Engineering and Information Technology Collected Works - Theses

Permanent URI for this collection

Search Results

Now showing 1 - 1 of 1
  • Item
    Thumbnail Image
    Patterns for agent-oriented software development
    Cheah, Wai Shiang. (University of Melbourne, 2010)
    Agent technology has been used in building various domain specific applications. However, agent technology has not been widely adopted by the software community. Reasons that have been proposed for the lack of adoption include the fragmented diversity of agent oriented methodologies and the lack of maturity of most, if not all, of the methodologies. Alternatively, agent patterns have been introduced by sharing the experience of engineering the agent system and allow the novices to solve the problem in a more systematic and structured way. This thesis is based on the premise that the comprehension of the agent patterns and the adoption of agent technology through patterns can be improved. An investigation into sharing the experience of what does work and does not work in engineering agent oriented software systems through software patterns has been explored by Oluyomi, a previous PhD student in the Intelligent AgentLab at the University of Melbourne. It has been reported that pattern classifications and agent pattern templates are inadequate and inconsistent for sharing the agent development experience to the wider software community. It has been suggested that the inadequacy and inconsistencies of the pattern classification and template structure occurred for several reasons including diverse ways of classifying the patterns, a lack of a clear rationale of the pattern elements and ambiguities of meaning on pattern elements. To enable the adoption of agent technology by the wider software community, we argue that a comprehensiveness and explicitness of agent patterns are needed in order to improve the comprehension of agent technology. Agent oriented modelling and a viewpoint framework have been introduced to develop multi-agent systems in a more unified way. By adopting the viewpoint framework and agent models to explicitly support the comprehension of the agent patterns, an improved agent pattern can be obtained that offers improved adoption of agent technology by software practitioners taking advantage of them. We adopted the viewpoint framework and agent oriented models to enhance Oluyomi�s work on pattern classification and designing agent oriented template structures. Consequently, we present an improved pattern classification and improved agent oriented template structure in this thesis. While various patterns have been introduced and improved agent oriented template structures have been described from one domain to another by Oluyomi, a particular pattern, task knowledge patterns, that is important but paid less attention by Oluyomi, is investigated in this thesis. An investigation to reveal the diverse view in a particular task knowledge pattern, �brokering patterns, which presented across various articles, is reported. We have updated the current agent patterns with the sharing of the task knowledge or problem solving method through the improved agent oriented template structure. The task knowledge patterns presented in this thesis have explicitly described the task knowledge within agent context (e.g. agent oriented models). Altogether, 10 task knowledge patterns are described. We study the reusability of the task knowledge patterns that are described through the improved agent oriented template structure. Assessment of the quality of the task knowledge patterns through proposed design metrics and demonstration of the reusability of the patterns in agent oriented software development are presented. The task knowledge patterns are reusable in the early development phase of the agent system. Appropriate guidelines to aid the reuse of task knowledge patterns in agent development process are proposed. Later, we demonstrate the adoption of the patterns to the rapid prototyping of an agent system for a case study of an adviser finder multi-agent system. Finally, a study on introducing task knowledge patterns to two Master students at Tallinn University of Technology in Estonia is described. Feedbacks from the informal study are promising. The feedbacks have shown that the improved template for task knowledge patterns has increased comprehension among agent developers and the task knowledge patterns are useful for software practitioners.