A Preliminary Study on a Robot's Prediction of Human Intention

Abstract

A wise man is said to be able to do the right thing at the right time. Thanks to the cognition, human beings are able to make reasonable judgments and then take specific actions accordingly. However, without explicit communication, even the most intelligent robot currently existing in the world may not be able to well understand the intentions of its human companion, not to mention its ability to act according to the human's intentions. In fact, “When to do?”, “What to do?”, and “How to do?” may not only confuse philosophers, but also often puzzle designers of robots. In this preliminary research, we introduce a role of prediction in the context of human-robot interaction. We study how a human pitches a ball by experiment and how a robot blocks the ball accordingly by simulation. The study is to investigate how a robot perceives and learns from the motion of a human participant. This paper proved that with machine learning techniques, a robot is able to accurately predict the human participant's intention of pitching. We also demonstrate how the learned knowledge can enhance the robot's intelligence and action.