Video recordings: https://www.youtube.com/watch?v=O8u5X0d-oXE
Date: June 27th 2022, 15h30 - 17h20 (CEST)
Organizers:
- Timm Faulwasser, TU Dortmund, Germany; timm.faulwasser@tuhh.de
- Thulasi Mylvaganam, Imperial College London, UK; t.mylvaganam@imperial.ac.uk
Schedule:
| Time (CEST) | Title and Speaker | |
| 15h30 | Welcome | |
| 15h35 | A software framework for nonlinear model predictive control on embedded hardware (GRAMPC)
Knut Graichen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany | |
| 16h25 | Exponential Decay of Sensitivity in Graph-Structured Nonlinear Programs
Victor Zavala, University of Wisconsin - Madison, USA | |
Speaker: Knut Graichen (Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany)
Title: A software framework for nonlinear model predictive control on embedded hardware (GRAMPC)
Abstract:
Model predictive control (MPC) nowadays is an established control method with significant advances over the last decade, for instance, concerning computational efficiency. Nevertheless, the application of MPC to fast systems with sampling times in the range of (sub-)milliseconds and weak computational hardware such as lean embedded automotive ECUs remains a challenge to this day.
The talk presents the MPC toolbox GRAMPC that is based on an augmented Lagrangian framework for continuous time nonlinear systems that was developed with the focus on real-time feasibility, minimal computational requirements and portability to embedded platforms with limited resources. The talk gives an introduction to the methodological background and applicability of GRAMPC and demonstrates its performance in comparison with state-of-the-art MPC solvers as well as for relevant industrial applications. The talk also outlines some recent extensions of GRAMPC, e.g. for networked systems.
Speaker: Victor Zavala (University of Wisconsin - Madison, USA)
Title: Exponential Decay of Sensitivity in Graph-Structured Nonlinear Programs
Abstract:
We study solution sensitivity for nonlinear programs (NLPs) whose structures are induced by graphs. These NLPs arise in many applications such as dynamic optimization, stochastic optimization, optimization with partial differential equations, and network optimization. We show that for a given pair of nodes, the sensitivity of the primal-dual solution at one node against a data perturbation at the other node decays exponentially with respect to the distance between these two nodes on the graph. In other words, the solution sensitivity decays exponentially as one moves away from the perturbation point. This result, which we call exponential decay of sensitivity, holds under fairly standard assumptions: the strong second-order sufficiency condition and the linear independence constraint qualification. We discuss how this property provides new and interesting insights on the behavior of complex systems and how it enables the design of new and scalable decomposition algorithms.
