Janick Frasch

Institute for Mathematical Optimisation (OvGU Magdeburg) Department of Electrical Engineering (KU Leuven)

Thursday, October 23, 2014, 16:00 - 17:00

Room 01-012, Georges-Köhler-Allee 102, Freiburg 79110, Germany

Sparsity exploitation is a crucial feature of convex quadratic programming algorithms in the context of dynamic optimization, particularly when tackling problems of long prediction or estimation horizons. As an alternative to established approaches (e.g., interior-point or condensing-based methods) we investigate the use of a dual Newton strategy that can tackle block structured QPs directly, but still exhibits desirable warmstarting capabilities. In the first part of the talk, we shed light on some details regarding the efficient implementation of this approach and report on the state of the open-source implementation qpDUNES. In the second part of the talk, we address a selection of open ideas for extensions of qpDUNES both regarding performance and reliability gains, as well as increased applicability of the method

Download Slides