Preview and feedforward in model predictive control: conceptual and design issues

Model Predictive Control has been, and continues to be, one of the major success stories of advanced control in industry. This is a result of several factors including its ability to deal with constraints. Early implementations were concentrated in the process industries. These implementations were characterised by several distinctive features including long sample periods and constant set-points. However, new techniques and faster sampling rates have opened the door to new classes of applications where neither of these earlier features apply. In this paper we re-examine the question of reference tracking in MPC and formulate a novel strategy (termed PF-MPC) aimed at problems where the reference is not constant. The strategy combines both preview and feedforward of the reference signal, with which we seek to improve nominal MPC tracking performance. This preliminary study stands as the conceptual basis for a future robust output-feedback MPC strategy focused on improving tracking of time-varying references.