Heterogeneous UAV Multi-Role Swarming Behaviors for Search and Rescue

This work examines the relationships among drone swarm characteristics and how they affect swarm performance. A swarm of autonomous Unmanned Aerial Vehicles (UAVs) is implemented in simulation to cooperatively gather situational awareness data during the first few hours after a major natural disaster. The swarm is controlled by prioritized sets of relatively simple, reactive behaviors, the totality of which result in the emergence of a form of collective intelligence. The behaviors are designed to optimize the coverage area of a simulated camera mounted on each UAV. The behaviors are implemented in simulation on swarms of sizes from 10 to 50 UAVs. The UAVs are each assigned one of 3 different roles, or “personality types”: Social Searcher, Antisocial Searcher, or Relay. 100 different size and personality type distribution configurations were compared over 1000 simulation runs to determine explore the effect of swarm parameters on search success over the target area. The results show that the swarm is successful in locating over 90 % of survivors in less than 40 minutes using the most effective personality type distributions. Results also show that, within in the context of the particular simulation scenario, the addition of any number of Relay UAVs tends to decrease performance of the swarm when compared to a distribution that does not include any Relays. More generally, the work demonstrates the value of the novel approach of applying design of experiment principles to simulation experiments in order to explore optimal swarm configurations.