[Club2] Einladung: SMT-based AI planning - Mi 21. Okt. 2020 14:00 - 14:20 (MESZ) (club2 at mailbroy.informatik.tu-muenchen.de)

[piller at in.tum.de]

Sie wurden zum folgenden Termin eingeladen.

Titel: SMT-based AI planning

Speaker: Malek Souissi

Type: Bachelor's Thesis Presentation

Abstract: A successful technique for AI planning, introduced by  Henry  
Kautz and Bart Selman in [KS92], is based on propositional satisfiability  
(SAT). The main idea lies in encoding a planning problem into a  
propositional formula that is satisfiable iff a plan exists for the given  
problem. A plan can then be extracted from the satisfying model. In some  
cases, it is of higher interest to find the optimal plan for a given  
problem (i.e. a plan with the lowest possible cost). Our aim is to adapt  
the SAT-based planning approach to solve the optimal planning problem by  
introducing action costs in an SMT-encoding of the problem. However, one  
crucial requirement for the completeness and efficiency of this approach is  
the completeness threshold. That is a bound k, such that, if no plan exists  
of length less or equal to k, then no plan exists for the corresponding  
problem (with a cost bound in case of optimal planning).

In this thesis we first port existing code for computing upper bounds on  
the diameter from c++ to sml. We then improve upon the existing algorithms  
and experiment with different approaches and configurations. In the second  
part of the thesis, we introduce an any-time algorithm for optimal planning  
which iteratively finds plans with lower costs until an optimal plan is  
found or a timelimit is reached. In this part, we experiment with different  
SMT-encodings and also SMT-solvers. We then try out some techniques that  
potentially reduce the time needed to prove optimality by relaxing the  
problem and assess the difference in performance.
Wann: Mi 21. Okt. 2020 14:00 – 14:20 Mitteleuropäische Zeit - Berlin
Wo: ls21coffee

Teilnahmeinformationen: Teilnehmen mit: Google Meet
https://meet.google.com/jci-yqqc-doe

Kalender: club2 at mailbroy.informatik.tu-muenchen.de
Wer
     * piller at in.tum.de- Veranstalter
     * ge72yet at mytum.de
     * club2 at mailbroy.informatik.tu-muenchen.de

Termininformationen:  
https://www.google.com/calendar/event?action=VIEW&eid=NXAyMmY4Z2dsb2JmcTNlNzNqNXVuZDl0YWggY2x1YjJAbWFpbGJyb3kuaW5mb3JtYXRpay50dS1tdWVuY2hlbi5kZQ&tok=NTIjc2U2ZWJlM3RvZmY0Y2g1bm11bmlibTVtOThAZ3JvdXAuY2FsZW5kYXIuZ29vZ2xlLmNvbWQ5MDQ4NWY1ODM2MTBiZGI2ZDVkYjFkNzJjNzg0NDE0MzQ3NDUxYTQ&ctz=Europe%2FBerlin&hl=de&es=0

Einladung von Google Kalender: https://www.google.com/calendar/

Sie erhalten diese E-Mail unter club2 at mailbroy.informatik.tu-muenchen.de,  
da Sie ein Gast bei diesem Termin sind.

Lehnen Sie diesen Termin ab, um keine weiteren Informationen zu diesem  
Termin zu erhalten. Sie können auch unter https://www.google.com/calendar/  
ein Google-Konto erstellen und Ihre Benachrichtigungseinstellungen für  
Ihren gesamten Kalender steuern.

Wenn Sie diese Einladung weiterleiten, kann jeder Empfänger eine Antwort an  
den Organisator senden und zur Gästeliste hinzugefügt werden. Außerdem  
könnte er weitere Nutzer einladen und Ihre Antwort ändern. Weitere  
Informationen: https://support.google.com/calendar/answer/37135#forwarding
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman46.in.tum.de/pipermail/club2/attachments/20201020/2bfc0f19/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/calendar
Size: 3258 bytes
Desc: not available
URL: <https://mailman46.in.tum.de/pipermail/club2/attachments/20201020/2bfc0f19/attachment-0001.ics>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: invite.ics
Type: application/ics
Size: 3317 bytes
Desc: not available
URL: <https://mailman46.in.tum.de/pipermail/club2/attachments/20201020/2bfc0f19/attachment-0001.bin>