[Ipopt] Handling unbounded solution sets

[Martin Schmidt]

Hi everybody,

I have a question concerning Ipopt's strategy for handling unbounded solution sets (see 3.7 in the implementation paper). In Sect. 3.7 it is stated that the objective function of the barrier problem is modified in a way such that divergence of variables with only one bound is penalized. Am I right in assuming that this "heuristic" affects all occurrences of the objective function of the barrier problem and its gradient. In particular, are the additional penalization terms at place in Eq. (18b), (19) and (20) in the filter line-search procedure.

Thanks in advance,
Martin Schmidt

---
Dipl.-Math. Martin Schmidt
AG Algorithmische Optimierung
Institut für Angewandte Mathematik
Leibniz Universität Hannover
Welfengarten 1, 30167 Hannover
Tel.: 0511-7623873
E-Mail: mschmidt at ifam.uni-hannover.de
Web: http://www.ifam.uni-hannover.de/~mschmidt/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20121121/d4940227/attachment.html>