Position Paper on Scalability of Evolutionary Computation

We have been researching methods developed for designing behaviors and a directed cyclic graph for an optimized behavior selection mechanism for autonomous mobile robots to use in unoccupied aerial vehicles (UAVs) performing swarm based automatic target recognition. Our work is based on managing level of behavior, where behavior algorithms that are initially developed using evolutionary computing methods in a relatively lowfidelity, disembodied modeling environment can be migrated to high-level, dynamic, complex embodied applications. We will demonstrate our concept using adaptive waypoints, which allow navigation behaviors to be ported among autonomous mobile robots with incremental adaptation, different degrees of embodiment, and staged optimization. We believe our evolutionary algorithms scale to provide useful mission goal tasks.