The European Control Conference offers pre-conference workshops addressing current and future topics in control systems from experts in academia, research institutes, and industry. Pre-conference workshops cover material or use presentation formats that are not found within the main conference to increase the interest for the event, enhance interaction and discussion amongst participants, and make useful connections to fields outside of control.
Main conference registration is a prerequisite for registering at ECC workshops.
Important: please note that you can register for only one workshop, since all workshops take place simultaneously. Recordings of workshops will not normally be made.
Information about the costs of attending a pre-conference workshop and how to register is available on the conference registration page.
The list of confirmed workshops is reported below. For further information contact the conference Workshops Chair, prof. Maria Elena Valcher.
ECC23 is hosting the following four confirmed workshops, on Tuesday, June 13th 2023 (full day):
Tuesday, June 13th 2023 @ 9:00 -17:00 (TBC)
Symbolic control and formal methods have been a promising direction in Systems and Control for several decades. Indeed, they allow in principle to provide formal guarantees on the behaviour of dynamical systems, even very complicated ones, by analysing models of such systems. However, such complicated systems are often naturally studied in a ‘model-free’, ‘data-driven’ fashion. Think for instance of a self-driving car, with many heterogeneous and complex components: this system is hard to model rigorously, but at the same time generates a huge amount of data via its many embedded sensors. Other examples include complex robots, smart grids, and many other modern systems. If data are certainly an opportunity for symbolic control of complex systems, they provide technical challenges that are not well understood nowadays: how can one reconcile the intrinsically random nature of data-driven approaches with certified guarantees? How can we certify asymptotic properties from finite-time observations? In recent years, these questions have found partial answers and deep relations with topics such as Interval Markov Chains, the Scenario Approach, Ergodic Dynamics or Monotone Systems. The event will consist of seven 40-minutes presentations from various leading researchers active in the topic. The goal of this full-day workshop is to introduce this recently emerged challenge to the community. As such, the talks will be aimed at a general public in control, with an emphasis on pedagogy to provide an entrance point to this exciting field in our community. The day will end with a panel discussion on the future directions of the topic.
Tuesday, June 13th 2023 @ 9:00 – 17:30
C-0-Auditorium (streaming at P-1-Aula Magna)
Lorenzo Jr. Sabug
When seeking the global minimum of a general non-convex problem, any sensible approach must concurrently learn, to gather knowledge about the cost and constraints, and optimize, to find better and better values of the decision variables inside the feasible region. When the cost and constraint functions are not available analytically, these two tasks must be carried out using only data samples – hence the name “black-box optimization”. Black-box optimization is highly relevant to many science and engineering fields. Examples are system design optimization using detailed simulations, system/control tuning via experiments, control adaptation, parameter identification for nonlinear models, training of neural networks and other nonlinear approximators, experiment design. This workshop will provide an overview of the main solution approaches in the literature, then introduce a new methodology and the related algorithm, named SMGO – Set Membership Global Optimization. Thanks to an interactive, hands-on session, the participants will be able to immediately test SMGO and compare it with other approaches. A final session and discussion will highlight challenges and promising research directions.
Tuesday, June 13th 2023 @ 9:30 -17:30
Small scale robotic platforms are fast becoming ubiquitous. We expect that the next decade will show an exponential growth in various applications. Among the many challenges, we may mention the unstructured environment and human interactions and the relatively low cost (and thus low performance) sensing modules available. All these limitations impose a rethinking of standard motion planning algorithms to emphasize redundancy, reliability and quick validation for unexpected mission changes. Consequently, both beginners and experts have to become accustomed with new results and tools at an increasingly hectic pace. We propose in this workshop to: i) cover new results from guidance, control and estimation for single and multi-agent systems; and ii) to validate them via simulation and experiments in interactive sessions. Our approach, proposed for this workshop, is to provide a complete (but obviously not exhaustive) implementation stack for the control of a set of small scale indoor robotic platforms (ground-based or aerial):
• provide a model-based guidance mechanism which, via flatness descriptions and subsequent spline parametrizations, gives a feasible path;
• propose various control strategies (linear, feedback linearizable, receding horizon control) to track the path;
• discuss and compare centralized versus distributed control implementations;
• present and solve estimation and formation control issues (for realistic mission requirements).
We consider the usual sources of difficulty: control saturation and path feasibility; obstacle and collision avoidance conditions; task assignment and formation control. We seek Master/PhD candidates and early-career researchers who wish to understand all the necessary steps (guidance, control and navigation) required in handling robotic platforms in small spaces (where internal dynamics are especially relevant and where GPS-like positioning systems are not always available). The tools that we provide may be of use not only for researchers whose main activity is in robotics but also for those, eg. from the control community, who want to test their theoretical results on systems inspired from real applications. The workshop requires a basic knowledge in the principles of constrained optimization, linear and nonlinear control (state-based), receding horizon framework. Familiarity with Matlab/Python as scripting languages, Gazebo (or similar) as simulation platform, and/or ROS as middleware for robotics, is welcomed. A strong point of our workshop proposal is the availability of a hardware platform which may be used by the participants for experimental validation. The lab is 500m from the conference venue, in the UPB campus. In fact, depending on the participant number, the ancillary equipment may be moved to the conference premises with ease.
Tuesday, June 13th 2023 @ 9:30 -17:00
Ravi N. Banavar
Autonomous multi-agent dynamical systems have become very important in the analysis of largescale interconnected systems appearing in biology, social science, flow analysis, epidemiology etc… as well as in the synthesis of large-scale systems for distributed sensing, estimation, computing and control in applications like robotics, communication and optimization. For agents that evolve on Euclidean spaces, multi-agent systems have been studied extensively using the tools of linear algebra and graph theory. Distributed linear multi-agent systems have been used in many applications. Another class of multi-agent systems that are equally important are oscillator networks. Here, each agent evolves on a circle and hence described by an angular coordinate. They are important in applications like neuronal networks, clock synchronization, electrical power networks, synchronization in animals, firefly flashing rhythms, propagation of charge density waves in metals and semi-conductors, chemical oscillators etc. This class of systems is called the Kuramoto oscillator. It turns out that there is a geometric framework to capture both the classes of systems when both these classes of dynamical systems are considered abstractly as evolving on Lie Groups (Rn , +) and (S1 ,+2π ) respectively. This not only unifies the analyses of both classes of systems but also is also extendable to multi-agent dynamical systems on any Lie group admitting a bi-invariant metric. This is the main theme of the workshop. Such an extension finds important physical applications as well. Since the configuration space of a rigid body is a Lie group SO(3) that admits a bi-invariant metric, these classes of systems will find applications to networks of mechanical systems as all mechanical systems are interconnections of rigid bodies. Especially, they are useful for attitude consensus or synchronization algorithms for a network of rigid-bodies. The proponents have come up with this framework recently and believe that this has lot of scope for further research on multi-agent systems in Lie groups.