MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractThe agent system specification, the agent system implementation and the agent system verification are three essential issues to build an agent system. Many works have been done for the first two issues in recent years. However, as a result of a lack in formal agent modeling techniques, little effort has been made to address the verification issue, which impedes the agent technique a smooth transition from the research lab to the desk of everyday computer engineers. Motivated by this fact and its significance, it is our objective in this dissertation to establish a systematic method for modeling and analysis of agent systems. An approach to combine the agent belief-desire-intention (BDI) theory and the Petri net transducer (PNT) theory is proposed. The resulting belief-planner-actuator model specifies individual behaviors of agents successfully and bridges the gap among belief, desire and intention of agents seamlessly. A set of agent communication protocols is developed to specify the agent social behavior. Theorems on analyzing the Petri Net underlying those protocols are proposed and proved. Based upon the proposed communication protocols, three agent social behavior models are proposed here: direct coordination, meeting-oriented coordination and blackboard-based coordination. To further exploits the power of the agent communication protocols, a framework to model the mobility of agents is proposed. The framework includes a set of stationary agents (SA) and mobile agents (MA). The agent learning ability is modeled based upon the probabilistic Petri net transducer theory. The individual agent learning behavior is then extended to multiple-agent systems, where the game theory and the agent learning model are combined to achieve a number of agent interaction strategies. These strategies include: self-interested learning, complete cooperative learning, bargaining learning and coordinated learning. Several simulation studies have been conducted to investigate the effectiveness of the proposed agent model. This model is further evaluated through its application to the WAVES (web based audio video educational systems) project and the results have indicated that the proposed method is ideal in analysis of agent systems.
Degree ProgramGraduate College
Systems and Industrial Engineering