Planning Human-Robot Interaction for Social Navigation in Crowded Environments.

Navigation is one of the crucial skills autonomous robots need to perform daily tasks, and many of the rest depend on it. In this paper, we argue that this dependence goes both ways in advanced social autonomous robots. Manipulation, perception, and most importantly human-robot interaction are some of the skills in which navigation might rely on. This paper is focused on the dependence on human-robot interaction and uses two particular scenarios of growing complexity as an example: asking for collaboration to enter a room and asking for permission to navigate between two people which are talking. In the first scenario, the person physically blocks the path to the adjacent room, so it would be impossible for the robot to navigate to such room. Even though in the second scenario the people talking do not block the path to the other room, from a social point of view, interrupting an ongoing conversation without noticing is undesirable. In this paper we propose a navigation planning domain and a set of software agents which allow the robot to navigate in crowded environments in a socially acceptable way, asking for cooperation or permission when necessary. The paper provides quantitative experimental results including social navigation metrics and the results of a Likert-scale satisfaction questionnaire.