Robust Data-Driven Output Feedback for Systems With Unknown Input Delays
In a history-feedback realization, an input delay increases memory without increasing the number of unknown physical coefficients. We use this structure to design output-feedback controllers for discrete-time systems with one input, one output, and known undelayed order, using noisy input-output data. For a known delay, we derive a necessary and sufficient linear matrix inequality for common-quadratic stabilization of all models consistent with the data. Optimizing its equation-error bound gives a robustness margin for the measured record. For an unknown constant delay in a known interval, we characterize noiseless delay ambiguity and give a sufficient condition for one controller to stabilize all models for every delay consistent with the error bound. Under stated excitation and nondegeneracy assumptions, this procedure is feasible for all sufficiently small equation errors. Numerical examples compare robustness at equal controller memory and show that certified stabilization can succeed while distinct delay models remain compatible with the same data.