Servicenavigation

Project Details

Back

Projekt Print View

Collective Information Processing - From Individual Sensory Inputs to Collective Motion and Decision Making

Subject Area Statistical Physics, Nonlinear Dynamics, Complex Systems, Soft and Fluid Matter, Biological Physics
Bioinformatics and Theoretical Biology
Human Cognitive and Systems Neuroscience
Term from 2016 to 2023
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 280037999
 
Final Report Year 2022

Final Report Abstract

In retrospective, during the Emmy Noether project, we were able to address the majority of research questions formulated in the initial proposal. However, due to unexpected developments, like lack of data on locusts as a model system, and new opportunities, like the access to new unique experimental data in the field, adaptations to the research plan had to be made. First, the focus shifted to fish as a the sole model system, as well as, to theoretical models on the role of perceptual and cognitive constraints on the emergence of collective movement (Rahmani, Peruani, and Romanczuk 2020; Bastien and Romanczuk 2020). Second, certain research questions, like the role of heterogeneity received less attention than planned (Jolles et al. 2020; Bierbach, Landgraf, et al. 2018), while others topics, including new ones, like the “criticality hypothesis”, moved more into the research focus (Klamser and Romanczuk 2021; Poel et al. 2022). A major breakthrough was the development of purely vision-based, individual-based flocking model (Bastien and Romanczuk 2020). We were able to demonstrate that already a minimal vision model is sufficient to generate organized collective behavior. Interestingly, in the minimal model the agents have no access to the relative distances or velocities of their neighbors, which are crucial ingredients of classical social force model. More generally, this work formulates a mathematical framework for exploration of perception-based interactions and how they differ from physical ones. and suggests novel approach for purely vision-based autonomous swarm robotic systems. We have intensely studied movement initiation in so-called startle cascades in fish schools in collaboration with experimental partners. We showed that risk is predominantly encoded in the physical structure of groups, i.e. their (visual) interaction network: The observed modulation of the collective response under risk can be traced back predominantly to the adjustment of the school structure (Sosna et al. 2019). Further, we analyzed in which parameter region the fish schools appears to operate, in particular how close to a phase transition. It has been suggested that collective information processing systems, including animal groups, should operate in the vicinity of critical points, where various aspects of collective computation become optimal (Mora and Bialek 2011). We could show that in the laboratory setting fish schools remain subcritical, trading-off sensitivity for robustness to noise (Poel et al. 2022). We were able to extend our research on collective escape response to large-scale escape cascades exhibited by schools of sulphur mollies under natural conditions. We provide evidence for benefits of the collective behavior in terms of predator deterrence (Doran et al. 2022), and suggest that under high predation threat in the wild the system operates indeed at a criticality in contrast to the safe laboratory environment (Gomez-Nava et al. 2022). We also explored the criticality hypothesis in the context of the flocking transition, i.e. transition from disordered to highly ordered group movement. Using a spatially explicit model of schooling prey responding to a predator agent (Klamser and Romanczuk 2021), we showed that while the prey system performs best at criticality, this is not due to optimal information propagation as typically assumed, but due to the dynamical structure of the group at criticality. We could also show that the critical point is maximally unstable to individual-level evolutionary adaptations due to maximal spatial self-sorting. This emphasizes, on the one hand, the crucial role of spatial self-organization, and on the other hand poses questions on the mechanism of self-organization towards criticality in animal groups. We studied the role of sensory constraints on coordination and collective information processing in agent-based model for collective behavior in complex environments with many potentially distracting environmental cues (Rahmani, Peruani, and Romanczuk 2020). Counterintuitively, large-scale coordination in such environments can be maximized by strongly limiting the sensory capacity of individuals, where due to self-organized dynamics the collective self-isolates from disrupting environmental information. We observe a fundamental trade-off between coordination and collective responsiveness to environmental cues. Our results offer important insights into possible evolutionary trade-offs in collective behavior. Further, we analyzed the properties of visual interaction networks (Poel, Winklmayr, and Romanczuk 2021), as well as the role of self-organized spatio-temporal patterns (Zhao, Huepe, and Romanczuk 2022a) in the context of collective information processing and behavioral contagion, as well as speed-accuracy trade-offs in collective decision making (Daniels and Romanczuk 2019; Winklmayr et al. 2020; Raoufi, Hamann, and Romanczuk 2021). With the assumption of a fixed-term 5-year professorship, the group will be able to continue its work on collective behavior and collective intelligence at the Humboldt University Berlin. Here, the embedding in the research cluster of excellence “Science of Intelligence” provides not only generous funding but also a uniquely stimulating and interdisciplinary environment for our future research.

Publications

 
 

Additional Information

© 2026 DFG Disclaimer / Copyright / Privacy Policy

Textvergrößerung und Kontrastanpassung