Resilience in State Estimation and System Identification

Abstract: Resilience is a particular robustness property which characterizes the sensitivity of some performance function of interest with respect to a class of disturbances (model uncertainties). For example, we say that a state estimator is resilient to a set of disturbances E if the estimation error induced by that estimator is (a) zero whenever the actual model uncertainty lies in E and (b) continuously dependent on the distance from the actual uncertainty to the set E. This talk will discuss a resilience-inducing optimization framework for secure state estimation and system identification in the scenario where E is a set of impulsive (or sparse) noise sequences. This type of noise signal can account typically for intermittent sensor failures or adversarial attacks in the context of cyber-physical systems. It can also arise artificially as a methodological device for example, in the identification, estimation, and control of switched systems. We consider both batch off-line and online recursive estimation.

Bio: L. Bako is an Associate Professor of control theory and signal processing at Ecole Centrale de Lyon, France. He is currently a visiting Associate Professor in the EECS department at the University of Michigan (within the group of Prof Ozay). He received an M.Sc. from Université de Poitiers (2005), a Ph.D. in Automatic Control and Computer Science from Université Lille 1  (2008), and a Research Accreditation (Habilitation à Diriger des Recherches) from Université de Lyon, France (2016). Before joining Ecole Centrale de Lyon, he was an Assistant Professor at IMT Nord-Europe (formally called Ecole des Mines de Douai) and a visiting researcher at the Center for Imaging Science at Johns Hopkins University. His research interests are in control theory, system identification, machine learning and optimization. He has served as an associate editor for many control conferences (IEEE CDC & ACC, IFAC-WC, ECC) and journals (International Journal of Robust and Nonlinear Control, Nonlinear Analysis – Hybrid Systems and IEEE Transactions on Automatic Control with service starting in Nov 2023).

***Event will take place in hybrid format. The location for in-person attendance will be room 1303 EECS.   Attendance will also be possible via Zoom. The Zoom link and password will be distributed to the Controls Group e-mail list-serv.

To join this list-serv, please send an (empty) email message to [email protected] with the word “subscribe” in the subject line. To cancel your subscription, send an (empty) email to the same address with the word “unsubscribe” in the subject line.  Zoom information is also available upon request to Sher Nickrand at [email protected].