Embedded Subspace Predictive Control, a Model Based Design Approach

This paper will present a model-based methodology for implementing an embedded, data-driven, subspace predictive controller. The control algorithm captures, in real-time, the characteristics of the plant and self-tunes through process modifications. A data-driven subspace approach is used, in which predictors are calculated directly from input and output data projections. Coding and performance aspects are discussed and tested in Model-In-the-Loop, Software-In-the-Loop and real application.