| Introduced over two decades ago, model predictive control (MPC) is nowadays arguably the most widely accepted advanced control design technique in control of industrial processes. It is featured by invoking system input and output constraints into system regulation and guaranteeing the closed-loop stability for nominal MPC systems. However, researchers currently notice that two main barriers hinder the further development of MPC: one is the ubiquitous model uncertainties of industrial processes; and the other is the implementation efficiency of MPC controllers. To improve its adaptability, this thesis proposes a novel MPC scenario---explicit robust model predictive control.; One of the contributions of this thesis is to separate MPC optimization from online implementation, and convert MPC design into multiple parametric sub-quadratic programming (mp-SQP). It is shown that the analytic solution to mpSQP problems can be represented by a set of piece-wise affine functions associated with state space partitions. Consequently, online MPC implementation is simplified as an affine function evaluation. Thanks to a novel prediction pattern introduced in this thesis, no high order uncertain terms occur in the MPC optimization, and the critical challenge of finite horizon robust MPC, high computational complexity; is solved skillfully.; As a natural extension of explicit RMPC, robust moving horizon state observation (RMHSO) is also covered in this thesis. The essential point that distinguishes RMHSO from conventional state observation is that RMHSO explicitly combines physical state constraints with the robust observer formulation. This thesis develops two offline RMHSO algorithms, namely, RMHSO with the forward open-loop prediction and RMHSO with the recursive closed-loop prediction. Roughly speaking, the former is less time-consuming than the latter, but the later is less memory-consuming than the former.; Keywords. MPC, robust MPC, robust moving horizon state estimation, stability, robustness, recursive closed-loop prediction, affine function control. |
