A Design Framework for Generating BDI-Agents from Goal Models

Goal models have been proposed in Distributed Artificial Intelligence for guiding software agents at run-time. In these approaches goal models consist of goal graphs representing AND/OR hierarchical decomposition along with inter-dependency links representing conflicts between goals or resources needed to satisfy them. Goal-oriented software requirements engineering exploits similar goal structures to support human analysts, at design-time, in analyzing the requirements of a system-to-be. We propose the use of goal models at different abstraction levels in engineering a Multi-Agent System (MAS), that is, not only at design time, but also as a part of the agent knowledge and choice strategy, at run-time. This approach aims at addressing crucial issues in complex distributed software such as evolvability and adaptivity. We define a tool-supported design framework that allows to specify an agent goal model and to automatically generate from it fragments of a BDI agent. We devise the design process as a transformation process from design, platform-independent models, to platform-specific models and then to code, followingModel Driven Architecture ideas. The design framework is demonstrated by referring to the Tropos methodology and to the JADE/Jadex MAS platform. The process is illustrated through an example and experimental results on various goal models are discussed.