Intelligent User Interfaces for Engineering Design Software
Thomas F. Stahovich
Associate Professor, Mechanical Engineering Department, Carnegie Mellon University
Newell-Simon Hall 1305 (Michael Mauldin Auditorium)
This talk explores two different research projects aimed at developing more powerful user interfaces for engineering design software. The first project is a system that observes a designer using traditional CAD software, and from this infers a set of reusable design rules. This system, called LearnIT II, can be thought of as a trainable design assistant. It uses decision tree learning to learn which design modifications the designer will make at any given point in the design process in order to obtain a design that complies with the design requirements. The program generates a set of rules that it uses to automatically create new designs satisfying new design requirements. In the second project, we are building sketch understanding techniques to enable pen-based user interfaces for CAD software. Our sketch understanding techniques are partially geometric and partially physics based. Geometric analysis of a sketch suggests possible interpretations of the symbols in the sketch. Qualitative physical reasoning is used to identify which of the possible interpretations make sense in the context of the complete sketch.
Professor Stahovich is an Associate Professor in the Carnegie Mellon Mechanical Engineering Department, where he runs the Smart Tools Lab. He is also a member of CALD, the Center for Automated Learning and Discovery. He received a B.S. in Mechanical Engineering from U.C. Berkeley in 1988, an S.M. in Mechanical Engineering from MIT in 1990, and a Ph.D. at the MIT Artificial Intelligence Lab in 1995. His research includes sketch understanding, automated design documentation tools, machine learning for design, design modification management tools, and geometric feature recognition.