Supporting Agent-Oriented Models within the Software Life Cycle

   page       attach   
Bin Lu

Agent-oriented models provide a more natural way to conceptualize software systems than object-oriented models. However, despite the simple nature of agent-oriented models, it is not straightforward to fit them in the processes of software engineering. This thesis investigates the use of the agent-oriented models and concepts with a different way with the agent-oriented methodologies within the life cycle of software development based on models proposed by Sterling and Taveter.
In this thesis, two systematic approaches are proposed to support the use of the agent-oriented models in requirements analysis and system design. Two new models, goal table model and agent type model, are introduced to help to specify and describe software requirements and system designs. Three models, scenario model, interaction model and environment model, are enhanced to improve the traceability of requirements and design. Three new agent-oriented templates, business vision, software requirements specification, and system design description, are proposed to systematically document agent-oriented systems and allow them to be used in standard software engineering processes. The three templates use agent-oriented models and comply with software industrial standards. A new tool, Pictorial Agent Nexus for Development and Analysis (PANDA), is analyzed and designed to support the use of agent-oriented models. A structure, System Evolution Directory, is proposed to help to organize models produced in the life cycle of software development. We also study the implementation of agent-oriented design models and discuss the gap between agent-oriented design models and agent-oriented platforms.
A significant proportion of the thesis concerns a case study, entitled Aircraft Turnaround Simulator. A complete set of documents for this case study has been developed and the system has been implemented independently be three developers from different countries with different backgrounds. A comparison is made of the three implementations.
The research suggests that the use of agent-oriented concepts and models improves the process of software development. The research extends and improves the models proposed by Sterling and Taveter. The research improves the use of agent-oriented models through proposing new approaches, documenting templates and new tool. It is believed that this research can improve the usability of agent-oriented models and help developers to build high quality systems.