Multi agent coordination through visual and tracking interaction using system identification method

Insects, serving as model systems for swarming robotic agents, pose a challenge for fully explicating the mechanisms underlying their onboard sensing and feedback supporting such coordinated motions. This study addresses this gap by measuring untethered insects engaged in solo in-flight visual tracking tasks and employing system identification techniques to construct an experimentally consistent model of their visual tracking behaviors. Integrating measured experimental delays and their variation into a visually interconnected swarm model, the study develops theoretical and simulated solutions and stability limits. Linear stability and bifurcation analysis reveal delay distributions leading to a family of solutions with stable and unstable swarm center of mass locations. Additionally, the study experimentally records multiple flying insects tracking a moving visual stimulus, quantifying visual tracking delay variation and revealing average delays and standard deviations for solitary and group behaviors. These results quantify the role of visuomotor delay heterogeneity in flying insects and their role in supporting swarm cohesion through implicit communication. Further, the study analyzes recordings of flying honeybees in groups approaching a moving stimulus, uncovering a novel three-stage decision-making process involving pairwise interactions. Through neighborhood identification and testing against three regulation candidates, the study highlights individual decision-making capabilities, feedback rule changes, and PID interaction rules within aerial groups, offering insights with implications for mathematics, biology, and robotics. Finally, hypervelocity impact on space debris propagation is analyzed based on the multi-agent system identification technique. The primary objective is to capture and analyze debris impacts across the collector’s layers, focusing on interactions and debris transfer between layers. Advanced image processing techniques, including debris distribution analysis for each layer, were employed to evaluate the patterns and transitions of debris. Among the tested models, the second-order model structure provided the best approximation of debris distribution for single plates. This study provides valuable insights into optimizing future space debris collection systems and improving our understanding of debris dynamics in layered shielding structures. The findings contribute significantly to the fields of bio-inspired robotics, swarm dynamics, and space technology, paving the way for designing more effective multi-agent systems and resilient debris mitigation strategies.