Probabilistic validation of deep learning-based MPC controllers
Event details
| Date | 28.02.2020 |
| Hour | 11:00 › 12:00 |
| Speaker | Prof. Sergio Lucia - TU Berlin |
| Location | |
| Category | Conferences - Seminars |
Abstract:
Solving nonlinear model predictive control problems in real time is still an important challenge despite of recent advances in computing hardware, optimization algorithms and tailored implementations. This challenge is even greater when uncertainty is present due to disturbances, unknown parameters or measurement and estimation errors. To enable the application of advanced control schemes to fast systems and on low-cost embedded hardware where exact explicit MPC is not possible, recent works have proposed to approximate a complex model predictive controllers using deep learning.
This talk focuses on probabilistic validation techniques that can be used to compute safe sates or general performance guarantees that can be used together with constraint tightening techniques. The potential of the proposed approach is demonstrated with simulation results of several linear and nonlinear examples.
Bio:
Sergio Lucia obtained his M.Sc. in Electrical Engineering from the University of Zaragoza in 2010 and the Dr.-Ing. degree from TU Dortmund in the field of optimization and automatic control in 2014. He then joined the Otto-von-Guericke Universität Magdeburg, and visited the Massachusetts Institute of Technology as a Postdoctoral Fellow in 2016.
Since May 2017, he is Assistant Professor and holds the chair “Internet of Things for Smart Buildings” at the TU Berlin and the Einstein Center Digital Future. His research efforts focus on decision-making under uncertainty, distributed control, machine learning and embedded optimization in the framework of the Internet of Things. Applications of interest include smart buildings and energy systems. Prof. Lucia is an active member of IFAC and IEEE CSS, he is Associate Editor of the Journal of Process Control and currently IT Chair of the National Organizing Committee for the IFAC World Congress 2020.
Solving nonlinear model predictive control problems in real time is still an important challenge despite of recent advances in computing hardware, optimization algorithms and tailored implementations. This challenge is even greater when uncertainty is present due to disturbances, unknown parameters or measurement and estimation errors. To enable the application of advanced control schemes to fast systems and on low-cost embedded hardware where exact explicit MPC is not possible, recent works have proposed to approximate a complex model predictive controllers using deep learning.
This talk focuses on probabilistic validation techniques that can be used to compute safe sates or general performance guarantees that can be used together with constraint tightening techniques. The potential of the proposed approach is demonstrated with simulation results of several linear and nonlinear examples.
Bio:
Sergio Lucia obtained his M.Sc. in Electrical Engineering from the University of Zaragoza in 2010 and the Dr.-Ing. degree from TU Dortmund in the field of optimization and automatic control in 2014. He then joined the Otto-von-Guericke Universität Magdeburg, and visited the Massachusetts Institute of Technology as a Postdoctoral Fellow in 2016.
Since May 2017, he is Assistant Professor and holds the chair “Internet of Things for Smart Buildings” at the TU Berlin and the Einstein Center Digital Future. His research efforts focus on decision-making under uncertainty, distributed control, machine learning and embedded optimization in the framework of the Internet of Things. Applications of interest include smart buildings and energy systems. Prof. Lucia is an active member of IFAC and IEEE CSS, he is Associate Editor of the Journal of Process Control and currently IT Chair of the National Organizing Committee for the IFAC World Congress 2020.
Practical information
- General public
- Free
Organizer
- Laboratoire d'Automatique (LA) - Emilio Maddalena