PACMAS: A Personalized, Adaptive, and Cooperative MultiAgent System Architecture

In this paper, a generic architecture, designed to support the implementation of applications aimed at managing information among different and heterogeneous sources, is presented. Information is filtered and organized according to personal interests explicitly stated by the user. User pro- files are improved and refined throughout time by suitable adaptation techniques. The overall architecture has been called PACMAS, being a support for implementing Personalized, Adaptive, and Cooperative MultiAgent Systems. PACMAS agents are autonomous and flexible, and can be made personal, adaptive and cooperative, depending on the given application. The peculiarities of the architecture are highlighted by illustrating three relevant case studies focused on giving a support to undergraduate and graduate students, on predicting protein secondary structure, and on classifying newspaper articles, respectively.