Evaluation of Navigation Commands Based on Conceptual Attributes to Enhance Human-Robot Interaction

Abstract

Natural voice commands can be used as an interaction modality to improve assistive robots. Recent developments with voice modality have failed to model the unique attributes of human communication such as incompleteness, inconsistency, uncertainty and unpredictability. Human-human conversations are volatile with uncertainty and unpredictability. These attributes should be evaluated properly in order to design more robust natural vocal command model. Aside from the language attributes, there are other factors amounts to familiarity of vocal commands. Therefore in this paper, we evaluate the unique language attributes of vocal commands related to robot navigation and other factors affecting the human robot encounters. Human study was carried out in order to identify above attributes and understand the effect on navigation commands. Further experiments were conducted to understand the differences between human-robot and human-human conversations.