COMBINING DIRECT DATA DRIVEN AND MODEL PREDICTIVE CONTROL WITH SET MEMBERSHIP UNCERTAINTY

To extend the always studied system with statistical uncertainty, this paper considers the control problem for the system with set membership uncertainty, thus being more suited in practice. As the most important element in applying model predictive control strategy is the unknown output predictor, the new idea of direct data driven is used to identify the unknown state in the framework of set membership uncertainty, such as process noise and measurement process. After our recursive state estimation is substituted in our own mathematical derivations, the output predictor is yielded sequently. Then one quadratic programming problem is constructed to satisfy the goal of model predictive control. An ellipsoid optimization algorithm is proposed to solve this quadratic programming problem, whose optimal control input corresponds to the center of the final ellipsoid. Finally, one simulation example is used to prove the efficiency of our proposed theories.