Robustness of collective decision dynamics
Naomi Leonard, Princeton University
Abstract:
Assorted multi-agent system models and feedback algorithms have been designed to converge to collective patterns that resemble groups from nature. Typically, however, the resemblance is lost when the system models are perturbed: natural group behavior is observed to be highly robust to disturbance and uncertainty even as individuals in the group are assumed to be “minimalists” in their use of feedback. I will describe methodology to investigate the role of directed interconnection topology on robustness of collective decision dynamics. I will show how the methodology is being used to evaluate performance of natural groups from observational data in an effort to identify underlying mechanisms of robust behavior.