Tutorial Sessions

Tutorial sessions address state-of-the-art control theory and its industrial applications. While session formats vary, tutorial sessions often start with a longer 40- or 60-minute talk on the underlying theory or application area. After the lead presentation, there are usually several 20-minute talks highlighting particular aspects or applications of the topic area in further detail.

We are pleased to offer 7 tutorial sessions this year.

WeA25: Formal Methods in Transportation Flow Networks
Presenters: Sam Coogan (Univ. of California Los Angeles), Calin Belta (Boston Univ.), Murat Arcak (Univ. of California Berkeley), Negar Mehr (Univ. of California Berkeley), Roberto Horowitz (Univ. of California Berkeley), Sadra Sadraddini (Boston Univ.), Eric Kim (Univ. of California Berkeley), Sanjit Seshia (Univ. of California Berkeley).
Time: Wednesday, July 6, 9:30am - 11:30am
Location: Arlington

The ubiquity of sensing and communication networks provides new opportunities for correct-by-design traffic management to increase efficiency, mobility, and safety. Such advances are critical, as increasingly populous urban areas have led to a growing need for efficient use of existing transportation infrastructure. Indeed, inefficient traffic management is pervasive; for example, ninety percent of traffic signals in the United States are pre-timed, and only sixty percent of these are retimed at intervals less than five years. This inefficiency leads to congestion, the costs of which have increased five-fold in the past three decades to $120 billion annually and include 5.5 billion hours of additional travel time and 2.9 billion gallons of wasted fuel. Next generation transportation systems must mitigate this inefficiency through advances in connected infrastructure, connected vehicles, and automation. In addition, these advances must coexist with legacy technology into the foreseeable future. This complexity makes the goal of improved mobility and safety ever more daunting. Addressing these challenges requires scalable and automated verification and synthesis techniques for transportation systems. Such approaches should leverage recent advances in formal verification and synthesis of control systems to provide automated tools that guarantee safety and improve mobility. To ensure these approaches are scalable, reliable, and adaptable, future research must identify and exploit intrinsic structural properties found in transportation systems.

WeB25: Recent Advances on Computational Methods for Power Flow Equations
Presenters: Dhagash Mehta (Univ. of Notre Dame), Daniel Molzahn (Argonne National Lab), Konstantin Turitsyn (MIT), Emiliano Dall'Anese (National Renewable Energy Lab), Steven Low (California Institute of Tech.)
Time: Wednesday, July 6, 1:30pm - 3:30pm
Location: Arlington

The power flow equations are a system of polynomials that are at the heart of many electric power system computations. These equations model the steady-state relationship between the power injections and voltage phasors in an electric power system. The optimization and control needs envisioned for the design and operation of future power systems requires substantial advances in power flow analyses. Recently, novel methods with solid mathematical bases have been developed and employed to efficiently find multiple solutions of the power flow equations as well as to solve related optimization problems. Most of these methods arise from algebraic geometry, a mature area in mathematics that studies the relationship between algebraic equations and the geometry of their solution space. The methods we will focus in this tutorial session are convex relaxations using semidefinite programming, Grobner basis techniques, the polynomial homotopy continuation method, and monotone operator theory.

WeC25: Foundations of Infrastructure Cyber-Physical Systems
Presenters: Anuradha Annaswamy (MIT), Aranya Chakrabortty (North Carolina State University), Alefiya Hussein (University of Southern California), Milos Cvetkovic (MIT), Tariq Samad (Honeywell), Jakob Stoustrup (Pacific Northwest National Lab)
Time: Wednesday, July 6, 4pm - 6pm
Location: Arlington

Infrastructures have been around as long as urban centers, supporting a society's needs for its planning, operation, and safety. As we move deeper into the 21st century, these infrastructures are becoming smart; they monitor themselves, communicate, and most importantly self-govern, which we denote as Infrastructure Cyber-Physical Systems (CPS). With foundations from CPS, the time is appropriate to look at the design of smart infrastructures, which is the focus of this tutorial. After introducing the overall topic, the tutorial will focus on four main pillars of Infrastructure CPS: (i) Human Empowerment, (ii) Transactive Control, (iii) Co-design, and (iv) Resilience. The discussion will be followed by two examples, one on the interdependence between natural gas and electricity, and the other on the nexus between power and communication infrastructures, both of which have been investigated extensively of late, and are emerging to be apt illustrations of Infrastructure CPS.

ThA12: Control and Systems Challenges on a Molecular Scale
Presenters: Martha Grover (Georgia Tech), Rolf Findeisen (OVG Univ. Magdeburg), Christian Wagner (Juelich Research Center), Michael Maiworm (OVG Univ), Ruslan Temirov (Juelich Research Center), Stefan Tautz (Juelicj Research Center), Murti Salapaka (Univ. of Minnesota), Srinivas Salapaka (Univ. of Illinois Urbana Champaign), Richard Braatz (MIT), Reza Moheimani (Univ. of Texas Dallas), Joel Paulson (MIT), Venkatasailanathan Ramadesigan (IIT Bombay), Venkat Subramanian (Univ. of Washington), Daniel Griffin (Georgia Tech), Xun Tang (Georgia Tech), Anthony Fowler (Univ. of Newcastle), Mohammad Maroufi (Univ. of Newcastle), Michael Ruppert (Univ. of Newcastle)
Time: Thursday, July 7, 9.20am - 11.20am
Location: Provincetown

Feedback control has typically been applied to macroscopic systems, but in principle it should be possible to apply the same concepts to micro- and nano-scale systems.  Practical challenges include the difficulty of sensing and the limited amount of actuation that is available.  In addition, the system dynamics are typically nonlinear and stochastic, and accurate models are not always available.  Despite these challenges, technology for sensing and modeling continue to improve, and it is becoming more practical to apply feedback control at small scales.  The purpose of this tutorial session is to highlight some of the ongoing research in control at molecular scales, and to make the participants aware of possible future directions for control.

ThA21: A Tutorial, Application, and Recent Developments in Retrospective Cost Adaptive Control
Presenters: Dennis S. Bernstein (Univ. of Michigan, Ann Arbor), Jesse B. Hoagg (Univ. of Kentucky), Yousaf Rahman (Univ. of Michigan), Antai Xie (Univ. of Michigan), Frantisek Sobolic (Univ. of Michigan), Ankit Goel (Univ. of Michigan)
Time: Thursday, July 7, 9.20am - 11.20am
Location: Exeter

The proposed session is devoted to an adaptive control technique that has several unique features in relation to the many adaptive control methods that are currently under development. Retrospective cost adaptive control (RCAC) is first and foremost a discrete-time control methodology. This choice is motivated by the realization that controllers are implemented by digital computers with sampled data, but is necessitated by the fact that RCAC is based on a technique that appears to have no continuous-time counterpart. In particular, RCAC is based on the principle that a control law can be improved by re-optimizing its gains based on actual past performance. This re-optimization operates on a past window of data to determine a control law that would have provided better performance if it had been used, which of course is not possible. Nevertheless, the re-optimized control law is used at the next step, and this procedure represents the mechanism of adaptation. In this way, RCAC learns from actual performance data. The proposed session will feature three talks on RCAC, each of which is supported by a submitted paper. These talks are designed to serve several purposes for a mixed audience, including attendees interested in theory, attendees interested in applications, and attendees interested in learning about the features and challenges of adaptive control.

ThC06: Graphical Models Methods in Modeling, Analysis, Identification and Control of Networks of Dynamic Systems
Presenters: Donatello Materassi (Univ. of Tennessee), Murti Salapaka (Univ. of Minnesota)
Time: Thursday, July 7, 4pm - 6pm
Location: Salon F

There is an extensive scientific literature about the description of joint probability distributions via graphs, where each node represents a random variable and the edges describe a form of coupling among the variables. These graphical models do not necessarily represent forms of input/output relation among the variables involved. Instead, they typically represent convenient factorizations of their joint probability distribution that need to be acyclic in order to be meaningful. In the area of dynamic systems, and especially control theory, it is instead common to and network models involving stochastic processes that influence each other according to a directed graph where feedback loops may be present, as well. In this case the graph connections do indeed represent an input/output relation. Thus, the network structures underlying these two classes of models differ at a fundamental semantic level. Furthermore, it is not straightforward to relate networks of dynamic systems to graphical models of random variables. Indeed, introducing a notion of factorization associated with the interconnection graph present technical difficulties due not only to the potential presence of loops, but also because stochastic processes involve an infinite number of random variables. Despite these differences, methodologies can be borrowed from probabilistic graphical models and used for the analysis, identification and control design in the domain of networks of dynamic systems. Some of these methodologies can be imported with no significant modifications, while others need to be substantially revisited. The main goal of this session is to bridge the conceptual and methodological gap between graphical models of random variables and networks of dynamic systems creating a single unified language and theoretical framework for these two different classes of models. This is achieved by drawing parallels between similar approaches and highlighting the main differences.

FrA09: On NASDAQ Order Book Dynamics: New Problems for the Control Field
Presenters: Ross Barmish (Univ. of Wisconsin Madison), Scott Condie (Brigham Young Univ.), Sean Warnick (Brigham Young Univ.), James Primbs (California State Univ. Fullerton), Donatello Materassi (Univ. of Tennessee)
Time: Friday, July 8, 10am - 12pm
Location: Salon I

The NASDAQ is an electronic stock market, which is run entirely by high-speed computers. With this as context, the main objectives of this tutorial are as follows: (1) To demonstrate that NASDAQ order book dynamics provide a rich source of many new research problems of a control-theoretic nature; (2) to provide context and motivation driving new order book research in the control area by presenting a review of the critical issues and results in the finance literature; (3) to provide a detailed explanation of order book “mechanics” and a review of existing literature providing the basis for state space and simulation models; (4) to describe some specific new examples of Order Book Control Problems that are amenable to solution using the tools and expertise of our community. This includes control problems involving regulation in volatile markets, stock-price manipulation and robustness issues arising from dynamic model uncertainty; (5) to provide broader context for the research based on the role of the order book in a larger financial network, the tutorial includes a review of some existing results in the area of Network Identification which are relevant to areas being surveyed.