Plant input-mapping-based predictive control of systems through band-limited networks

Abstract

A new control method for network-based control systems, with guaranteed closed-loop stability, is proposed. The method substantially enhances the conventional networked model-based predictive control (NMBPC) schemes. In conventional NMBPC methods, associated with a pre-specified range of possible time delays, a vector of stabilising control signals is determined and transmitted to the plant side of the network, where, based on the actually occurred time delay, just one entry of the control vector is selected and applied to the plant. In practice, stability issues may arise when the network time delay is not small enough. A modified method, using the plant-input mapping (PIM) discretisation technique, with the possibility of guaranteeing the closed-loop stability is introduced. In order to alleviate the deficiency of some existing transfer function-based methods, which implicitly assume a non-realistic zero initial condition at the outset of every sampling instant, a state-space representation is proposed. Simulation studies on well-known benchmark problems demonstrate the effectiveness of the proposed PIM-based NMBPC method.