CKSRI Seminar Series 2024 "Robust Data-Driven Predictive Control for Unknown Linear Time-Invariant Systems"

Thank you to everyone who joined the CKSRI Seminar!
We are honoured to have Prof. Tao Liu, Assistant Professor of Electrical and Electronic Engineering at the University of Hong Kong, as our distinguished speaker. 

Abstract:
This seminar has invited Prof. Tao Liu, Assistant Professor of Electrical and Electronic Engineering at the University of Hong Kong to discuss the Robust Data-Driven Predictive Control for Unknown Linear Time-Invariant Systems.
Model-based control relies heavily on accurate mathematical models, which can be difficult to obtain in practice. To address this issue, data-driven control, specifically for linear time-invariant (LTI) systems, has garnered significant attention from various research fields. Recently, model predictive control, an essential and effective control technique for constrained systems, has also been expanded into data-driven predictive control, which does not rely on an explicit system model. However, many of these control methods necessitate the pre-collection of adequate and informative data, meaning that the pre-collected data must satisfy the persistently exciting (PE) condition of a sufficiently high order. In practice, obtaining data that meets the PE condition from an unknown system can be challenging, especially for unstable systems. 
In this talk, we aim to develop a novel, robust, data-driven predictive control scheme for unknown LTI systems that eliminates the need for the PE condition on pre-collected data.
We construct a set containing all systems capable of generating the given pre-collected data, thereby removing the requirement for the PE condition of a sufficiently high order on pre-collected data. At each time step, we derive an upper bound for a given objective function for all systems in the set and design a feedback controller to minimize this bound. The optimal control gain at each time step is determined by solving a series of linear matrix inequalities. We demonstrate that if the synthesis problem is feasible at the initial time step, it remains feasible for all future time steps. Unlike existing data-driven predictive control schemes based on behavioral system theory, our approach imposes less stringent conditions on pre-collected data, enabling easier implementation.

Short Bio: 
Tao Liu (M’13) received his B.E. degree from Northeastern University, China, in 2003 and his Ph.D. degree from the Australian National University (ANU), Australia, in 2011. From 2012 to 2015, he worked as a Post-doctoral Fellow at ANU, University of Groningen, and University of Hong Kong (HKU). He became a Research Assistant Professor at HKU in 2015 and is now an Assistant Professor. His research interests include power system analysis and control, complex dynamical networks, data-driven control, distributed control/optimization, and event-triggered control.