Pseudo-Linearisation-based Model Predictive Controller Design Using an Ultra-Local Model

Pseudo-linearisation-based model predictive control (MPC) which combines a data-driven control and model-based control, is proposed to improve transient response for rapidly changing reference signals. The conventional MFC cannot be applied as the inner loop of the controller with MPC as the outer loop because the dynamic characteristic of MFC is always equal to one. To solve this problem, the conventional MFC was extended to a model-matching-type MFC by adopting a proportional-leading-type MFC as the inner loop for the proposed method. In addition, virtual reference feedback tuning was introduced to optimise an important parameter of the MFC. Based on the experimental results for artificial muscle control with strong asymmetric hysteresis characteristics, the proposed method improves the transient response by rapidly changing the reference signals and is a practical approach that can be applied to nonlinear systems because it does not require system identification or a model structure.