BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Memento EPFL// BEGIN:VEVENT SUMMARY:Learning with Guarantees: Neural Feedback Controllers with Built-i n Stability and Robustness Certificates DTSTART:20251103T150000 DTEND:20251103T153000 DTSTAMP:20260430T201403Z UID:ae745c41140daeb52ac0fcf956beba35b251b72a2ac0bc0cf3af59e4 CATEGORIES:Conferences - Seminars DESCRIPTION:Nicholas Barbara\nAustralian Centre for Robotics\, The Univers ity of Sydney\nLearning-based control strategies such as deep reinforcemen t learning (RL) and imitation learning (IL) are powerful tools for general -purpose robotic control. They rely on simple\, gradient-based optimisatio n schemes\, and typically parameterise control policies with black-box dee p neural networks\, which are known to be universal approximators for nonl inear systems. However\, black-box approaches to feedback control suffer f rom a fundamental lack of closed-loop stability and robustness. In this ta lk\, we consider the problem of learning stabilising neural policies with built-in stability and robustness guarantees for nonlinear systems. Our ap proach leverages recent developments in robust neural networks\, which are neural networks that automatically satisfy internal stability and robustn ess properties of their own. We present two novel approaches: (1) parametr ising control policies with robust neural networks to improve their empiri cal robustness in open loop\; and (2) parametrising control policies by co mbining robust Recurrent Equilibrium Networks (RENs) with a nonlinear vers ion of the classical Youla-Kucera parametrisation to achieve closed-loop s tability (contraction) and robustness (Lipschitz) guarantees. Our resultin g "Youla-REN" policy class is parametrised to automatically satisfy these closed-loop guarantees\, making it compatible with standard gradient-based optimisation pipelines deep RL/IL. We discuss our recent theoretical resu lts\, provide illustrative numerical examples\, and pose directions for fu ture research on the application of robust learning-based control in real- world robotic systems.\n\nNicholas Barbara is a PhD candidate at the Austr alian Centre for Robotics\, within the University of Sydney. He received h is Bachelor of Engineering (Hons 1\, University Medal) and Bachelor of Sci ence from the University of Sydney\, Australia\, in 2021. His current rese arch involves developing new tools to learn control policies with stabilit y and robustness guarantees for robotic systems. His research interests in clude learning-based control\, robust machine learning\, robotics\, and sp acecraft GNC. \n  LOCATION:ME A3 31 https://plan.epfl.ch/?room==ME%20A3%2031 https://uni-syd ney.zoom.us/j/83218340533 STATUS:CONFIRMED END:VEVENT END:VCALENDAR