Robust MPC suitable for closed-loop re-identification, based on probabilistic invariant sets

ANDERSON, ALEJANDRO; D'JORGE, AGUSTINA; GONZALEZ, ALEJANDRO H.; FERRAMOSCA, ANTONIO; KOFMAN, ERNESTO

SYSTEMS AND CONTROL LETTERS

ELSEVIER SCIENCE BV

Lugar: Amsterdam; Año: 2018 p. 84 - 93

0167-6911

Recently, a set-based Model Predictive Control (MPC) scheme suitable forclosed-loop re-identication was proposed. That approach solved the potentialconict between the persistent excitation of the system and the stabilizationof the closed-loop conguration by using invariant-set-stability instead ofequilibrium-point-stability. However, as a rst approach, the resulting MPCis only nominal (which in re-identication scenarios is a main drawback) anddid not exploit the knowledge of the probabilistic distribution of the excitationsignal, ensuring convergence to large regions that conservatively containthe excited system evolution. In this work, we propose a robust extension ofthe set-based MPC (for parametric uncertainty) which, furthermore, include aProbabilistic Invariant Sets (PIS) as target set, allowing this way a substantialreduction of the region where the re-identication is safely made. Simulationresults show the practical benets of the novel robust strategy.