Performance of a networked human-drone team: command response and interaction effects

Drones are flying machines which can be employed for sensing in different altitudes. While past research has made remarkable progress in providing autonomous drone navigation, the integration of humans and drones to collaboratively solve a task poses new challenges. Studies are needed to understand the crucial features of systems that integrate autonomous drone behavior and human-drone interactions. In this work, we study the performance of such a networked system for human-drone teams. We make use of a prototype consisting of an off-the-shelf drone with an on-board camera, autopilot, and Wi-Fi connectivity, and a custom smartphone app. The app supports the drone and performs vision-based autonomous navigation control and human gesture recognition, and provides a graphical user interface for the human. To study the benefit of collaboration, a concrete indoor use case is selected, namely book-shelf inventory. The drone scans the books in a shelf, while the human acts as a supervisor who may temporarily take over control to reposition the drone. In a measurement study, we investigate the command response delays of the networked system, effects of human intervention, and the performance of gesture-based interaction.