Learning-based control

Aim

Formulating analysis and synthesis methods to better understand fundamental limitations and robustness properties of closed-loop systems featuring learning components (e.g. parameter estimator, data-driven controller) and propose improved solutions. We would like to contribute to the understanding of when and why a learning-based controller is preferable to more standard solutions and what are the properties that such a controller should have. 

Tools

We combine standard subjects in the field of systems theory such as adaptive control and robust control with methodologies from reinforcement learning and information theory.

Selected publications

  • Parsi A., Iannelli A., Smith R.S. - "An explicit dual control approach for constrained reference tracking of uncertain linear systems" - IEEE Transactions on Automatic Control (special issue: Learning for Control), 2022. Link
  • Yin M., Iannelli A., Smith R.S. - "Maximum Likelihood Estimation in Data-Driven Modeling and Control" - IEEE Transactions on Automatic Control, 2021. Link
  • Iannelli A., Smith R.S. - "A multiobjective LQR synthesis approach to dual control for uncertain plants" - IEEE Control Systems Letters, vol. 4, 2020. Link
  • Yin M., Iannelli A., Smith R.S. - "Data-Driven Prediction with Stochastic Data: Confidence Regions and Minimum Mean-Squared Error Estimates" - European Control Conference 2022. Link
  • Scampicchio A., Iannelli A. - "An update-and-design scheme for scenario-based LQR synthesis" - American Control Conference 2022. Link
  • Yin M., Iannelli A., Smith R.S. - "Maximum Likelihood Signal Matrix Model for Data-Driven Predictive Control" - Learning For Dynamics and Control Conference 2021. Link

Collaborations

M. Yin, A. Parsi, R. S. Smith (ETH Zürich)

To the top of the page