Human action prediction for human robot interaction

Abstract

Coordination can achieve higher performance when the actors are aware of each others action plans. While such action plan coordination is straightforward between robotic agents, because each robot has its plan in a digital form that is easily communicable, it is more challenging in human-robot interactions. One challenge is making the robots aware of the human action plan, which is challenging because the human action plans are not in a convenient form for communications and humans translating their plans into such a form is not time-efficient. Challenges related to human action prediction include the fact that the underlying human planning model and its properties are not known; although it may be reasonable to presume that the human actors are performing some (sub)optimal reasoning process based on the their knowledge of the state of the system. This paper studies the use of the behavioral systems modeling approach for the human action prediction problem.