Spatial Self-Organization in Animal Groups and Human Crowds

Self-organization processes enable animal groups and human crowds to coordinate their behaviors and achieve collective tasks that are far beyond the capabilities of any single individual. A manifestation (among many others) of the ‘collective intelligence’ of crowds is the spontaneous lane formation in densely populated pedestrian corridors, which increases the flow efficiency. These puzzling phenomena can now be experimentally investigated in great detail thanks to sophisticated tracking systems which make the identification and quantification of the behavioral interactions among individuals possible. Experiments then feed computer models which describe each individual’s behavior and response to the surrounding agents. These Individual Based Models are useful to decipher the mechanisms that underlie the processing of information at the level of a single individual and give rise to the observed collective response. In this talk, we will provide examples of this complex systems approach from animal groups (social insects, fish, mammals) or human crowds and will show how it can provide a deep insight into the behavioral mechanisms involved in collective intelligence

A particular aspect of the complex systems approach lies in the determination of Continuum Models that describe swarms, schools, herds or crowds through average quantities such as the number of individuals per square meter at a given location and a given time. These quantities obey partial differential equations which bear strong similarities with the equations of a gas like air or a fluid like water. Continuum models give access to large-scale observables such as the motion of the crowd boundary, the direction and intensity of the flow or the amount of ‘turbulence’. These observables are easily identified on the experimental data and can be used to calibrate the coefficients of the Continuum Models. The establishment of a mathematical correspondence between the Continuum and Individual-Based Models provides a systematic link between large scale observables and the individuals’ behavior. This talk will report examples where this correspondence can be performed and improvements that can be expected from this approach from the view points of data analysis and model predictability.

Joint work with Guy Theraulaz (CNRS & University of Toulouse, Centre de Recherches sur la Cognition Animale)