Infant-Robot Interaction as an Early Intervention Strategy

University Of Southern California, Los Angeles CA

Investigators

Abstract

Typically developing (TD) infants use movement to explore their environment, to interact, and to control their bodies. By moving, they learn which movements lead to desired outcomes: obtaining a smile from a caregiver, reaching a toy, etc. In contrast to TD infants, infants at risk (AR) for developmental delays often have difficulty moving and decreased motivation for movement. Consequently, less movement experience results in less exploration and may contribute to delays in development. The goal of this project is to develop a robot that can interact with an infant and encourage the infant to explore different types of movements. The robot will guide and engage the infant to produce movements in ranges that the infant is not experiencing on his/her own. The proposed research has the potential to advance knowledge about which aspects of movement infants can adjust and how to most effectively guide their movement to help them learn to control their bodies. End users (therapists and parents) are participating in the development of the infant-robot interaction system with the goal that it be easily adaptable to other conditions, such as autism. The project includes education and training components: 1) a K-12 STEM outreach component ties the research to educating youth about STEM and motor disabilities, and 2) graduate students in engineering, computer science, and biokinesiology receive training and mentorship to become effective interdisciplinary researchers. This research project is developing an interactive infant-robot system by: 1) analyzing existing movement data from TD and AR infants to produce expected behavior distributions for movement characteristics; 2) using those results to inform the contingent feedback for the robot to determine which movement characteristics infants can act to adjust; and 3) determining whether personalized, robot-guided feedback (motivating the infant in the desired movement) is more effective than a standard robot reward (fixed robot movement pattern). Wearable sensor and 3D vision technology is being used to quantify infant movement characteristics: limb movement quantity, duration, peak acceleration, and amplitude. A non-cumbersome light-weight wearable eye-tracker is being used to measure visual attention. The hypotheses to be tested are that: 1) infants are able to perceive and act on information to adjust the amplitude, peak acceleration, and duration of their limb movements; and 2) personalized, robot-guided feedback is more effective than a fixed robot reward at eliciting target movement behaviors.