Designing response surface model-based run-by-run controllers: a worst case approach

This paper presents a framework for carrying out run-by-run (RbR) control via a deterministic worst-case approach. In particular the RbR controller developed tries to minimize the worst-case performance of the plant. This yields a methodology for handling uncertainty. A consequence of using the deterministic approach is that we no longer need any assumptions on the statistics of the noise. Rather, what we require is that the noise be bounded. Thus, we can also deal with non-Gaussian and correlated noise. We provide results comparing the performance of the controller to a recursive least squares (RLS) based controller.