Lecture Notes in Computer Science vol:3394 pages:88-105
Fourth Symposium on Adaptive Agents and Multi-Agent Systems location:Leeds, UK date:March 29 - April 1, 2004
So far, the main focus of research on adaptability in multi-agent systems (MASs) has been on the agents' behavior, for example on developing new learning techniques and more flexible action selection mechanisms. In this paper, we introduce a different type of adaptability in MASs, called time management adaptability. Time management adaptability focuses on adaptability in MASs with respect to execution control. First, time management adaptability allows a MAS to be adaptive with respect to its execution platform, anticipating arbitrary and varying timing delays which can violate correctness. Second, time management adaptability allows the execution policy of a MAS to be customized at will to suit the needs of a particular application. We discuss the essential parts of time management adaptability: (1) we employ time models as a means to explicitly capture the execution policy derived from the application's execution requirements, (2) we classify and evaluate time management mechanisms which can be used to enforce time models, and (3) we introduce a MAS execution control platform which combines both previous parts to offer high-level execution control.
Adaptive Agents and Multi-Agent Systems III: Adaptation and Multi-Agent Learning