Detecting Swarm Degradation: Measuring Human and Machine Performance.

Swarms comprise robotic assets that operate via local control algorithms. As these technologies come online, understanding how humans interact with these systems becomes more important. The present work replicated a recent experiment aimed at understanding humans’ competence in identifying when and the extent to which swarms experience degradation (defined as assets breaking from consensus), as asset loss is expected in deployed swarm technologies. The present work also analyzed cluster formation in swarm simulations and explored its relationship with actual degradation. The present work replicated past findings showing people are not competent in detecting and estimating swarm degradation in flocking tasks. However, the cluster analysis showed clusters formed in simulations correlate with swarm reliability. Future work ought to expand investigations of methods to optimize cluster analysis techniques for real-time use. The implications of this work provide suggestions to interface designers on features to display to operators in human-swarm interaction.