System identification of emergent dynamics in multi-agent systems

Multi-agent systems, whether biological or engineered, exhibit robustness and adaptability through decentralized decision-making and local interaction rules. Understanding the underlying principles governing these systems is critical for both scientific discovery and engineering design.This dissertation investigates system identification in multi-agent systems from multiple perspectives. A heuristic test for emergence recognition is developed, grounded in a mathematical formulation, to characterize collective order from a macroscopic viewpoint. Flight trajectories of insects and birds are analyzed using graph-theoretic methods to estimate agent-to-agent connectivity. To further understand how these interactions depend on factors such as metric distance or topological neighbor order, a generalized interaction function identification framework is formulated. The analysis reveals species-specific behavioral trends such as distance-dependent attraction and velocity alignment, which provide insight into swarm coordination. Beyond biological systems, the framework is extended to analyze impulse responses in engineered systems, particularly spacecraft–debris collisions in low Earth orbit (LEO). By analyzing altitude deviations as impulse responses, a method for detecting debris impacts is proposed. Collectively, these findings advance the understanding of emergent dynamics in complex multi-agent systems and establish a unified methodology for identifying interaction mechanisms across different domains.