Evaluating agent oriented software testing customized performance metrics

Abstract

The thesis comprising of six chapters, presents the measures for evaluating the characteristics of software agents. It briefly describes the methological way of developing agent based systems. It also introduces a high level agile methodology wherein quality test cases are created, executed and analyzed. Further, action based testing is integrated to the testing process of an agent based system. Testing measures used in procedural programming have also been utilized for evaluating agent s quality. But, they are not appropriate for evaluating agent oriented software. Multi Agent System (MAS) is complex in nature, due to the increase in complexity of developing a multi agent system, the existing metrics are less sufficient for evaluating the quality of an MAS. This is due to the fact that the agent reacts in an unpredictable manner. Existing metrics for measuring MAS quality fails to addresses potential communication, initiative behavior and learnability. In this work, new metrics are proposed to measure the agent quality. The use of action based testing in an agent environment increases the readability of test cases and also plays a vital role in increasing the maintainability of an agent oriented systems. newline