ABSTRACT
Model predictive control (MPC) was proposed in the 1970s first by industrial circles. The first method obtains the detail description of the system. The second method developed to the subject of "system identification." Models can be classified into two categories: one is the physical model and the other is the mathematical model. A linear system satisfies the superposition principle, which makes the mathematical computations much more convenient and a general closed-form solution exists. In synthesis approaches, when the uncertainty is considered, feedback MPC (i.e., MPC based on closed-loop optimization, where a sequence of control laws are optimized) is better than open-loop MPC (i.e., MPC based on open-loop optimization, where a sequence of control moves are optimized) in performance (feasibility, optimality). In Constrained nonlinear quadratic regulator (CLNQR), if the receding horizon optimization is applied, i.e., the outside mode controller of CNLQR is solved at each sampling instant and the closed-loop system is not necessarily asymptotically stable.
