in the case of young infants, arousal variations trigger regulatory behaviours that engage caregivers in helping behaviors. We conclude that this attachment-based architecture provides a generic process that permits a robot to request interventions from a human caregiver during relevant events.