Self-referential Biological Inspiration: Humans Observing Human Swarms to Identify Swarm Programming Techniques

Biology has elegantly solved numerous engineering problems that are far beyond our current collective abilities. Taking the apprentice role, with Nature as the master crafter, scientists and engineers can learn much by examining organic solutions. A biologically-inspired approach typically comes after trying unsuccessfully to tackle a specific problem using classical methods. Observing relatively simple creatures that exhibit the necessary abilities in their natural habitat often yield new and successful approaches. Turning this approach on its head, our research looks at higher-level creatures (humans) without a specific problem in hand. Our strategy is to create a crucible from which we can extract elemental components of complex behaviors to develop a toolbox of general programming techniques for multirobot control algorithms. By conducting multiple observations of human swarms performing constrained experiments, patterns of successful behavior emerge. Using these patterns, human swarm algorithms have been reconstructed and applied to simulated robots. Another benefit of using human swarms is the high-level communication available for giving commands (it is easy to tell humans what to do). Unfortunately, this can also make it difficult to reverse engineer human swarm solutions. It is hard to identify and quantify use of hand or voice signals, subtle cues from body language and facial expressions. So, to support algorithm mining, we have developed a prototype system for implementing virtual human swarms. The distributed system controls what a human at a remote keyboard sees and hears, as well as, restricting and monitoring their communication. For example, the software can enforce reduced sensor capabilities, prevent global communication, and limit possible agent actions. The virtual swarm software can record all actions and communications during an experiment facilitating algorithm extraction and the cataloging of swarm programming component behaviors.