The Design Challenges of Drone Swarm Control

摘要

It is widely recognised that swarms are the likely next step for Unmanned Aerial Vehicle (UAV) or drone technology. Although substantially increased autonomy for navigation, data collection and decision-making is very much part of the "collective artificial intelligence" vision, this expected development raises questions about the most productive form of interaction between the swarm and its human oper-ator(s). On the one hand, low-level "micro-management" of every unit clearly nullifies many of the advantages of using swarms. On the other, retaining an ability to exercise some control over the swarm's objectives and real-time behaviour is obviously paramount. We present two families of control methods, direct and indirect, that we believe could be used to design suitable, i.e. simultaneously intuitive, easy to use, powerful and flexible, Graphical User Interfaces (GUI) that would allow a single operator to choreograph a swarm's actions. Simulation results are used to illustrate the concept and perform a quantitative performance analysis of both control methods in different scenarios. Human factors aspects related to drone swarm control are identified and both control methods are discussed from the human operator's usage point of view. We conclude that the direct approach is more suitable over short time-scales ( "tactical" level), whilst indirect methods allow to specify more abstract long-term objectives ( "operational" level), making them naturally complementary.