Dependable Control and Decision group

Performance and safety guarantees from NOISY data! *New Preprint*

Published:13.06.21 — Solely using noisy historical data, we synthesize a robustly safe output-feedback policy, for which the suboptimality and safety gaps are explicitly and tightly upperbounded in terms of the level of corrupting noise. This is achieved by developing a new tractable optimal control method, the Behavioral Input-Output Parametrization (BIOP), which first exploits behavioral theory to identify a precise impulse response, and then sets up a quasi-convex optimal control procedure with guaranteed high performance.  Check out our preprint! 

State estimation using data: we have a new preprint on arXiv!

Published:01.06.21 — Our new technical report "Data-driven Unknown-input Observers and State Estimation" leverages the recently developed behavioral system theory to propose a completely data-driven framework for unknown-input observers and regular state estimation. 

Our technical report on data-driven control is out

Published:01.06.21 — The paper, entitled "A Data-Driven Convex Programming Approach to Worst-Case Robust Tracking Controller Design", uses behavioral system theory to provide a convex-programming-based framework for solving worst-case optimal tracking problems in a data-driven fashion. It also guarantees that input/output constraints are always satisfied when the output data are noisy and inputs have disturbances.

We present our poster in L4DC Conference 2021

Published:01.06.21 — Our paper on data-driven robust optimal control has been accepted to the 3rd Annual Learning for Dynamics and Control (L4DC) Conference. It will be virtually presented on June 7, 2021.

New Arxiv techrep: Hamiltonian DNN have no vanishing gradients!

Published:31.05.21 — In our ArXiv work we prove that Hamiltonian deep networks do not suffer from vanishing gradients, if they are suitably designed.

Hamiltonian DNNs will be presented in L4DC Conference 2021

Published:31.05.21 — Our paper on Hamiltonian deep neural networks has been accepted in the 3rd Annual Learning for Dynamics & Control (L4DC) Conference. It will be presented during the poster sessions on June 7 – 8, 2021.

Pulkit Nahata obtains his PhD: congratulations!

Published:29.05.21 — Following a successfull presentation of the work of his thesis, entitled "Hierarchical Control of Islanded Microgrids with Flexible Structures", Pulkit Nahata was awarded the PhD from the Doctoral School of Electrical Engineering at EPFL. Pulkit is the first student graduating from the DECODE group: congratulations on this wonderful achievement and best wishes for your career!!

Mahrokh's portrait for the #NCCRWomen campaign

Published:16.04.21 — New video! Mahrokh Ghoddousi Boroujeni, PhD student of the DECODE group and member of the NCCR Automation, describes how she uses mathematical tools and data analysis for controlling the operation of the electricity grid.

New paper in Optimal Control Applications and Methods

Published:16.04.21 — Our paper on clustering of large-scale electric networks for facilitating the deployment of optimization-based energy management systems has been published in OCAM.