Intelligent agent characterization and uncertainty management with fuzzy set theory: a tool to support early supplier integration

This research performed an analysis of the existing categories for agent classification and characterization. The methodology developed fuzzy set theory based intelligent agents designed to promote integration of suppliers into the early stages of product development. The first stage in the project personified the anticipated agents by giving them roles that correlated with humanistic tendencies or professions. The project produced a multidimensional profile for agents by providing an analysis of three personality traits: aggressiveness, attractiveness, and vulnerability. The value associated with each of these traits is obtained through different functions which consider the degree of anticipated interaction between the agent, other agents, and the external environment. Further, the types of uncertainty associated with decision making are identified and assigned. The personality traits are used to describe the expected behavior of the agent and are assigned at the initiation o f a given task. Conversely, the measures of uncertainty are invoked at the completion of the task to provide a series of metrics related to the quality of information or task performed. The developed generic fuzzy intelligent agent model was programmed into two of the types of agents using JAVA. The agents successfully communicated with each other while permitting examination of the uncertainty. The result of the methodology is the creation of a frame that can be used to develop multiple fuzzy agents. Further, this research presents one of the first instantiations of a fuzzy agent. Finally, a hypothetical application of these agents is proposed to support architectural innovation in product development. This example promotes early supplier integration into the product development phase through the use of electronic communication and management of the interactions through intelligent agents. The details of computer programming are left for presentation in a succeeding article.