Portraits of self-organization in fish schools interacting with robots

In this paper, we propose an enabling computational and theoretical framework for the analysis of experimental instances of collective behavior in response to external stimuli. In particular, this work addresses the characterization of aggregation and interaction phenomena in robot–animal groups through the exemplary analysis of fish schooling in the vicinity of a biomimetic robot. We adapt global observables from statistical mechanics to capture the main features of the shoal collective motion and its response to the robot from experimental observations. We investigate the shoal behavior by using a diffusion mapping analysis performed on these global observables that also informs the definition of relevant portraits of self-organization.

► A data-driven approach is proposed to identify pattern formation in complex systems.
► Time-series of collective behavior induced by engineered stimulation are analyzed.
► Real data are obtained from observations of a fish school interacting with a robot.
► Aggregation and interaction patterns can be described through global observables.
► Diffusion mapping analysis can be used for model reduction and interpretation.