Reference: Iwasaki, Y. Model Abstraction in Qualitative Reasoning. KSL, November, 1989.
Abstract: Model formation is an important step in engineering problem solving, but relatively little work has been done to automate the process. I formulation a model, choosing an appropriate level of abstraction is crucial since the computational cost of analysis increases rapidly with the size of the model. We propose to develop a theory of model abstraction. We will develop a formalism to define abstraction levels, in a way that allows information about the goal to be used to select an appropriate level. We will test the proposed theory by constructing a program that, given a question about device behavior, can generate models at different abstraction levels, reason about its behavior with appropriate models, and integrate conclusions drawn at multiple levels into one coherent description of the behavior. We expect this research to contribute to understanding of engineering model formulation process and enable automated analysis of complex systems.