Speaker: Romain Postoyan (CNRS, CRAN, Université de Lorraine – France) joint work with Mathieu Granzotto and Jamal Daafouz (CNRS, CRAN,
Université de Lorraine – France), Lucian Busoniu (TU Cluj – Romania) and Dragan Nesic (University of Melbourne – Australia)
Abstract: Value iteration (VI) is a ubiquitous algorithm for optimal control, planning, and reinforcement learning schemes. Under the right
assumptions, VI is a vital tool to generate inputs with optimal properties for general nonlinear systems and cost functions. However,
several important questions remain largely unexplored in the literature in particular: what can we say about the stability of systems whose
inputs are obtained by VI? and what about their robustness properties?
These properties are fundamental in numerous safety-critical control applications, it is thus of primary importance to make sure that the use
of VI to construct the control inputs is done in such a way that robust stability of the closed-loop system is ensured. In this talk, we will
provide answer to these questions for the case where the plant dynamics is known, deterministic, nonlinear and in discrete-time. We will then
see how stability can be exploited to significantly improve existing near-optimality bounds of the literature, i.e., the mismatch between the
value function given by VI at each step and the targeted optimal value function. We will also provide some first elements on when to rigorously
stop the algorithm and how to implement it.
Bio: Romain Postoyan received the “Ingénieur” degree in Electrical and Control Engineering from ENSEEIHT (France) in 2005. He obtained the
M.Sc. by Research in Control Theory & Application from Coventry University (United Kingdom) in 2006 and the Ph.D. in Control Theory from
Université Paris-Sud (France) in 2009. In 2010, he was a research assistant at the University of Melbourne (Australia). Since 2011, he is
a CNRS researcher at the “Centre de Recherche en Automatique de Nancy (CRAN)” (France). He received the `Habilitation à Diriger des
Recherches (HDR)” in 2019 from Université de Lorraine (Nancy, France).
He serves/served as an associate editor for the journals: Automatica, IEEE Control Systems Letters and IMA Journal of Mathematical Control and
Information. His domains of expertise include: hybrid systems, networked control systems, event-triggered control, dynamic programming, nonlinear
Date(s) - 25 Nov 2021
14 h 00 min - 16 h 00 min
CatégoriesFiled under: DYCOP, Séminaires