Séminaire d’automatique @ Paris-Saclay

This series of seminars aims at disseminating throughout the iCODE community the recent advances concerning the domains of automation and mathematical control theory, as well as related emerging domains.

Organizers:


14h00-15h00. Thibault Liard (LS2N, Nantes, France)

Boundary sliding mode control of system of transport equations

We study the asymptotic behavior of a system of transport equations subject to unknown boundary disturbances. Our aim is to construct a boundary feedback law, based on a sliding mode procedure, which rejects the disturbance in finite time and which globally stabilizes the equilibrium point zero. The main novelty of our approach consists in defining a sliding variable and a corresponding sliding surface on which the global exponential stability is ensured. We will extend this approach to an equation of conservation laws with simulations.

Biography. Thibault Liard earned his PhD in observation and control for some conservative systems under the supervision of Alain Haraux and Yannick Privat. From 2017 to 2019, his Postdoctoral positions were focused on the study of strongly coupled PDE/ODE systems where the PDE consists of a system of conservation laws and the ODE represents the trajectory of particles. From 2019 to 2020, he was a Postdoctoral researcher at the ERC Advanced Grant project DyCon under the supervision of Prof. Enrique Zuazua where he was working on inverse problems of conservation laws. From 2021, he is Postdoctoral researcher at LS2N with Swann Marx where he is focusing on control and stabilization of conservation laws.
His main research interests are control theory, calculus of variations, inverse problem, coupled PDE/ODE, hyperbolic systems and wave-front tracking algorithm.

Practical details. the seminar will be also broadcast live using Microsoft Teams. If you want to attend this seminar (or any of the forthcoming online Automatiques seminars), and if you do not already have access to the AutoSeminar group on Teams, simply send an email and you will be invited. Please specify which email address the invitation must be sent to (this has to be the address associated with your Teams account). You will find the link to the seminar on the « General » AutoSeminar channel on Teams, approximately 15 minutes before the beginning.

The technical side of things: you can use Teams either directly from you web browser or using the « fat client », which is available for most platforms (Windows, Linux, Mac, Android & iOS). We strongly recommend the latter option whenever possible. Please give it a try before the seminar to anticipate potential problems.


14:00-15:00 Claudio Gaz
 (Post-Doc researcher, Department of Computer, Control and Management Engineering (DIAG), Sapienza Università di Roma, Italy)

Title.  The role of modelling and parameter identification for controlling robotic and biological systems.

Abstract. Mathematical models are widely employed to describe phenomena in diverse domains, as natural or social sciences, or engineering. While model-less approaches (i.e., machine learning techniques) may neglect an explicit knowledge of the laws behind the system under study, the implementation of advanced control strategies (i.e., optimal or robust control) requires more effort with respect to model-based approaches, which, conversely, need a reliable and identifiable mathematical model. 

Having the symbolic structure of the model representing the considered system, a crucial process consists in the identification of the intrinsic model parameters for the actual observed system, in order to have algebraic or differential equations able to reliably describe the process under study. 

In robotics, a dynamic model is the relationship between joint motion (positions, velocities and accelerations) and applied joint torques. The knowledge of accurate dynamic models is of fundamental importance for many robotic applications, such as for planning minimum energy trajectories, when regulating force or imposing a desired impedance control at the contact, or when implementing strategies for the sensorless detection and isolation of unexpected collisions. By means of a dynamic observer of the unknown actuation faults (a.k.a. the residual vector), it is possible to retrieve an estimation of the external disturbances, thus unforeseen collisions. Furthermore, when a collision is sensed, possible countermeasures may be taken, as reaction maneuvers like human reflexes. Moreover, human-robot collaboration strategies in industrial settings are also achieved by means of the residual vector: for instance, the orientation of a workpiece held by the end-effector of a manipulator can be changed by simply pushing or pulling the robot structure, while preserving its position. 

Beyond robotics, parameters identification is a critical issue even in biomedical contexts: for instance, the tuning of artificial pancreas devices for insulin-resistant patients. A recently published mathematical model accurately describing human glucose homeostasis is exploited to generate virtual patients: in particular, the glycemic profile of a healthy patient, together with the identified insulin-resistant patient parameters, can be successfully used to tune an external controller infusing insulin by subcutaneous injections. 

Biography.  Claudio Gaz is a Post-Doc researcher at the Department of Computer, Control and Management Engineering (DIAG) of Sapienza Università di Roma (Italy), where he received a Master Degree in Control Engineering with the highest mark in 2011 and a Ph.D. in Automation and Operational Research in 2016. He received in 2019 the French national qualification as maître de conférences for the class 61 (Génie informatique, automatique et traitement du signal). His main interests are mathematical modeling, parameter identification and control of robotics and biological systems. In particular, he dealt with the dynamic parameters identification of well-known manipulators, such as the KUKA LWR, the Universal Robots UR10, the Kinova Jaco Arm 2 and the Franka Emika Panda. He was a visiting researcher at Airbus (Airbus Group) in Suresnes (France) and at the German Aerospace Center (DLR) in Oberpfaffenhofen (Germany). He is currently collaborating also with the Italian National Research Council (CNR-IASI) on issues concerning the control of glycemia for insulin-resistant patients. 

10:00-11:00 Jean Auriol (Chargé de recherche, CNRS, L2S, CentraleSupélec)

Title. Robust backstepping stabilization of linear hyperbolic PDEs systems. Application to a drilling problem.

Abstract. Linear hyperbolic systems naturally arise when modeling industrial processes for which the dynamics involve a transport phenomenon (related applications include electric transmission lines, traffic flow, oil well drilling…). These systems are the source of complex control and engineering problems (mostly due to the transport phenomena and the presence of destabilizing terms), which have impact in terms of environmental safety and economic feasibility. In this presentation, we develop operating methods for the control of such hyperbolic systems. More precisely, using a backstepping approach combined with a rewrite of the system as a difference equation, we design an explicit control law (and the corresponding dual observer) that guarantees the robust output feedback stabilization of a system of two hyperbolic PDEs.  The proposed control law  introduces three degrees of freedom (by means of tuning parameters) that enable a trade-off between performance and robustness, between disturbance rejection and sensitivity to noise. The proposed approach can be extended to higher dimensional systems and networks interconnected systems. Finally, we conclude this presentation by considering the problem of toolface control for directional drilling operations with the bit off-bottom. The torsional dynamics of such a system can be modeled as a non-linear hyperbolic system for which a robust backstepping-based state-observer is  designed to monitor at all times the torque and the RPM. Using these estimations, we design an algorithm that controls the toolface orientation. The different algorithms are tested against real field data.

Biography. Jean Auriol received his Master degree in civil engineering in 2015 (major: applied maths) in MINES ParisTech, part of PSL Research University and in 2018 his Ph.D. degree in control theory and applied mathematics from the same university (Centre Automatique et Systèmes). His Ph.D. thesis, entitled Robust design of backstepping controllers for systems of linear hyperbolic PDEs, has been nominated for the best thesis award given by the GDR MACS and the Section Automatique du Club EEA in France. From 2018 to 2019, he was a Posdoctoral Researcher at the Department of Petroleum Engineering, University of Calgary, AB, Canada, where he was working on the implementation of backstepping control laws for the attenuation of mechanical vibrations in drilling systems. From December 2019, he is a Junior Researcher (Chargé de Recherches) at CNRS, Université Paris-Saclay, Centrale Supelec, Laboratoire des Signaux et Systèmes (L2S), Gif-sur-Yvette, France.His research interests include robust control of hyperbolic systems, neutral systems, networks and interconnected systems.

11:00-12:00 Federico Bribiesca-Argomedo (Associate Professor, Department of Mechanical Engineering, Institut National des Sciences Appliquées de Lyon)

Title. Handling interconnections in hyperbolic-PDE/ODE systems with reduced over-actuation.

Abstract. Linear hyperbolic PDEs are a common representation for natural or artificial processes where some quantity: matter, energy, information, etc., propagates with a finite speed on a spatial, or at least space-like, domain. In particular, systems of coupled hyperbolic PDEs are a common occurrence, since balance laws rarely appear in an isolated manner, and information in a system (e.g., the effect of actuation on a process) tends to propagate in all, or at least several, spatial directions. In this talk we will focus on the use of the infinite-dimensional backstepping method to simplify coupling structures in systems of hyperbolic PDEs, allowing for constructive control designs that do not require the use of « one actuator per transport equation, » thus reducing the need for over-actuated systems. The focus will be on results showing how the method extends to cases where ODE dynamics are present on the actuated and/or unactuated boundaries of hyperbolic systems. Particular attention will be paid to the robustness of such designs, which can require an « infinite bandwidth » in their more naive forms, with respect to small delays in the control loop. A more practical control design will be presented, using adequately designed filters to restrict the bandwidth of the resulting controller while preserving the stability of the closed-loop system.

Biography.Federico Bribiesca-Argomedo received the B.Sc. degree in mechatronics engineering from the Tecnológico de Monterrey, Monterrey, Mexico, in 2009, the M.Sc. degree in control systems from Grenoble INP, Grenoble, France, in 2009, and the Ph.D. degree in control systems from GIPSA-Laboratory, Grenoble University, Grenoble. He held a post-doctoral position with the Department of Mechanical and Aerospace Engineering, University of California, San Diego, San Diego, CA, USA. He is currently an Associate Professor with the Department of Mechanical Engineering, Institut National des Sciences Appliquées de Lyon, Lyon, France, attached to Ampère Laboratory. Research interests include control of hyperbolic and parabolic partial differential equations and nonlinear control theory. Past and current applications include tokamak safety factor profiles, electrochemical models of Li-ion batteries and energy distribution networks.

SEMINAIRE DE Claudio Gaz

14:00-15:00 Claudio Gaz (Post-Doc researcher, Department of Computer, Control and Management Engineering (DIAG), Sapienza Università di Roma, Italy) Title.  The role of modelling and parameter identification…
Lire la suite