Multi-Role UAV Swarm Behaviors for Wide Area Search using Emergent Intelligence

This work continues a line of research examining the relationship between drone sensor swarm characteristics and performance of the swarm as a holistic system. A swarm of autonomous Unmanned Aerial Vehicles (UAVs) was implemented in simulation to cooperatively perform a Wide Area Search for survivors in a post-disaster scenario. The sensor swarm was controlled through a behavior-based artificially intelligent paradigm whereby each UAV executes one of three sets of simple, uncoupled search behaviors - Social Searcher, Antisocial Searcher, Relay. The relationship between proportions of these behavior sets and time to find 90% of survivors was examined in swarms of 10-50 UAVs. 100 different swarm size and role distributions were generated using Design and Analysis of Computer Experiments (DACE) methodology. 1000 (100 swarm size x 10 replications) simulation runs revealed an effective distribution pattern which drones could follow to automatically reallocate roles based on current swarm size. Simulations also revealed multiple configurations of role distributions at various swarm sizes that could consistently locate over 90% of survivors in under 20 minutes.