Advisor — Master Thesis "Active Interconnections as a Mechanistic Model for Collective Behavior"

Student

Bao Duc Cao

Abstract

Collective behavior in animal groups arises from local interactions under limited perception and physical constraints. However, many existing models rely on idealized particles with simplified sensing and unrestricted motion, making it difficult to explain how coordinated patterns emerge from realistic sensorimotor processes. In this thesis, we develop a mechanistic framework based on Active InterCONnect (AICON) that integrates perception, internal state estimation, and motor feasibility into a model of collective behavior. Agents perceive neighbors through bearing and apparent-size cues within a limited field of view and maintain internal state estimates that explicitly account for uncertainty and prediction over time. Rather than prescribing fixed alignment or attraction rules, the framework generates coordinated motion through simple goal-directed gradient descent defined over internally estimated quantities. Through systematic simulations, we reproduce a range of collective patterns, including polarized motion, milling, and subgroup formation. A global sensitivity analysis further demonstrates how sensory noise, field-of-view constraints, estimation uncertainty, and motor limits jointly shape macroscopic organization. By linking biologically grounded perception with physically feasible actuation, this work advances a more embodied and mechanistic understanding of collective behavior and supports the development of decentralized robotic systems operating under realistic constraints.

Related Publication

Publication pending.

More Information

More information can be found here.