No abstract

Development of software agents according to belief-desire-intention (BDI) model usually becomes challenging due to autonomy, distributedness, and openness of multi-agent systems (MAS). Hence, here, a domain-specific modelling language (DSML), called DSML4BDI, is introduced to support development of BDI agents. The syntax of the language provides the design of agent components required for the construction of the system according to the specifications of BDI architecture. The implementation of designed MAS on Jason BDI platform is also possible via model-to-text transformations built in the DSML. The comparative evaluation results showed that a significant amount of artefacts required for the exact MAS implementation can be automatically achieved by employing DSML4BDI. Moreover, time needed for developing a BDI agent system from scratch can be reduced to one-third in the case of using DSML4BDI. Finally, qualitative assessment, based on the developers' feedback, exposed how DSML4BDI facilitates development of BDI agents.

In agent-oriented software engineering (AOSE), the application of model-driven development (MDD) and the use of domain-specific modeling languages (DSMLs) for Multi-Agent System (MAS) development are quite popular since the implementation of MAS is naturally complex, error-prone, and costly due to the autonomous and proactive properties of the agents. The internal agent behavior and the interaction within the agent organizations become even more complex and hard to implement when the requirements and interactions for the other agent environments such as the Semantic Web are considered. Hence, in this study, we propose a modeldriven MAS development methodology which is based on a domain-specific modeling language (called SEA_ML) and covers the whole process of analysis, modeling, code generation and implementation of a MAS working in the Semantic Web according to the well-known Belief-Desire-Intention (BDI) agent principles. The use of new SEA_ML-based MAS development methodology is exemplified with the development of a semantic web-enabled MAS for electronic bartering (Ebarter). Achieved results validated the generation and the development-time performance of applying this new MAS development methodology. More than half of the all agents and artifacts needed for fully implementing the E-barter MAS were automatically obtained by just using the generation features of the proposed methodology.