Vertical hierarchical MPC for constrained linear systems

• Autores: Justin P. Koeln, Vignesh Raghuraman, Brandon Hencey
• Localización: Automatica: A journal of IFAC the International Federation of Automatic Control, ISSN 0005-1098, Nº. 113, 2020
• Idioma: inglés
• Texto completo no disponible (Saber más ...)
• Resumen

  • A hierarchical Model Predictive Control (MPC) formulation is presented for discrete-time linear systems with state and input constraints. A vertical hierarchical controller, with one controller per level, reduces the computational burden associated with the solution of centralized MPC having long prediction horizons and short time steps. To guarantee satisfaction of state and input constraints in the presence of both known and unknown disturbances, a robust MPC formulation is used at each level while waysets are used as a novel coordination mechanism between controllers at different levels. These waysets are implemented as terminal state constraints on lower-level controllers and are computed on-line based on the optimal state trajectory of upper-level controllers. To achieve the computational efficiency necessary for on-line calculation, waysets are represented as constrained zonotopes. State and input constraint satisfaction is proven for a hierarchical controller with an arbitrary number of levels and two numerical examples demonstrate the key features, performance, and scalability of the approach.