Agent-oriented Requirements Modeling

A standard way of describing requirements for system design is according to function. One describes high level functions of the operation of the system and decomposes these into lower level functions which perhaps overlap. For complex systems the number of functions can be large and they may interact, making it important to have a methodical way of describing the structure of functions and their interactions. Agent-Oriented Software Engineering methodology addresses complex systems by three strategies: decomposition, abstraction, and organization. This is a useful and effective way to solve problems because it gives one a clear way to decompose the problem into agents. In addition, an agent-based system is endowed with some degree of automation and self-adaptation in that agents are autonomous entities and they are “intelligent” in some aspects. This makes the system more flexible and robust. In this paper, we demonstrate our belief that, agent technology, with its virtue of domain knowledge capture capability and high-level abstraction for interactions, can be applied to Requirements Engineering. In Section1 of the paper we demonstrate how we are inspired to adopt extended UML and XML for representing and encoding requirements knowledge. We compare agent-oriented requirements modeling with traditional descriptive functional-based requirements representation and argue that the new modeling approach has some fundamental advantages. In Section2 we introduce related work on agent-oriented Software Engineering, Requirements Engineering and AUML, an integration of these techniques brings us distinctive features and benefits. In Section3, we give readers an impression of what the new requirements look like and the steps to build agent-oriented UML diagrams including identification of agents and their interactions. Later on we discuss the detailed notation systems and apply the agent-oriented approach to represent a portion from an existing Rail Track requirements document in Section4. Then we introduce in Section5 our specially designed agent-oriented UML CASE tool and illustrate how to do requirements transformation in detail with this tool, in this section we also introduce the tool’s integrated functionalities of automatic framework code generation and architecture consistency validation. Finally we evaluate our approach, make a conclusion and discuss open issues and possible further work.