Regensburg 2022 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
BP: Fachverband Biologische Physik
BP 6: Statistical Physics of Biological Systems 1 (joint session BP/DY)
BP 6.1: Vortrag
Montag, 5. September 2022, 15:00–15:15, H16
Dynamics and Fair Risk Sharing in Groups of Intelligent, Egoistic Individuals — Samuel Monter1, •Veit-Lorenz Heuthe1, Emanuele Panizon2, and Clemens Bechinger1 — 1FB Physik, Universität Konstanz, Konstanz, Germany — 2Department of Quantitative Life Science, ICTP, Trieste, Italy
Many animal species organize in social groups of fascinating complexity. The evolutionary biologist W.D. Hamilton hypothesized that the gregariousness of some animals can be explained solely from the egoistic motivation to decrease the risk of predation [1]. As a quantitative measure of this risk, he considered the Voronoi area around each animal. Many collective behavior studies try to capture this motivation by imposing interaction rules or neglect the driving motive altogether when modeling the dynamics of animals. In this study we train a swarm of individuals in a Multi Agent Reinforcement Learning (MARL) framework according to Hamilton's hypothesis, i.e. to decrease their predation risk. Thus, we gain insights into the dynamics of an ensemble of selfishly motivated individuals unbiased by any a priori assumption about interactions. We find that the individuals learn to cluster into groups which exhibit dynamic steady states resembling the behavior of natural swarms. Additionally, the predation risk is shared evenly within the groups, counterintuitive to the selfish motivation of each individual. Our findings suggest that gregariousness could indeed be driven by selfish motives in accordance with Hamilton's hypothesis.
[1] W. D. Hamilton, Journal of theoretical Biology 1971, 31, 295-311.