A new approach for social navigation and interaction using a dynamic proxemia modeling

Abstract

This paper treats the problem of social navigation and human–robot interaction. In most previous works addressing this issue, the proxemia concept has been considered static with regards to the activity’s nature. Furthermore, the F-Formation type is not considered. The different activities of people can not be considered similarly. For example, the way to navigate socially without disturbing a group drinking coffee or discussing a poster is different, just like the interaction location when serving coffee or explaining a poster. The main contribution of this paper is to propose a dynamic proxemia modeling approach “DPMA” for social navigation and interaction. It is based on proxemia and spatial modeling, which allow the robot to navigate socially by considering the comfort of people, and to interact in appropriate interaction conditions. It considers the F-Formation type and the nature of the activity, additionally to the different social spaces and the navigation and interaction constraints. Then, it uses a social map to encode them. The performance and the efficiency DPMA are evaluated both in three simulation environments using the ROS Stage simulator and on a real robot used as a service robot in a mediation event example. Promising results are given in this paper.