Low-Dimensional-Agent-Model-Based Nonlinear MPC for Flexible Vibration Suppression

In this paper, a nonlinear model predictive control strategy is proposed based on a low-dimensional agent model to suppress the vibration of flexible beams governed by an Euler-Bernoulli beam equation. Dimensionality reduction is achieved for the high-dimensional vibration data via a deep autoencoder, effectively mapping the high-dimensional spatiotemporal data into a compact temporal domain. By the sparse identification of nonlinear dynamical systems with control algorithm, the agent model is identified from the reduced vibration data and used to derive optimal control actions based on the agent model outputs over a predefined time horizon. Numerical simulations demonstrate the proposed vibration control strategy's effectiveness.