Data-Driven Output Prediction and Control of Stochastic Systems: An Innovation-Based Approach
From MaRDI portal
Abstract: Recent years have witnessed a booming interest in data-driven predictive control of dynamical systems. However, ill-conditioned solutions may occur in face of stochastic systems, causing inaccurate predictions and unexpected control behaviours. In this article, we develop new data-driven solutions to output prediction and control tasks of stochastic systems with unmeasurable states by utilizing the innovation form, where uncertainty is recast as innovations that can be directly estimated from data using subspace identification techniques. A generalized fundamental lemma is first put forward by incorporating innovations into inputs. This yields a new innovation-based data-driven output predictor of stochastic system, which helps to bypass both model identification and state estimation. Importantly, it is shown to bear equivalence to the Kalman filter-based predictor in both open- and closed-loop if innovations are "precisely" derived. Besides, a formal analysis of closed-loop stability of prediction error is presented considering imprecise innovation estimates. Based on this, a new data-driven control formulation is developed, which is shown to be equal to certainty-equivalent model predictive control with precisely known innovations. Numerical simulations are carried out to validate the theoretical results and show that the innovation-based data-driven control scheme yields considerable control performance improvement over existing formulations.
This page was built for publication: Data-Driven Output Prediction and Control of Stochastic Systems: An Innovation-Based Approach
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6508042)
