Master's thesis · 2016
Child-Robot Interaction: Learning by Imitation using Motion Generalization
A NAO robot that learns arm movements by watching people, built to play teaching games with children with cerebral palsy.
Abstract
The work is part of a partnership between the Institute of Systems and Robotics (ISR) and the Cerebral Palsy Association of Coimbra (APCC), developing child-robot interaction for the association's children.
The system pairs a NAO robot with a computer and a Kinect RGB-D depth sensor, connected through the Robot Operating System (ROS). It detects users, learns arm movements taught by a person, reproduces a person's arm movements on the robot, recognises movements and talks through voice commands.
In imitation mode the robot mirrors the user; in learning mode the user teaches new movements, and when a movement is not yet known the robot asks for it to be repeated until it can be learned and generalised. The skills were tested with adults and children, first one by one and then together in continuous interaction.
- Degree
- MSc in Electrical and Computer Engineering
- University
- University of Coimbra, Faculty of Sciences and Technology
- Lab
- Institute of Systems and Robotics (ISR), with APCC
- Supervisors
- Prof. Paulo Peixoto and Prof. Diego Faria
- Keywords
- Kinect, NAO, ROS, motion learning, child-robot interaction
The robot at work
Video 1
Video 2
Video 3