-------- Original-Nachricht -------- Betreff: 20050518.VORTRAG: Prof.Dr. Michal Pechoucek, Prague Datum: Fri, 6 May 2005 13:26:17 +0200 Von: OeFAI Information Server oefaiinf@ai.univie.ac.at An: undisclosed-recipients:;
VORTRAG *******
Oesterreichisches Forschungsinstitut fuer Artificial Intelligence(OFAI) Freyung 6/6, A-1010 Wien Tel.: +43-1-53361120, Fax: +43-1-5336112-77, Email: sec@oefai.at -------------------------------------------------------------------------
Prof.Dr. Michal Pechoucek, Czech Technical University and Gerstner Laboratory, Prague
THE USE OF ACQUAINTANCE MODELS FOR EFFICIENT CONTRACTION IN COMPLEX COMPETITIVE ENVIRONMENTS
Intelligent coordination in complex multi-agent environments requires sophisticated mechanisms for suboptimal task decomposition and efficient task delegation among agents.
Besides the quality of coordination (i.e. efficiency of decomposition and delegation) we need to handle also computational efficiency restriction such as fast response time and limited communication traffic among agents but also the amount of disclosed private knowledge while the semi-collaborative agents negotiate collaboration. We present a novel contraction mechanism based on the use of the acquaintance model, a structure where agents store the information about possible collaborators. We suggest an approach of iterative construction of partially-linear acquaintance models that is beneficial mainly in complex agent communities. The Use of the suggested approaches is an alternative to socially uniformed contraction mechanisms such as a contract-net-protocol or various private value auction techniques, that proved to be inadequate in complex negotiation domains.
This approach has shown to be useful in various application domains, especially for collaboration in semi-trusted virtual environments such as planning humanitarian aid logistics.
*********
Zeit: Mittwoch, 18. Mai 2005, 18:30 Uhr pktl.
Ort: Oesterreichisches Forschungsinstitut fuer Artificial Intelligence, OFAI Freyung 6, Stiege 6, 1010 Wien.
OESTERREICHISCHES FORSCHUNGSINSTITUT FUER ARTIFICIAL INTELLIGENCE
o.Univ.-Prof. Ing. Dr. Robert Trappl