ABSTRACT This paper presents a distributionally robust tube‐based model predictive control method for nonlinear systems affected by external random disturbances. The reference trajectory is computed by solving a rolling optimization problem for the nominal system, while an auxiliary controller with chance constraints is solved in real time to guarantee that the trajectory of the disturbed system remains within an invariant tube centered on the reference trajectory. Since the distribution information of disturbances is unknown, a distributionally robust ambiguity set is constructed based on disturbance samples collected online to reformulate the chance constraints, thereby ensuring that the state of the disturbed system satisfies the state constraints with a certain probability. It is demonstrated that this method can ensure the disturbed system remains stable, with its state staying near the reference trajectory. Furthermore, its robustness and compliance with constraints have been validated on a nonlinear continuous stirred‐tank reactor.
Yan et al. (Wed,) studied this question.