On the Definition of Meta-models for Analysis of Large-Scale MAS

Multi Agent Systems (MASs) are undergoing a transition from closed monolithic architectures into open architectures composed of a huge number of autonomous agents. Large-scale MASs involve hundreds or perhaps thousands of agents to share a common goal. However, the current practice of MAS design tends to be limited to individual agents and small face-to-face groups of agents. With this kind of approaches it is very difficult to model large-scale MAS. To cope with this limitation an Abstract Agent can be used as a modelling artifact for MAS analysis and design. An Abstract Agent may represent, at modelling stage, groups of agents (MAS, organization, federation, etc.) which can take on the qualities of being an agent. An Abstract Agent acting in organizational structures can encapsulate the complexity of group of agents (simplifying representation and design) and modularize its functionality. In MAS analysis and design there are some models commonly used to specify MAS. In this work we present extensions to these models in order to include the Abstract Agent into them.