Adaptive Agent Model: Software Adaptivity using an Agent-oriented Model-Driven Architecture

Model-Driven Architecture (MDA) promotes the development of software systems through successive building and generation of models, improving the reusability of models. Applying the same principles to the area of Agent-Oriented Software Engineering (AOSE) advances the ideas behind MDA even more significantly, due to the inherent adaptivity of software agents We describe an appropriate set of models originating from requirements specification and transformable to models understandable and executable by agents thus demonstrating an Agent-oriented Model-Driven Architecture (AMDA) approach. In AMDA, agents use hierarchical business knowledge models with business process rules at the top, business rules to control policy and logic in the middle and a base layer defining business concepts. Being externalised, knowledge is easily configurable by human beings and applied by software agents. A real case study is used to illustrate the process. The main advances over the object-oriented MDA are (i) the addition of component dynamics (ii) the use of agent-executable rule-based business models and (iii) a proposed higher level of abstraction with the direct representation of business requirements.