Supporting the platform extensibility for the model-driven development of agent systems by the interoperability between domain-specific modeling languages of multi-agent systems

Abstract

The conventional approach currently followed in the development of domain-specific modeling languages (DSMLs) for multi-agent systems (MASs) requires the definition and implementation of new model-to-model and model-totext transformations from scratch in order to make the DSMLs functional for each different agent execution platforms. In this paper, we present an alternative approach which considers the construction of the interoperability between MAS DSMLs for a more efficient way of platform support extension. The feasibility of using this new interoperability approach instead of the conventional approach is exhibited by discussing and evaluating the model-driven engineering required for the application of both approaches. Use of the approaches is also exemplified with a case study which covers the model-driven development of an agent-based stock exchange system. In comparison to the conventional approach, evaluation results show that the interoperability approach requires both less development time and effort considering design and implementation of all required transformations.

Refactoring Legacy Software for Layer Separation

“…As another possible study, the separation mechanism can be performed with the use of domain-specific languages and model transformations as we experienced (e.g. in [48], [49], [50], [51]) for software agent architecture models. In this way the model of the layered architecture can be generated and re-used in later extensions.…”

AgentDSM-Eval: A framework for the evaluation of domain-specific modeling languages for multi-agent systems

“…Second, a single evaluator group was used instead of two different groups, which could pose a threat to the execution phase. We experienced using both single and double evaluator groups in our previous empirical studies on evaluating both different MAS DSMLs (e.g., Kardas et al, 2017;Miranda et al, 2019) and DSMLs in other industrial domains (e.g., Saritas and Kardas, 2014;Arslan and Kardas, 2020). Using a single group may raise the risk that the evaluators take advantage of their prior development experience using the MAS DSML while developing the same MAS without using this MAS DSML (or vice-versa).…”