An Architecture for Navigation of Service Robots in Human-Populated Office-like Environments*

This paper proposes a simple and efficient solution for robot navigation in officelike, indoor environments. The proposed methodology does not rely on metric localization and uses a laser range finder to detect the transitions among the nodes of a topological map of the environment, which is represented by a directed graph. The navigation between two nodes of the topological map is performed by using a vector field that is computed on the fly in function of the detected walls of the corridors, which are common in office-like environments, and of the people and other obstacles encountered in the robot’s path. This vector field is computed independently of the robot’s global path and is designed in a way that the robot’s behavior is socially accepted by the humans of the environment. Because metric localization is not necessary and the vector field computation is very simple, the method is computationally efficient, and this allows the robot to navigate at high speeds. In this paper we illustrate the methodology through real world experiments with a nonholonomic service robot navigating in an office-like environment.