Multi-vehicle Control and Autonomy for Swarming Quadrotors

This talk will present a cooperative-control framework designed to enable swarms of quadrotors to autonomously detect moving people and vehicles in an uncertain and complex environment. The talk will first describe inspiration from prior work in swarming and pursuit behavior in wild malarial mosquitoes. Then I will describe a distributed control for concurrent mapping and search. Swarm agents have a finite field of view and they collaborate to map the environment represented by a graph. Nodes of the occupancy graph indicate the target likelihood ratio and inform the swarm’s search strategy. The evolving priorities of the swarm are represented by a spatially varying sampling priority surface that is used to identify tasks of high value. The candidate tasks are assigned to agents in the swarm using a consensus-based auction algorithm. Ongoing efforts seek to demonstrate the swarming framework using a multi-vehicle testbed under development at the University of Maryland’s outdoor netted Fearless Flight Facility. To enable reliable outdoor flight in wind, we are investigating improvements in attitude stability that are obtained by equipping each quadrotor with an onboard wind-velocity probe.