A0808
Title: Moment-driven predictive control of mean-field collective dynamics
Authors: Chiara Segala - Università della Svizzera Italiana (Switzerland) [presenting]
Abstract: The aim is to present a feedback control strategy for large-scale systems of interacting agents governed by nonlinear dynamics. Starting from an agent-based model with nonlocal interactions, the mean-field limit is considered to derive a tractable control formulation. The approach combines linearization techniques and Riccati equations to compute suboptimal feedback laws embedded in a model predictive control framework. Controls are updated using macroscopic quantities, specifically the first and second moments of the agent distribution. This circumvents the complexity of solving high-dimensional Hamilton-Jacobi-Bellman equations. A feedback law is constructed for the linearized mean-field model and iteratively refined based on the evolving moments. The resulting control strategy is efficient, scalable, and robust to partial observations. Theoretical performance estimates guide the selection of linearization points and update frequency. Numerical results show the effectiveness of the method in stabilizing self-organizing behaviors, with applications to opinion dynamics and collective motion.
