Experiments in human multi-robot systems.

This thesis seeks to determine the constraints on the possible coordination control for systems composed of a human and group of mobile robots. In particular the constraints imposed by realtime action, robustness and human safety. We intend to put forward the findings on the first (to the best of the author's knowledge) real-world human / multi-robot system that achieves a complex navigation task.The focus is on the development of an autonomous robotic system to create a flexible formation of a group of mobile robots around a human in an unconstrained environment. A suitable environment, robot platforms, control architecture and experimental set-up are described.The robots are virtually linked to each other by application of artificial potential forces that create and maintain a flexible formation. The robots accompany a human, who is an integral part of the team, through a known environment which may contain obstacles. Each robot is capable to detect the position of the robot peers, the human, and obstacles within the sensor range. All robots attempt to maintain specified distances to the human and to every other member of the group, as a result of attractive / repulsive forces. The procedure guarantees cohesion of formation whilst avoiding collisions among participating components.The control procedure is achieved by generating a minimal generic control model and next devising the required robot controller. By virtue of our methodology, we first test and refine our controller through simulation, followed by a set of experiments carried out in real-world environments to quantify the performance and reliability of the method. We provide performance metrics that illustrate: (a) comparison of the utility of different control methods, (b) examination of tradeoffs between different metrics, and (c) details the method of comparison for future research in this area. The refined model / controller, based on these metrics, is shown to be more successful and reliable.