Measuring the reactivity of intelligent agent programs

The booming of intelligent agent technology over past few decades brings a surging number of agent applications in various areas. There also have a large number of designs as well as programming languages been introduced in the literature in the agent oriented area. However, very little work has been dedicated to define quality measures for the development of an agent-based system. Previous efforts mostly focus on adopting classical measures such as using coupling (degree of program dependency) and cohesion (degree of function relationship in single module) to measure the quality of agent programs. I maintain that its time to work on a set of software quality measures that are specific to the distinct characteristics of agent-based systems. In this thesis, two methods are purposed to measure the reactivity of agent systems, which provide indications on how agent systems respond to changes in their environment in a timely fashion. A prototype tool is developed integrated with Jason, a well-known agent-oriented programming platform, to calculate reactivity of each agent in agent system based on their plan libraries, and conducted several experiments to demonstrate the reliability of reactivity measure. In addition, an agent behavioural profile is introduced which is an overview of relationships of actions in agent plan library. Based on agent behavioural profile, definitions of agent behavioral profile identity, entailment as well as preservation were proposed, which ensure original agent's behaviours could be preserved while performing reactivity enhancement.