How to make CAD tools more useful to designers through re- representation

This paper describes research that explores computational approaches to automatically developing multiple representations of objects, with the aim of overcoming the limitation of fixed representations. The claim is that if CAD tools used a more fluid system of representation they would be more useful to designers by preventing fixation and by providing a means of resolving the ambiguity in internal representations. A method for producing multiple representations has been developed and implemented to demonstrate that simple sketches can be re-represented automically. The method uses a neural network to find features in an arbitrary canonical representation and to restructure a sketch based upon groupings of similar features. The results demonstrate the feasibility of automated computer-based re-representation. This research provides the foundations for support for multiple representations in CAD tools that would make CAD tools more useful to designers. Conference theme: Digital