Skip to main navigationSkip to main content
The University of Southampton
Southampton SIAM Student Chapter

A Large-Scale Reformulation Technique for Nonlinear Bilevel Programming Based on Sequential Quadratic Programming Seminar

15:00 - 16:00
28 November 2018
The Ketley Room (B54 / lvl 4)

For more information regarding this seminar, please email Marton Benedek at .

Event details

A technical system's dynamic behavior is often qualitatively known such that a general model can be formulated, but parameter values within the model are unknown since they depend on system specifications. In this talk, I will firstly give a brief introduction to the field of parameter identification for dynamical systems. Transcription techniques, in particular the single shooting and the full discretization approach, will be presented, as well as some illustrating examples. The second part of the talk will be about nonlinear bilevel optimization. Similar to parameter identification, this problem class exhibits highly nonlinear constraints, since a lower level problem has to be solved. I will present a strategy to reformulate it as a single level problem, in which a set of constraints is introduced that represents the process of solving the lower level problem based on a sequential quadratic programming method. This requires additional optimization variables and finally leads to a sparse, but large-scale NLP. The focus is on providing ideas and first numerical results.

Speaker information

Kai Schäfer, University of Bremen. Center for Industrial Mathematics

Privacy Settings