For more details, contact the Vice Program Chair for Tutorial Sessions, Professor Joao Gomes da Silva Jr: email@example.com
Alexandre S. Bazanella, Frank Allgöwer
Data-driven control is a research topic that has received considerable attention at least since the early 1990’s, and that recently has experienced a boost. The early work was centered around the design of dynamic output feedback controllers, mostly within the classical model reference approach. Many different design methods have been developed along this line, and this approach has become a rather mature subject.
More recently, following the huge availability of data for virtually all systems, new perspectives on data-driven control have emerged. In attacking the development of data-driven methodologies for more general control policies, whether explicit (such as LQR) or implicit (such as predictive control), it was found that the tools of subspace identification are instrumental and that behavioral systems theory is a natural and powerful framework. Accordingly, an independent data-driven control theory has developed, setting itself apart from the model-based paradigm, with an important part of this new theory built upon Willems fundamental lemma. The fundamental aspects of control theory have already been approached within this framework, such as controllability and stability or dissipativity analysis. Various methodologies for control design have been developed, with predictive control playing a central role.
In this session we will start by presenting the classical data-driven control methodologies, showing the important contribution that the data-driven paradigm has already brought to control technology and practice. Then we will present various facets of the behavioral approach, from its fundamental concepts to the state-of-the-art. We will cover the framework of data informativity, which provides conditions on the data that allow us to solve a wide range of problems. For the LQR problem, it will be shown that regularization of the control objective allows to achieve robustness and provides flexibility that can improve performance significantly. Different approaches for the design of predictive controllers will be discussed, methods that guarantee closed-loop stability and robustness will be presented, and the session will be concluded with a discussion of real-world applications.
Stephane Lafortune, Christoforos N. Hadjicostis, Feng Lin, Rong Su
This tutorial session will consider the effect of deception attacks on compromised sensors and actuators at the supervisory control layer of cyber-physical control systems.
The problem will be modeled and analyzed in the framework of the theories of diagnosability and supervisory control of discrete event systems, where discrete transition models are used. Both attacks and defense against attacks will be considered. First, robust estimation and diagnosis in the presence of sensor attacks will be analyzed. Then, the problem of synthesizing covert attacks will be formulated and its solution discussed in different contexts. Finally, the problem of synthesizing supervisors that are resilient to covert attacks on sensors and actuators will be studied. The presentations will not assume detailed prior knowledge of diagnosis or supervisory control of discrete event systems.
Infectious diseases are a global health threat: the world has recently faced outbreaks of Ebola, SARS, MERS, tropical diseases, and is facing the ongoing HIV and COVID-19 pandemics. Mathematical epidemiologists, as well as researchers from the systems-and-control community, have shown the fundamental role of mathematical models in the monitoring, prediction, prevention and control of epidemics. The proposed Tutorial Session aims at providing a broad overview of the fundamental developments in the modelling and the control of infectious diseases, seen as complex phenomena that embrace multiple scales, from within host (infection, i.e. interplay between pathogens and immune system) to between hosts (contagion, i.e. spread of the disease between individuals). Three talks will focus on diverse aspects to propose a well-rounded illustration of the relevance of control-theoretic methods to model and contrast infectious diseases at all scales. The first talk will focus on modelling and control of infectious diseases in the host, including the dynamic interaction between pathogens and the immune system of the host. Control strategies will be discussed aimed at designing personalised therapies and optimal treatments to clear the infection.
The second talk will deal with epidemic dynamics over networks where the nodes represent individuals and the links represent interactions that can lead to contagion between hosts. Both static and time-varying networks will be considered, as well as the case of multi-strain and multi-virus models, and mitigation actions. The third talk will adopt a holistic viewpoint on epidemics (including opinion dynamics and socio-economic aspects along with public health issues) and will discuss their control across scales, resorting to multi-pronged contrast strategies that leverage both pharmaceutical interventions, such as drugs and vaccines, and non-pharmaceutical interventions, including hygiene, use of personal protective equipment, physical distancing. Examples based on real data and simulation results will be provided to bridge theory and real case studies. Some open problems will also be discussed to stimulate research ideas and goals for the future.
Maria Laura Delle Monache, Cecilia Pasquale
This tutorial provides an overview of new and emerging areas of traffic control and estimation. The goal is to give a broad view on how the control and estimation problems are evolving in the last few years. In particular, the session will emphasize how new disruptive technology, as for example connected and automated vehicles (CAVs) or Artificial Intelligence (AI), are impacting the way we think of traffic control and estimation. Classical and new methods will be presented to give a broad outlook on the current used techniques.