Nonlinear Network Dynamics for Bio-Inspired Collective Decision-Making

点击翻译按钮获取中文摘要

This project will build upon model-based investigation of group decision-making in animals, to create a rigorous methodology for bio-inspired control of multi-agent systems. It is well known that bird flocks and fish schools show a remarkable ability to choose between mutually exclusive competing options with speed, accuracy, robustness and adaptability. This project will derive a general and flexible framework for modeling collective decision making, by applying singularity and bifurcation theory to graph-based representations of the dynamics of networked systems. The mathematical framework will be validated on experimental data from observations of swarming honeybees and schooling fish. The resulting methodology will find application to design and control of, for example, mobile sensing networks, transportation networks, and power networks. These applications have significant potential societal impact in environmental monitoring, hazard response, transportation, energy, healthcare, and manufacturing. Leveraging this exciting, cross-disciplinary topic, the project will engage middle and high school girls through researcher visits into schools, and student and teacher visits to the Principle Investigator's lab.An analytically tractable, abstract agent-based model for collective decision-making dynamics that captures nonlinear phenomena observed in animal groups will be used. This will enable a new methodology for coordinated decision-making that extends current capabilities of networks of dynamic agents to reliably perform demanding tasks in complex, changing environments. A principled approach, based on singularity theory -- a robust bifurcation theory -- will be used for this model to connect biology to engineering design. Realizations of collective decision-making between alternatives will be derived and conditions for high performance proved by building in value-sensitive decision-making and addressing heterogeneity in preferences across the group. Explorations will go beyond biologically plausible parameter ranges to identify and prove conditions for high-performing engineered collective decision-making dynamics.