Robust Data-Driven Control of LPV Systems With Safety Guarantees

TY - JOUR

T1 - Robust Data-Driven Control of LPV Systems With Safety Guarantees

AU - Zhou, Lingan

AU - Liu, Wenjie

AU - Li, Yifei

AU - Wei, Yuzhou

AU - Wang, Gang

AU - Sun, Jian

N1 - Publisher Copyright: © 2025 John Wiley & Sons Ltd.

PY - 2026/2

Y1 - 2026/2

N2 - Direct data-driven control approaches based on the fundamental lemma by Willems et al. offer a promising alternative to model-based approaches by bypassing explicit system identification. However, their extension to linear parameter-varying (LPV) systems presents challenges due to the scheduling-dependent dynamics and the need for accounting for safety constraints. This paper proposes a direct data-driven control framework for unknown LPV systems that guarantees pointwise-in-time safety constraints via semi-definite programming (SDP). By leveraging input-state-scheduling data and employing Petersen's lemma, we develop a tractable parameterization of admissible LPV trajectories and reformulate state constraints as conditions on positively invariant (PI) sets. The framework is extended to handle both offline and online process disturbance, with a reduced-complexity SDP formulation introduced for disturbance with known spectral characteristics. Numerical results validate the effectiveness and robustness of the proposed approach.

AB - Direct data-driven control approaches based on the fundamental lemma by Willems et al. offer a promising alternative to model-based approaches by bypassing explicit system identification. However, their extension to linear parameter-varying (LPV) systems presents challenges due to the scheduling-dependent dynamics and the need for accounting for safety constraints. This paper proposes a direct data-driven control framework for unknown LPV systems that guarantees pointwise-in-time safety constraints via semi-definite programming (SDP). By leveraging input-state-scheduling data and employing Petersen's lemma, we develop a tractable parameterization of admissible LPV trajectories and reformulate state constraints as conditions on positively invariant (PI) sets. The framework is extended to handle both offline and online process disturbance, with a reduced-complexity SDP formulation introduced for disturbance with known spectral characteristics. Numerical results validate the effectiveness and robustness of the proposed approach.

KW - LPV system

KW - data-driven control

KW - safety-critical control

UR - https://www.scopus.com/pages/publications/105014946377

U2 - 10.1002/rnc.70185

DO - 10.1002/rnc.70185

M3 - Article

AN - SCOPUS:105014946377

SN - 1049-8923

VL - 36

SP - 1020

EP - 1032

JO - International Journal of Robust and Nonlinear Control

JF - International Journal of Robust and Nonlinear Control

IS - 3

ER -