Towards the Development of Agent-Based Organizations through MDD

ions for conceptual and computational modeling. Thus, the developers often require the necessary acquisition of new skills to understand and design with the MAS methodologies. As a consequence, the creation of applications is very hard and difficult for the MAS developer, because there is no agreement about a common group of components that can be used across different MAS methodologies and platforms. Therefore, a major challenge when designing MAS is to provide efficient February 28, 2012 3:6 WSPC/INSTRUCTION FILE VO ̇Agent-based ̇MDD ̇review Towards the development of agent-based organizations through MDD 3 tools that can be used by non-expert users. Synthesizing a unified set of components from existing agent-oriented methodologies is a challenge. However, the Model-Driven Development approach can facilitate and simplify the design process and the quality of agent-based software since it allows the reuse of software and transformation between models. MDD basically proposes the automatic generation of code using transformations from models that have platform-independent components. These models are translated into more specific components (or code) that depend on the execution platform, which integrates specific details about the system. In the MAS literature, researchers are beginning to strive to formulate a set of models that guide the MAS development process using the Model-Driver approach. Some works have concentrated their efforts on creating a very generic unified model for analyzing and modeling different methodologies. Some of the most significant proposals are: TAO, FAML, Agent UML(AUML), and AML. These proposals create only a conceptual framework to develop and design MAS, but they are not intended to get the MAS deployments to run on specific platforms. Other works, such as PIM4AGENT and CAFnE, have a unified meta-model (a little less generic), but these works can generate the MAS deployment to run on specific platforms. Finally, other approaches use MDD as a modelling tool for some MAS methodologies, but they only generate MAS deployments for a single platform. Some of the most significant proposals are: PASSI, TROPOS, and INGENIAS. However, despite some of the earlier proposals (MDD in MAS uses the concept of organization in their meta-model), none of them focus the organizational development as is proposed by the Virtual Organizations approach, where it is necessary to create different deployments: one for the organization level and another for the agent level. Thus, our purpose is to use the MDD approach for the design of Virtual Organizations. This work proposes an approach for developing MAS that can be implemented in different organization-oriented platforms applying the ideas of MDD. This paper first presents a relatively generic Virtual Organization meta-model, which was created mainly using a bottom-up perspective iteratively over organization-oriented agent methodologies. This paper then proposes two transformation models for translating the unified model of the Virtual Organization to two different platforms. This process generates code templates automatically (specific target deployments) and then the developer can write any additional code in these templates if deemed necessary. This allows the MAS development to be an easy and fast process. These transformations are proposed as examples, and they allow the feasibility of the proposal to be verified. The organization-oriented target platforms used are: THOMAS 18 and E-Institutions. However, this transformation process is not limited exclusively to these agent platforms but is open to other platforms, simply by defining new transformation rules. ahttp://users.dsic.upv.es/grupos/ia/sma/tools/Thomas bhttp://e-institutions.iiia.csic.es February 28, 2012 3:6 WSPC/INSTRUCTION FILE VO ̇Agent-based ̇MDD ̇review 4 J. Agüero, C. Carrascosa, M. Rebollo, V. Julián The rest of the paper is structured as follows: Section 2 briefly describes the main concepts used in this work. It reviews the different technologies and platforms used to cope with organization development and MDD in the area of MAS. Section 3 details the different meta-models as orthogonal views that describe the complete system to be modelled at a high level of abstraction. Sections 4 and 5 explain how the proposed models can be used to design and develop a complete system. The former details the steps that developer must follow. The latter shows how transformation rules can be defined to generate automatic transformations between models. THOMAS and E-Institutions have been chosen to illustrate the process, and a usage scenario is described. Finally, conclusions of this work are presented in Section 6.