Dresden 2026 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 57: Statistical Physics of Biological Systems IV (joint session BP/DY)
DY 57.9: Vortrag
Freitag, 13. März 2026, 12:00–12:15, BAR/SCHÖ
Visual-based Collective Shepherding in Swarm Robotic System — •Yating Zheng1,2 and Pawel Romanczuk1,2 — 1Department of Biology, Humboldt Universität zu Berlin, Berlin, Germany — 2Research Cluster of Excellence 'Science of Intelligence', Berlin, Germany
Collective shepherding presents a rich example of two interacting multi-agent systems coupled through non-reciprocal interactions. While most existing models assume that shepherd agents have global knowledge of the flock-an unrealistic premise for physical or biological systems-we introduce a vision-based, locally interacting model that captures the essential physics of shepherd-flock coordination. The model produces robust, self-organized behavior among shepherds without explicit communication, and we analyze how key control parameters, such as flock size and the number of shepherds, shape the resulting dynamics.
The framework also performs effectively in more challenging regimes, including the manipulation of non-cohesive agents and passive (non-self-propelled) agents, demonstrating its broad dynamical applicability. We further validate the model on a mixed-reality swarm-robotic platform, where physical robots successfully shepherd a virtual flock.
Overall, these results provide a minimal yet powerful physics-based description of multi-agent herding using only local visual information, offering insight into non-reciprocal collective behavior and enabling scalable real-world implementations in swarm robotics.
Keywords: collective shepherding; heuristic mechanism; visual-based model; swarm robotics