Title: Equivalence of SS-based MPC and ARX-based MPC

Abstract: This paper presents the equivalence of two kinds of Model Predictive Control (MPC) based on the state-space (SS) model and the autoregressive model with exogenous (ARX) terms model. Firstly, based on the Cayley-Hamilton (CH) theorem, a SS model can be transformed into a unique equivalent ARX model, but this CH-based SS-to-ARX transformation is sensitive to process and measurement noises. We then present two more generalized SS-to-ARX transformations based on the Observer-Theory (OT) and Kalman Filter (KF). The same structure of their final transformed ARX models explains that the OT-based SS-to-ARX transformation is robust to noises. Thus, an observable linear SS model can be equivalently transformed into an ARX model by designing an observer via pole placement method. This paper suggests for the first time how to play the role of SS-to-ARX transformation, which can be used in an interpretive and adaptive MPC framework. That is, an interpretative SS model is transformed into its equivalent adaptive ARX model in offline design, and then the adaptive ARX model updates online. An AFTI-16 MPC numerical example is used to illustrate the equivalence of SS-based MPC and ARX-based MPC problem and investigate the robustness of different SS-to-ARX transformations to noises