posted on 1992-01-01, 00:00authored byAjay K. Modi, Carnegie Mellon University.Engineering Design Research Center.
Abstract: "Process design is a complex activity that requires the design agent to possess characteristics currently missing in most artificial design systems. Soar is an integrated software architecture for knowledge-based problem solving, learning and interaction with external environments. We report on two systems developed using the architecture. The first, CPD-Soar, designs distillation sequences, while the second, Interval-Soar, performs simple arithmetic tasks. The structure and behaviour of both systems is described and discussed in depth. In so doing, it is depicted how design and design-related tasks can be cast within the Soar framework, hence demonstrating the functioning and strengths of the architecture's problem-solving mechanisms.The systems provide evidence for an [sic] hypothesis about learning; namely, that the knowledge learned by an agent while performing a taskis strongly dependent upon the models the agent brings to the problem-solving experience. Specifically, it is shown that the richer the model an agent has of its evaluation functions, the more general the knowledge it learns. In relation to this, we describe how the functionality of Interval-Soar can be used to further improve upon the performance of CPD-Soar. We also discuss a number of other issues that arose when building the two systems."