Optimal control

Optimal control of transport dominated problems with reduced order modeling

Optimal control for transport-dominated systems becomes tractable and practical when paired with carefully constructed reduced-order models. Transport phenomena (shocks, moving vortices, concentration fronts) concentrate energy in low-dimensional, moving coherent structures that standard global bases struggle to represent. However, nonlinear model order reduction techniques like the sPOD-Galerkin method capture those features with far fewer degrees of freedom. That dramatic reduction in state dimension cuts the cost of forward and adjoint solves by orders of magnitude, enabling faster gradient evaluation, real-time or many-query optimization, and easier incorporation of constraints and parameter dependencies. When designed and validated carefully, ROM-based optimal control preserves essential physics (stability, conservation) while making control synthesis, sensor/actuator placement, and uncertainty-aware design feasible for problems that would otherwise be computationally intractable.

Our work presents an in-depth theoretical and numerical analysis of an example transport problem applied in an optimal control context and is shown in our paper

2024

  1. OCsPOD
    Optimal control for a class of linear transport-dominated systems via the shifted proper orthogonal decomposition
    Tobias Breiten, Shubhaditya Burela, and Philipp Schulze
    arXiv, 2024