Akce ČVUT

Přednáška - dr. Lukáš CHRPA - Reformulating Planning Problems

16.12.2013   14:30-15:30

Abstract:

Despite a significant progress in developing AI planning techniques many planning problems remain hard and challenging. Current planning engines although often using informative heuristics are still vulnerable to certain peculiarities in planning domains and problems. Learning and exploiting characteristic knowledge which can be identified in many planning domains and/or classes of planning problems seems to be a very promising way how to improve the planning process. Reformulating planning domains and/or problems is a reasonable way of encoding such knowledge directly into domain and problem descriptions because such an approach is independent on used planning engines. Macro-operators that encapsulate sequences of original planning operators are well known type of knowledge which provides 'shortcuts' in the search space. Another type of knowledge, entanglements that are relations between planning operators and predicates, is used to prune unpromising alternatives in the search space.

Místo konání

Karlovo nám. 13, budova A (strojní fakulta), 4. patro, cislo 404

Pořadatel

Katedra kybernetiky a katedra počítačů FEL ČVUT

Kontaktní osoba

Ing. Veronika Šínová, sinova@fel.cvut.cz, 224357667