(473e) Gray-Box Modeling and Feedback Control of Active Matter

Feedback control is essential to the performance of dynamical systems, helping to drive nonequilibrium systems from one state to another. In this work, we apply feedback control to active and living matter by using a combination of light actuation, simulations, and data-driven modeling. We demonstrate the use of “gray-box models” that incorporate partially-known dynamics and approximate the difficult-to-model terms (e.g., many-body interactions) using a neural network. We generate training data based on particle-based simulations and light-triggered motile bacteria, and learn the constitutive equations that enable closure of the governing equations. We then use our model in Model Predictive Control (MPC) to manipulate the continuum density fields of interacting active matter under external actuation. We use the concepts of observability and time-delay embedding to model, forecast, and control the complex behaviors of active matter using continuum density data alone.