Mixed human-robot team navigation in the GUARDIANS project

Abstract

A theoretical framework for generating navigation behaviour patterns in mixed human-robot groups in complex environments is proposed. This framework represents an essential part in the development of a multiple robot-human system for assisting fire-fighters in search and rescue operations in the GUARDIANS project. In order to produce the desired behaviours an artificial potential field method has been developed. We distinguish a three classes of agents: robots, humans and obstacles, and apply different potential functions to them. Depending on the situation, we switch from one function to another; this allows to generate desired behaviour patterns as well as to avoid certain local minima. Typical behaviour patterns are singled out and their stability is discussed. Stability analysis is based on geometric considerations, that permits to avoid bulky computations and provide graphic demonstrations of convergence. The proposed framework can be used in other robotic applications where a group of heterogenous agents is deployed.