Predictive control with embedded feedback linearization for bilinear plants with input constraints

This paper proposes a combined control approach for bilinear dyadic plants consisting of a feedback linearization and of a predictive controller. The embedded feedback linearization allows to restate the predictive problem as a linear one, while the predictive controller can be designed either to improve performance or to avoid reaching input bounds, which are known to be a possible cause of instability for standard feedback linearization. Due to the embedded feedback, however, input constraints on the physical plant are projected into state-dependent constraints on the output of the predictive control. The design of the controller under these constraints is possible either using a minimum-time control design approach or by one-step predictor with time-varying constraints. The final implementation results into a multirate system. Simulations are used to illustrate the operation of the system and the different role of the predictor according to the sample rate ratio, as well as other implementation parameters.