DocumentCode :
2851397
Title :
Dynamic social knowledge: a cognitive multi-agent system cooperation strategy
Author :
Da Costa, Augusto Cesar Pinto Loureiro ; Bittencour, G.
Author_Institution :
Dept. de Autom. e Sistemas, Univ. Fed. de Santa Catarina, Florianopolis, Brazil
fYear :
1998
fDate :
3-7 Jul 1998
Firstpage :
415
Lastpage :
416
Abstract :
The adopted cooperation strategy is one of the most important aspects to make an agent community achieve a common goal. Basically there are two kinds of cooperation strategies for cognitive multi agent systems: contract net protocol and negotiation. Contract net protocol requires a small amount of knowledge about the other agents in the community, but it is not possible to persuade another agent to perform a given task or to join some agents in an agent group to perform a given task during a cooperation process. On the other hand, negotiation allows an agent to communicate directly with another agent. This direct communication allows agreements when the optimal solution is not available, but the required amount of social knowledge and the communication overhead needed to update it is much greater than in the contract net protocol. A new cooperation strategy for open cognitive autonomous multi agent systems, called Dynamic Social Knowledge is proposed. This strategy combines the advantages of both contract net protocol and negotiation, and makes intensive use of rule based reasoning
Keywords :
cognitive systems; cooperative systems; inference mechanisms; protocols; software agents; adopted cooperation strategy; agent community; cognitive multi agent system cooperation strategy; communication overhead; contract net protocol; direct communication; dynamic social knowledge; negotiation; open cognitive autonomous multi agent systems; rule based reasoning; Contracts; Data structures; Multiagent systems; Proposals; Protocols; Real time systems;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Multi Agent Systems, 1998. Proceedings. International Conference on
Conference_Location :
Paris
Print_ISBN :
0-8186-8500-X
Type :
conf
DOI :
10.1109/ICMAS.1998.699240
Filename :
699240
Link To Document :
بازگشت