The colloquium will take place in Room V7.01, Pfaffenwaldring 7, at the campus of the University of Stuttgart on June 7, 2022.
Welcome
Abstract: Computing power doubles every two years, and is called Moore’s Law. This exponential rate of change enables accelerating developments in sensor technology, AI computing and in robotics and automotive. Machines to make products in modern factories will be smart and self-learning. Cars will become like an iPad on wheels. Our group is world champion in soccer playing robots. The question is when will they be better than humans? What we learn with playing robot soccer is also applicable to service robotics for care at home, and to autonomous guided vehicles for agriculture, logistics and industrial applications. We learn our robots to navigate, but when will robots start to learn to us. Are humans in the end necessary? And how does the future of schools and universities look like?
Bio: Maarten Steinbuch (born 1960 in Zeist, NL) is a high-tech systems scientist, entrepreneur and communicator. He holds the chair of Systems & Control at Eindhoven University of Technology (TU/e), where he is Distinguished University Professor. He is also Scientific Director of Eindhoven Engine. The research of his group spans from automotive engineering (with a focus on connected cars and clean vehicles) to mechatronics, motion control, and fusion plasma control. He is most known for his work in the field of advanced motion control and mechatronics, as well as in robotics for high precision surgery. Steinbuch is a prolific blogger and a key opinion leader on the influence of new technologies on society.
Abstract: The fundamental lemma proposed by Jan Willems and co-authors in 2005 is pivotal for many recent research efforts on data-driven control and behavioral systems theory. It states that, under suitable assumptions, any input-output trajectory of a linear time-invariant (LTI) system can be described as a linear combination of previously recorded trajectories. Recently, there have been several extensions of the fundamental lemma, e.g., to linear-parameter varying systems and to nonlinear systems. Moreover, predictive control based on such non-parametric system descriptions is receiving substantial research interest. In this seminar, Timm will discuss progress of data-driven output feedback MPC of stochastic LTI systems. Leveraging polynomial chaos expansions (PCE) of random variables, the origins of which date back to Norbert Wiener, Timm answers the question of how to formulate the fundamental lemma for stochastic systems. The seminar will illustrate that the knowledge or estimation of past noise realizations allows the construction of Hankel matrices which in turn enable propagation of non-Gaussian and Gaussian uncertainties with non-parametric system descriptions. Put differently, the seminar will show how to achieve uncertainty propagation for stochastic LTI systems without explicit model knowledge. In the final part of the talk, Timm will turn towards data-driven stochastic MPC to demonstrate that the proposed non-parametric system description can be used in stochastic optimal and predictive control. Timm's findings will be illustrated by examples from different applications.
Bio: Timm Faulwasser has studied Engineering Cybernetics at the University of Stuttgart, with majors in systems and control and science theory and philosophy. From 2008 until 2012 he was a member of the International Max Planck Research School for Analysis, Design and Optimization in Chemical and Biochemical Process Engineering Magdeburg. In 2012 he obtained his PhD from the Department of Electrical Engineering and Information Engineering, Otto-von-Guericke-University Magdeburg, Germany. From 2013 to 2016 he was with the Laboratoire d’Automatique, École Polytechnique Fédérale de Lausanne (EPFL), Switzerland, while 2015-2019 he was leading the Optimization and Control Group at the Institute for Automation and Applied Informatics at Karlsruhe Institute of Technology (KIT), where he successfully completed his habilitation in the Department of Informatics in 2020. In November 2019 he joined the Department of Electrical Engineering and Information Technology at TU Dortmund University, Germany. Currently, he serves as associate editor for the IEEE Transactions on Automatic Control, the IEEE Control System Letters, as well as Mathematics of Control Systems and Signals. His main research interests are optimization-based and predictive control of nonlinear systems and networks with applications in energy, process systems engineering, mechatronics, and beyond.
Coffee Break
Small snacks are provided
Abstract: The study of large scale and complex interconnected systems is of great importance in today's networked world with applications ranging from distributed power generation to deep space exploration. A great challenge for these systems is to understand the interplay between the dynamic properties of the individual systems comprising the networks, the underlying information exchange network, and the interaction protocols governing the collective behavior. In this talk we will explore necessary and sufficient conditions for a network of passive dynamical systems to reach an output agreement, i.e., the trajectories of each system will synchronize. This leads to a refinement of classical passivity theory that we term maximal equilibrium passivity. We then show that the steady-state behavior of these systems are in fact solutions to a family of classic network optimization problems, and as a result we draw connections between notions of duality in static optimization to cooperative control. This network optimization perspective also leads to synthesis methods for controllers to guarantee the desired behavior of the network and provides new insights to classical problems such as feedback passivation.
Bio: Daniel Zelazo is an associate professor of aerospace engineering and director of the Philadelphia Flight Control Laboratory at the Technion-Israel Institute of Technology, Haifa. He received the B.Sc. and M.Eng. degrees in electrical engineering and computer science from the Massachusetts Institute of Technology, Cambridge, in 1999 and 2001, respectively. In 2009, he completed his Ph.D. degree at the University of Washington, Seattle, in aeronautics and astronautics. From 2010 to 2012, he was a postdoctoral research associate and lecturer at the Institute for Systems Theory and Automatic Control, University of Stuttgart, Germany. He is currently an associate editor of IEEE Control Systems Letters and subject editor for the International Journal of Robust and Nonlinear Control. His research interests include topics related to multi-agent systems, graph theory, and control systems.
Abstract: Motivated by applications in the energy domain, we shall discuss the key role of data-driven and distributed optimization in addressing the challenges posed by the presence of uncertainty and the large-scale structure of the involved systems. In particular, we shall present recent results on stochastic control with probabilistic constraints and decision making for multi-agent systems characterized by both discrete and continuous decision variables.
Bio: Maria Prandini received the laurea degree in Electrical Engineering, summa cum laude, from the Politecnico di Milano in 1994, and the Ph.D. degree in Information Technology from the University of Brescia in 1998. She was a postdoctoral researcher at the University of California at Berkeley from 1998 to 2000. She also held visiting positions at Delft University of Technology (1998), Cambridge University (2000), UC Berkeley (2005), and ETH Zurich (2006). In 2002, she joined the Department of Electronics, Information and Bioengineering at the Politecnico di Milano, where she is Full Professor since 2018. She is currently Chair of the Automation and Control Engineering Program. Her research interests include stochastic hybrid systems, randomized algorithms, distributed and data-driven optimization, multi-agent systems, and the application of control theory to transportation and energy systems. She was elected Fellow of the IEEE in 2020 and IEEE Control Systems Society Distinguished Member in 2018. In 2017, she was August-Wilhelm Scheer Visiting Professor and Honorary fellow of the TUM Institute for Advanced Studied. Since January 2022, she is Visiting Professor in Engineering at the University of Oxford for a period of 3 years. She is and has been active in the IEEE Control Systems Society (CSS), the International Federation of Automatic Control (IFAC), and the Association for Computing Machinery (ACM), contributing to their activities in different roles. She is currently IFAC Vice-President Conferences for the triennium 2020-23. Previously, she was IEEE CSS Vice-President for Conference Activities in 2016 and 2017, elected member of the IEEE CSS Board of Governors (2015-2017), appointed member of the IFAC Policy Committee in (2017-2020), and a member of SIGBED Board of Directors (2019-21). She currently serves on the editorial boards of Automatica and IEEE Transactions on Control of Network Systems as an associate editor. She is general chair of the IEEE Conference on Decision and Control 2024 and of the Mediterranean Conference on Control and Automation 2022.·
Lunch and poster session
University Canteen / V9.0.144 (next to V7.01)
Abstract: Motivated by advances in Machine Learning, the past few years have seen renewed interest in designing controllers directly from data. In this talk we will discuss both classical approaches and new emerging ones. We will start by examining some new results concerning the sample and computational complexity of learning dynamical systems, as well as the implications of these results for designing controllers using a two step procedure: Identification followed by controller design. We will also argue that when the goal is to design controllers, the "loss function" used in learning should reflect closed loop, rather than open-loop performance, a task that can be accomplished by using a gap-metric motivated approach. This is particularly important when identifying open-loop unstable plants, since typically in this case the open loop distance is unbounded. Next, we will cover some new methods that, motivated by the classical Willems' Lemma for LTI systems, avoid the plant identification step by directly identifying a controller from the data, and briefly indicate how to extend some of these methods to the switched and non-linear cases. The talk concludes by comparing the "data driven control" framework against methods that also seek to learn controllers from data using "control agnostic" Machine Learning (ML) based methods. With some simple examples we will illustrate the challenges faced by "control agnostic" ML and argue that these methods are unlikely to succeed in moderately difficult cases.
Bio: Mario Sznaier is currently the Dennis Picard Chaired Professor at the Electrical and Computer Engineering Department, Northeastern University, Boston. Prior to joining Northeastern University, Dr. Sznaier was a Professor of Electrical Engineering at the Pennsylvania State University and also held visiting positions at the California Institute of Technology. His research interest include robust identification and control of hybrid systems, robust optimization, and dynamical vision. Dr. Sznaier is currently serving as chair of the IFAC Technical Committee on Robust Control and Editor In Chief of the section "AI and Machine Learning Control" of the journal Frontiers in Control Engineering. Past recent service include Program Chair of the 2017 IEEE Conf. on Decision and Control, General Chair of the 2016 IEEE Multi Systems Conference and Associate Editor for Automatica (2005-2021). He is a distinguished member of the IEEE Control Systems Society and a Fellow of the IEEE for his contributions to robust control, identification and dynamic vision. A list of publications and current research projects can be found at http://robustsystems.coe.neu.edu.
Abstract: Natural gas processing facilities form a part of the energy infrastructure associated with agile power production. At this stage, they are an important part of carbon emissions reduction. Energy efficiency and adaptability of the facility dictate that high-performance control be applied and that this be reconfigurable with modest design effort. Traditionally, mechanical thermos-fluid systems have been modeled by bond graph methods involving differential-algebraic equations, where the algebraic aspects are connected to conservation of mass constraints. Frank Allgöwer has studied Model Predictive Control applied directly to such systems. Here we present methods of modeling these systems subsuming the conservation of mass into the linear state equations, thereby removing the algebraic parts and admitting mutliinput-mutlioutput linear control design methods directly. These models of compressible fluid flow capture the resonant wave modes along with the bulk fluid flows of interest for the facility. We show how the models may be interconnected in a simple and familiar fashion subsuming Mason’s Gain Rule. The complex system is then combined with antialiasing filters, model reduction and control design to illustrate the applicability of the approach and the presence and role of mass conservation.
Bio: Bob Bitmead is Distinguished Professor in Mechanical & Aerospace Engineering at the University of California, San Diego. He holds degrees in Applied Mathematics and Electrical Engineering from Sydney University and Newcastle University, both in Australia. He has held faculty positions at the Australian National University and James Cook University of North Queensland. He is a control theorist with a long experience in control applications in many industrial sectors. His theoretical work is strongly informed and guided by these applications. He was the recipient of the 2014 ASME Rufus Oldenburger Medal and of the 2015 IEEE Controls Systems Transition to Practice Award. Bob was President of the IEEE Control Systems Society for 2019. He was a member of the IFAC Council from 1996 to 2002 and is Editor-in-Chief of the IFAC Journal of Systems & Control. He is Fellow of IEEE, IFAC and the Australian Academy of Technological Sciences and Engineering. Bob brews his own beer and is an accredited and active Australian Rules Football umpire training with the San Diego Lions Australian Football Club.
Coffee Break
Small snacks are provided
Abstract: In this talk we outline the challenges that exist for parameter estimation for non-linear systems where un-measured inputs (disturbances) significantly affect the behavior of the system. Neglecting these may give rise to gross systematic errors (bias) in the estimated parameters, while, on the other hand, accounting for them gives rise to computational challenges. Classical estimation techniques such as maximum likelihood, Bayesian and optimal prediction error methods all require non-linear estimation and while this can be done using particle filtering and smoothing techniques we show that by relaxing the family of estimators used, the computational load may be reduced without sacrificing consistency of the parameter estimates. Based on this insight we construct a stochastic gradient based method for estimation of parameters in non-linear differential-algebraic-equation models subject to stochastic disturbances. The methods uses pure simulations of the model and by taking this notion one step further we also illustrate how one can split the construction of the parameter estimator and the inference in time using deep learning techniques so that the estimator can be constructed a priori and inference can be performed instantaneously.
Bio: Håkan Hjalmarsson was born in 1962. He received the M.S. degree in Electrical Engineering in 1988, and the Licentiate degree and the Ph.D. degree in Automatic Control in 1990 and 1993, respectively, all from Linköping University, Sweden. He has held visiting research positions at California Institute of Technology, Louvain University and at the University of Newcastle, Australia. He has served as an Associate Editor for Automatica (1996-2001), and IEEE Transactions on Automatic Control (2005-2007) and been Guest Editor for European Journal of Control and Control Engineering Practice. He is Professor at the Division of Decision and Control Systems, School of Electrical Engineering and Computer Science, KTH, Stockholm, Sweden and also affiliated with the Competence Centre for Advanced BioProduction by Continuous Processing, AdBIOPRO. He is an IEEE Fellow and past Chair of the IFAC Coordinating Committee CC1 Systems and Signals. In 2001 he received the KTH award for outstanding contribution to undergraduate education. He was General Chair for the IFAC Symposium on System Identification in 2018. His research interests include system identification, learning of dynamical systems for control, process modeling control and also estimation in communication networks.
Abstract: Stochastic feedback systems give rise to a variety of notions of stability: median, mean, and variance stability conditions differ. These conditions can be stated explicitly for scalar discrete-time systems with (almost) arbitrary distributions of the stochastic feedback gain. The state variable in such systems evolves towards a heavy-tailed distribution and exhibits some non-intuitive characteristics. For example, one can use stochastic feedback to stabilise unstable systems where one does not even know the sign of the unstable pole or the sign of the system gain. A more dramatic example is an investment scheme which simultaneously yields unbounded expected profit and almost certain bankruptcy to every investor.
Bio: Roy Smith is a Professor in the Automatic Control Laboratory at the Swiss Federal Institute of Technology (ETH, Zurich) in Switzerland. From 1990 to 2010 he was on the faculty of the Electrical Engineering Dept. at the University of California, Santa Barbara. He received his undergraduate education at Canterbury University in New Zealand (1980) and a Ph.D. from the California Institute of Technology (1990). Roy Smith's research interests include: the identification and control of uncertain systems, and distributed estimation, communication and control systems. His application experience includes: process control, automotive engines, flexible space structures, aeromanoeuvring Mars entry vehicles, formation flying of spacecraft, magnetically levitated bearings, high energy accelerator control, airborne wind energy and energy control for buildings. He has been a long time consultant to the NASA Jet Propulsion Laboratory on guidance, navigation and control aspects of interplanetary and deep space science spacecraft. He is a Fellow of the IEEE & IFAC, an Associate Fellow of the AIAA, and a member of SIAM and NZAC.
Concluding Remarks