SPC: Subspace Predictive Control

Abstract Subspace identification has proven to be an excellent system identification method under peculiar industrial situations. Model predictive control on the other hand also turned out to be a very competitive method, especially in chemical industry. In practise, the identification and the control of a system are almost always considered as two separate problems. In the present paper some remarkable analogies between subspace identification and model predictive control are uncovered. Both methods can be combined in a very elegant way to form a numerically robust and easily implementable control/identification algorithm. This is the reason why we refer to it as subspace predictive control. The main result is that the system identification step and the controller design are done simultaneously. Starting from input and output measurements of the unknown system, only a QR-decomposition followed by a SV-decomposition are required to find the controller parameters.