An architecture for sensate robots : real time social-gesture recognition using a full body array of touch sensors

Abstract

Touch plays a central role in social expression but, so far, research into social touch behaviors for robots has been almost. non-existent. Embodied machines have the unique capability to sense human body language, which will enable robots to better comprehend, anticipate and respond to their human companions in a natural way.This thesis addresses the novel field of sensate touch by (1) creating the first. robot with full Body sensate touch and with on-screen visualization, (2) establishing a library of salient social gestures through behavioral studies, (3) implementing a first-pass touch gesture recognition system in real-time, an(d (4) running a small pilot study with children to evaluate classifications and test the device's acceptance/utility with humans. Such research is critical path to conceiving and advancing thee use of machine touch to better integrate robots in.to human social environments.All of the above will be incorporated into the huggable robotic teddy bear at the MIT Media Lab's Personal Robotics group and makes use of the Sensitive Skins circuit design created in Dan Stiehl's Masters thesis. This implementation substantially reduces his proposed total sensor numbers and type, modularizes sensors into two uniform shapes, and extends his valuable work on a single body sections to an evaluation of sensors over the entire surface of the robot.