IHMPC estável baseado em modelos não mínimos

GONZÁLEZ, ALEJANDRO HERNÁN; SOTOMAYOR, OSCAR; ODLOAK, DARCI

Juiz de Fora, Minas Gerais, Brasil

Congreso; XVII Congresso Brasileiro de Automática, CBA; 2008

Sociedade Brasileira de Automática (SBA), miembro brasileño de IFAC

In the literature on MPC (Model Predictive Control) based on state-space model, stability is usually assured under the assumption that the state is measured. Since the closed-loop system may be nonlinear because of the constraints, it is not possible to apply the separation principle to prove global stability for the output feedback case. It is well known that, in general, a nonlinear closed-loop system with the state estimated via an exponentially converging observer can be unstable even though it is stable with state feedback. One obvious alternative to overcome the state estimation problem is to adopt an output realigned model, in which the states represents the past inputs and outputs (Maciejowski, 2002; Wang and Young, 2006). In this case, no observer is needed since the state variables can be directly measured. However, an important disadvantage of this approach is that the realigned model is not of minimal order, which makes the standard Infinite Horizon MPC formulations (González et al. 2007) difficult to apply. Here, we propose a method to properly formulate an IHMPC based on an output-realigned model, which avoids the use of an observer and guarantees the closed loop stability.