IRIT - UMR 5505

MAY (Make Agents Yourself) is a tool to build architectures usable to support development and execution of multi-agent systems. MAY is released as an Eclipse plugin (that relies on the Xtext plugins).

Its main objective is to reduce the development effort of agent-based applications by bringing the development support closer to the business needs through the building of component-based software architectures and reuse of components. MAY is a part of an approach to better take into account the functional and non-functional requirements of such application by explicitly separating the production of the multi-agent system design from the production of the abstractions the latter relies on. Development is decomposed in two phases:

• The first one is about building an architecture with holes adapted to the needs of the development. In particular, the result of the design of the multi-agent system is an important element to guide this phase.
• The second one is about the programming of the behaviours of the agents of the multi-agent system using the architecture with holes.

MAY can thus be considered a tool to specify and implement a multi-agent platform kernel.

MAY is based on the building of home-made “species of agents” dedicated to the application. A species of agents defines the internal mechanisms that govern the execution of the agents (lifecycle, decision mechanisms…) as well as the mechanisms for communicating and interaction with their environment (sensors and actuators). A species of agents is defined inside an “ecosystem” that manages the platform aspect of the application architecture. An ecosystem defines the mechanisms to communicate and interact usable by the species of agents, but also everything that govern or concern the system globally (scheduling, graphical interfaces, agents’ environment…).

A species of agents and its ecosystem define the operational semantics of agents and can be interpreted as the execution engine of their behaviours. The mechanisms are software components with provided and required services that can be composed together to form the architectures of the species and of the ecosystem. The architecture is described using the textual architecture description language SpeADL (Species-based Architecture Description Language). The user of MAY first specifies the species of agents and their ecosystem using the SpeADL language. Then these specification are transformed to Java in order to allow the implementation of the architectural elements (the components). The generated part takes care of the mechanisms that make the dynamic creation and connection of the agents to the platform possible. This reduces as much as possible the development effort.

MAY is available with a library of reusable components that are dedicated to multi-agent systems.

Using MAY

MAY is a free software (LGPL) developed by the SMAC Team from IRIT as part of its work on the development support for multi-agent systems and is available as an Eclipse plugin.

Practical information (source code, tutorial, examples, mayling-list…) is available on the IRIT forge.

References

• Victor Noel, Jean-Paul Arcangeli, Marie-Pierre Gleizes. Une approche architecturale à base de composants pour l’implémentation des Systèmes Multi-Agents. In: Revue des Nouvelles Technologies de l’Information, Cépaduès Editions, 2012 (in press).
• Victor Noel, Jean-Paul Arcangeli. Frameworks, architectures et composants: revisiter le développement de systèmes multi-agents. In: Conférence Francophone sur les Architectures Logicielles (CAL 2011), Lille, 07/06/2011-08/06/2011, Philippe Aniorté (Eds.), Laboratoire d’Informatique Fondamentale de Lille, p. 23-32, June 2011.
• Victor Noel, Jean-Paul Arcangeli, Marie-Pierre Gleizes. Between Design and Implementation of Multi-Agent Systems: A Component-Based Two-Step Process. In: European Workshop on Multi-Agent Systems (EUMAS 2010), Paris (France), 16/12/2010-17/12/2010, Paris Descartes University, December 2010.