Data-Driven Control of Stochastic Systems: Representation, Prediction, and Optimal Control

Beschreibung

This thesis develops a unified framework for data-driven control of stochastic systems, combining the fundamental lemma of Jan C. Willems and Polynomial Chaos Expansion (PCE) to address representation, prediction, and control of stochastic systems without parametric models. It extends Willems' Fundamental Lemma to stochastic systems, enabling representing stochastic system behavior using recorded input/output/disturbance data. By condensing unmeasured or unmodeled disturbances into a residual disturbance that can be estimated and partially statistically modeled, the proposed data-driven stochastic prediction scheme ensures effective predictions with PCE-based confidence intervals. The thesis also investigates stochastic optimal control, introducing a data-driven output-feedback scheme with closed-loop guarantees.

Autor*in

Guanru Pan

Themen in »Data-Driven Control of Stochastic Systems: Representation, Prediction, and Optimal Control«

Data-driven Control Stochastic Control Optimal Control

Stimmen zu »Data-Driven Control of Stochastic Systems: Representation, Prediction, and Optimal Control«

Details

Link teilen