A Component-Based Multi-Agent System Modeling Methodology

This dissertation defines a new feature called apotential arc that is used to extend traditionalColored Petri Net(CPN) models to support agentbehavior modeling. Potential arcs serve to extend thetraditional CPN model with an explicit support fordistinguishing the representation of potentialconflicts and real conflicts, and thereby avoid theneed for eliminating all potential conflicts in thedesign stage.The research reported in this dissertation involves abroad array of tasks, including modelgeneration algorithms, analysis/simulationtechniques, and experimentation. The maincontributions of the research are : 1) defines a newfeature called a potential arc that is used toextend traditional Colored Petri Net (CPN) models tosupport agent behavior modeling; 2)presents a semi-automated MAS modeling methodology,which allow software engineers tocreate MAS systems based on the extended CPN model;3) presents an intelligent coordinatordesign methodology, resulting in a set ofcoordinators at different levels of intelligence; and4)evaluates the proposed modeling methodology bydeveloping models and performing simulationbasedanalysis.