Educational Robot Kiddo Learns to Draw to Enhance Interactive Handwriting Scenario for Primary School Children

Abstract

Handwriting and drawing skills play a significant role in early childhood education. Recent studies suggest that robots currently being used in education are social robots where most of them don't write directly as the way humans do by holding a pen and moving it over a paper, besides, use an indirect medium such as touch screen as the shared environment which creates a noticeable barrier in natural interaction. In this work, we develop a new robot named Kiddo that sits on a desk, has an animatronic face, can write with a whiteboard marker, and provides a shared environment where the robot and a child can write simultaneously. The robot learns to write shapes by maneuvering its hand just like we humans do which creates a vivid image to the child interacting with it. We have used CNN based model to recognize and draw 100 shapes collected from the National Curriculum Bengali, English, and Mathematics textbook of Bangladesh. To learn new shapes from children's demonstration an artificial intelligence system was build which can generate path trajectories from a camera image. We experimented with three kids where they drew 18 known and 4 new shapes on the shared whiteboard and Kiddo was also able to recognize and draw 17 of the known shapes and learn to draw all the new shapes from the children's demonstration in both simulation and in real.