Data-driven subspace approach to MIMO Minimum Variance Control performance assessment

A new data-driven approach is proposed for the estimation of the Minimum Variance Control (MVC) benchmark, which eliminates the need of estimating the interactor-matrix or extracting the model/Markov parameter matrices. Using the parity space, the proposed subspace approach gives equivalent estimation of the MVC performance bounds in multivariable feedback control system. The basic procedure is to identify a parity space of the system residual, instead of the process model, directly based on closed-loop data. Therefore, the MVC performance indices are estimated to make control performance assessment. The equivalence of the proposed approach to the conventional interactor-matrix based approaches for the estimation of the MVC-benchmark is proved and illustrated through simulations.