MASK : An AEIO Toolbox to Design and Build Multi-Agent Systems

Michel Occello (1) , Christof Baeijs (1) , Yves Demazeau (2, 1) , Jean-Luc Koning (1)

1 Leibniz - IMAG - Laboratoire Leibniz
2 CNRS - Centre National de la Recherche Scientifique

This article presents a development platform for designing and building multi-agent systems called MASK. The platform is organised along a set of four toolboxes, each of them covering one aspect of the AEIO (or "VOWELS") approach. This approach decomposes a multi-agent system into four different bricks: the agents (A), the environments (E), the interactions (I) and the organisations (O). The MASK platform offers for each of these basic entities the possibility t o reuse existing models and software components, as well as the possibility to design and build new ones. Throughout this article we present the existing models for the AEIO bricks, we show how they have been integrated in the MASK development platform, and we partly illustrate our AEIO toolbox approach with a concrete example in collective robotics. 1. Designing and Building Multi-Agent Systems 1.1. A MAS platform MASK (Multi-Agent System Kernel) is a software package to design and build multi-agent systems. This toolbox is used after the analysis stage in which the conceiver works out a detailed solution without taking into consideration the resources eventually required. Multi-agent analysis consists in breaking up a problem into a multi-agent solution. It attempts to specify conceptual agents and determines their skills and knowledge. The analysis phase can be tackled starting from the agents, the interactions, the organisations or the environment. The design stage's purpose is then to lead to an implementation of the envisioned application and to choose how to make the chosen models operational. Multi-agent-oriented design aims at building a Multi-Agent System (MAS) once what the agents have to do is known. A design approach is meant for leading to an operational MAS, i.e., integrating the agents, environment, interactions, and organisation within a MAS starting from the global specification drawn from the analysis stage. The main goal of the MASK platform is to provide the multi-agent system designer with a number of utilities packages embedded in a single software environment.

Domaines: Système multi-agents [cs.MA]; Intelligence artificielle [cs.AI]

Michel Occello, Christof Baeijs, Yves Demazeau, Jean-Luc Koning. MASK : An AEIO Toolbox to Design and Build Multi-Agent Systems. Knowledge Engineering and Agent Technology, IOS, 2004. ⟨hal-02189977⟩