I.J.Modern Education and Computer Science, 2013, 1, 42-55 Published Online January 2013 in MECS (http://www.mecs-press.org/) DOI: 10.5815/ijmecs.2013.01.06 Copyright © 2013 MECS I.J. Modern Education and Computer Science, 2013, 1, 42-55 An Ontology-Based Approach for Multi-Agent Systems Engineering Souleymane Koussoube Institut Africain d’Informatique (IAI), Libreville, Gabon Email: skoussoube@yahoo.fr Armel Ayimdji Département de Genie Informatique, Institut Universitaire de Technologie, Université de Douala, Cameroun Email: ayimdji@gmail.com Laure Pauline Fotso Département d’Informatique, University de Yaoundé I, Cameroun Email: laurepfotso@yahoo.com Abstract— This paper presents OBAMAS (Ontology- Based Approach for Multi-Agent Systems engineering), an ontology-based contribution to Agent Oriented Software Engineering. We propose a formal process for agentification, starting with an analysis phase which consists of the construction of three formal ontologies (a domain ontology, an ontology of functionalities, and an ontology of multi-agents systems) and their alignment to merge in a single one. The second step, which is a design phase, consists of the operationalization of the single ontology in order to infer in a more formal way the agents of the system. A case study is introduced to illustrate OBAMAS and to show its use and effectiveness in a real application, a distance learning system. Index Terms— Multi-agent system, Ontology, Ontology alignment, Operationalization, Description logics I. I NTRODUCTION These last years, progress in software engineering were realized by the development of more and more complex, dynamic and often distributed systems. These progresses sometimes consisted in granting more autonomy to software in order to gain efficiency and robustness. These efforts gave birth to multi-agents systems (MAS) [1] and then, to Agent Oriented Software Engineering (AOSE) whose ambition is to provide specific methods for MAS development. Indeed, although most of MAS development methods are based on already existing software engineering paradigms such as object oriented (OO) or knowledge-based systems (KBS), these paradigms are not really suitable for MAS. There are fundamental differences between the OO view and the Agent Oriented (AO) view. For example, to set up MAS, we have to take into account the autonomy of agents. Agents embody a stronger notion of autonomy than objects, and, in particular, they decide for themselves whether or not to perform an action on request of another agent. This distinction between objects and agents has been nicely summarized in the following slogan: “Objects do it for free; agents do it because they want to” [1]. AOSE methodologies were developed to deal with these peculiarities of agents. Our contribution is in this context. We propose OBAMAS (Ontology-Based Approach for Multi-Agents Systems engineering), an approach for MAS development based on the intensive usage of ontologies. The state of the art on AOSE methodologies [1, 2, 3, 4, 5, 6, 8] shows that many of them does not provide an explicit step of agents identification. In practice, the decomposition of the system into agents is essentially “intuitive”. Our objective in this work is to define a more formal approach for the identification of agents (agentification) that is to provide a complete guide with a set of formalized rules allowing inferring in an objective way the agents of MAS. The present work extends a previous one introduced in [9] by bringing a better structure of the approach, the deepening of the various steps, as well as a comparison with some existing methodologies. The rest of this article is organized as follows. Section II recalls the notions of agents and MAS, and then ontologies and their contributions in MAS. We present two key concepts in section III: the alignment and the operationalization of ontologies. In Section IV, we give a quick presentation of OBAMAS through the description of its different steps followed by an illustration of the methodology with a representative case study (a distance learning system) to show its use and effectiveness in a real application. In section V, we compare our methodology to related works in literature. A summary of the special features of our methodology is presented in section VI while section VII concludes this work. II. AGENTS, MULTI -AGENTS SYSTEMS AND ONTOLOGIES Agents have been studied and defined in several ways. We adopt a definition which is very often used in the MAS community: