Robust Least-Squares Optimization for Data-Driven Predictive Control: A Geometric Approach

The recent paper on arXiv introduces a geometrically robust least-squares optimization technique that extends traditional methods in predictive control. This innovative approach interprets least-squares as enforcing approximate subspace inclusion between measured and true data spaces, addressing uncertainties through a geometric lens. The study's application to robust finite-horizon linear-quadratic tracking demonstrates its effectiveness, yielding stronger robustness and favorable scaling in scenarios with small uncertainty. The closed-form solution for the inner maximization not only enhances computational efficiency but also provides a clear geometric interpretation, marking a notable advancement in the field of data-driven predictive control.