Dynamic Non-Functional Requirements Based Model-Driven Agent Development

The Belief-Desire-Intention (BDI) agent model is one of the more favoured models used to develop various agents due to both its flexibility and conceptual adaptability. This is mainly with regards to its ability to distinguish between a series of plans through a tailored plan selection function such that the plan most suitable to the current situation is selected. Since BDI agents were first proposed however, their creation process has mainly been entirely functional. This means that in order to properly implement them, specific expertise and knowledge is required. In sight of this, we propose in this paper an entirely model-driven approach based on the Extended Non-Functional Requirements framework. This approach allows agents to be directly derived from an accompanying extended NFR-framework model. The developer therefore only requires knowledge and expertise in this specific framework, which is significantly easier to understand. This derived agent can then select plans from the model's entities with respect to their hierarchy, and influenced by any contributions that they may make; selecting plans which exhibit the highest benefit to the overall system. This model-driven approach is focused on operating within a dynamic environment, which to our knowledge, is the first of its kind. This approach is evaluated experimentally.