Robust cautious data driven control with guaranteed mean square stability

The paper presents a cautious and robust approach for data driven control synthesis. It proposes to parameterize a closed-loop LTI output predictor by Least Squares (LS) estimated and stochastically uncertain Markov parameters, completely characterizable by measured input and output (I/O) data. Direct embedding of I/O data, carrying uncertain Markov parameter information, into the robust infinite horizon controller design method does not only guarantee mean square stability of the closed-loop system under stochastic model uncertainties, but also reject the effect of disturbance term over a pre-defined performance output. Example shows the effectiveness of the elaborated method.