Distributed Control meets Machine Learning

This project will develop machine learning algorithms and theory to build data-driven dynamic models that are suitable for distributed control. In order to apply methods from distributed control, a system model needs to exhibit some sort of sparsity structure, typically when inferring models from data, incorporating structural constraints makes the problem intractable. In this project we will seek to design ML algorithms that impose structured model realizations. It is important that we are able to characterize the sample complexity and the model error.

This project is best suited to a student with a strong background in convex optimization and control theory.