NCCR Automation

Interested in a principled framework to apply predictive control (MPC) to nonlinear stochastic systems? Then check out our recent paper "Predictive control for nonlinear stochastic systems: Closed-loop guarantees with unbounded noise" on arXiv: https://lnkd.in/g-686jhr with Melanie Zeilinger. We provide a general framework that incorporates probabilistic reachability, predictive control, and stochastic contraction metrics. Closed-loop properties include satisfaction of probabilistic state constraints and expected performance bounds. In addition, online computational complexity is comparable to a certainty equivalent (nominal) formulation. NCCR Automation

Are you interested in data-driven control? Then, you should definitely check our latest research: “From Data to Predictive Control: A Framework for Stochastic Linear Systems with Output Measurements” with Andrea Carron, Melanie Zeilinger, Johannes Köhler We've developed a novel framework, Data to Predictive Control (D2PC), that bridges the gap between data-driven techniques and predictive control for discrete-time stochastic linear systems. D2PC offers: A parameter identification method based on Expectation-Maximization that enables integration of pre-defined constraints. An asymptotically correct uncertainty quantification approach, allowing us to establish probabilistic sets over parameters. A robust dynamic output-feedback controller synthesis strategy. A predictive control scheme ensuring recursive feasibility and chance constraint satisfaction. D2PC marks a significant step forward in integrating data into robust and predictive control schemes. We demonstrate its effectiveness on a 10-dimensional spring-mass-damper system. Check out our paper for the full details: https://lnkd.in/dBQwu9mf Additionally, we offer an extensive code framework implementing our pipeline: https://lnkd.in/d5pdADE6 NCCR Automation ETH Foundation #ControlSystems #ControlTheory #ModelPredictiveControl #MPC

It's not just hard work at NCCR Automation! Our Research Explanation and Application Laboratory Master's students (plus some friends) enjoyed a games night last week. Good luck to everyone working on their Master's projects at the moment - don't forget to take some time out from work. Back row (L->R)Andrei Kharitenko, Cara Koepele, Anne-Laure Matthijs, Jan Brändle, Jonathan Hilberg, Diego Parravicini Front row (L->R) Junan Lin, Philippe Brigger, Kai Zhang

#Daline, a new open-source toolbox developed by Mengshuo Jia (NCCR Automation, ETH Zürich, ETH - Power Systems Laboratory (PSL)) and his collaborators, provides over 50 data-driven linearization algorithms, making it easy for researchers and engineers to choose the most appropriate method for their tasks while enjoying the state-of-the-art performance of the latest methods with minimal coding.   Daline isn't just for theoretical research; it has real-world applications too. For example, in collaboration with Swissgrid AG, one data-driven method supported by Daline has been used and adapted to support remedial actions for a European power grid, including over 11,000 connection points and 18,000 power lines, in order to improve grid reliability when incorporating renewables. Additionally, using Daline as a simulation platform, Mengshuo Jia and his collaborators have shown that large language models can be designed to perform power system simulations. #LLM #Energy #EnergyTransition https://lnkd.in/dK9Su4pB

To scale up innovation, we need to bring research closer to real-world industrial needs. A new book by Silvia Mastellone, Alex van Delft, Alisa Rupenyan and Efe Balta looks into how that can be achieved, and where the most exciting progress is right now. #TechnologyTransfer #Automation #Control #ControlTheory https://lnkd.in/dH_vuven

Ease yourself into Monday with the latest #inControl podcast episode Host Alberto Padoan meets Francesco Bullo (UC Santa Barbara) to explore his research journey, from early work in geometric control and robotics to robotic networks, cyber-physical systems, mathematical sociology, all the way to contraction theory and neural networks! https://lnkd.in/dqPvna2k

Track 2 of the AI+X Summit, Revolutionizing Urban Mobility: Smart Solutions for Smart Cities is hosted by NCCR Automation and a number of our researchers, including John Lygeros, Florian Dorfler, Carlo Cenedese and Alberto Padoan Make sure to get your tickets before the early bird rates end on the 15th July!

Solar panels, heat pumps, batteries, ... ❓ How do you manage all these different electricity producers and consumers in your home and save money on your electricity bill? This is what Cara Koepele is investigating. Learn more: https://lnkd.in/eFqrhSXE

Do you ever think about just how complicated it is to keep a train at exactly the right temperature? In our latest #researcherblog, Ahmed Aboudonia sets out the work he has been doing with SBB CFF FFS to use #MPC to improve the automated activation of heating and cooling, resulting in energy savings and a comfortable journey. https://lnkd.in/dBAZCEXz

Kids getting bored? We have a cool activity - beginner-friendly introduction to 3D printing & CAD design. Check out the details below: When: July 29 - 31, 2024 | Monday - Wednesday | 09:00 - 16:00 Where: ETH Student Project House, Clausiusstrasse 16, 8006 Zürich Who: Girls, 12 - 16 years old Cost: CHF 200* Limited spaces. Register today! *Includes lunch. For scholarships please contact David Cleres or Lara Fritsche-Riparip #summercamp #codingcamp #girlscodetoo