Dual-Mode-Based Model Predictive Control for Markovian Systems With Improved Optimizing Prediction Dynamics
针对离散时间不确定马尔可夫跳变系统，提出一种基于观测器的输出反馈双模控制策略，通过改进优化预测动力学平衡初始可行性、控制性能和计算效率，并用宏观经济系统仿真验证有效性。
This article investigates the model predictive control (MPC) problem for discrete-time uncertain Markovian jump systems (MJSs) via an improved optimizing prediction dynamics (IOPD) approach. To address immeasurable system states, an observer-based output feedback controller is designed within the MPC framework, accompanied by a novel dual-mode control strategy that optimizes the tradeoff among initial feasibility, control performance, and computational efficiency. The first control mode, associated with the terminal constraint set, is derived from an off-line infinite-horizon optimization problem. The second mode, which steers the system state toward the terminal set within a prescribed time, is determined via online optimization, where dynamically structured perturbations expand the initial feasible region of system state, reduce computational burden, and improve closed-loop performance. The challenges posed by immeasurable states and nonlinear variable coupling are systematically resolved using matrix decomposition and parameter transformation techniques. Sufficient conditions are established to guarantee the recursive feasibility of the IOPD-MPC algorithm and the mean-square stability of the closed-loop system. Numerical simulations on a macroeconomic system validate the efficacy of the proposed method.