A Dispersed Autonomy Architecture for Information-Gathering Drone Swarms

Teams of cooperating unmanned aircraft can perform spatial-temporal sampling of complex phenomena at scales not achievable by single platform sensor systems. In highly-dynamic environments, like severe storms, the sensing platforms must adapt in response to the evolving phenomena. Because onboard computation is limited, a dispersed computing architecture enables inclusion of high-performance models and computationally expensive algorithms in the decision-making loop. This paper presents the design and implementation of components of a dispersed autonomy architecture for information-gathering drones. Components of this architecture are described and demonstrated for several field campaigns, including the sampling of severe supercell thunderstorms by loosely coordinated UAS teams in order to study tornado formation.