From Boundary Crossings to Global Connectivity: A Minimal Mechanism in Structured Agent-Based Landscapes
This study investigates a minimal mechanism through which local mobility heterogeneity producesglobal reconfiguration in structured agent-based systems. Agents move in a multi-attractorlandscape, where a small fraction exhibits higher-mobility exploratory dynamics while theremainder remain locally constrained. By systematically comparing random-walk exploration,interface-sensitive dynamics, novelty-biased exploration, and a flat-landscape control, I isolatethe conditions under which large-scale connectivity emerges.As the fraction of exploratory agents increases, the system transitions from a fragmentedregime to an increasingly connected transition network. Event-level analysis shows thatconfigurational switching is strongly localized near inter-attractor boundaries, indicating thatinterfaces act as critical gateways through which transitions occur. These localized eventsaccumulate over time, generating an expanding network of transitions that progressivelyintegrates the landscape.Importantly, the core effect persists under minimal random-walk exploration, demonstrating that neither optimization nor goal-directed behavior is required. By contrast, whenthe landscape structure is removed, connectivity becomes operationally trivial and theboundary-mediated mechanism disappears. This comparison shows that heterogeneous mobility becomes systemically effective only in the presence of a structured landscape. Large-scalereconfiguration emerges from the interaction between movement heterogeneity and spatialconstraints.These results identify a minimal generative principle for global connectivity in agent-basedsystems. Large-scale reconfiguration can arise from simple stochastic dynamics, providedthat movement occurs within a structured environment with meaningful interfaces. Thisshifts the focus from agent-level strategies to system-level organization, highlighting the roleof boundaries as mediators of emergent connectivity
Authors
- Fabio Nelli
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-05-12
- DOI
- https://doi.org/10.5281/zenodo.20145381
- Primary Topic
- Modular Robots and Swarm Intelligence
- Type
- preprint