| In Proceedings of IJCAI-01 Resolving Ambiguities to Create a Natural Computer-Based Sketching Environment (2008) | |||||||||||||||
Abstract | |||||||||||||||
| Current computer-based design tools for mechanical engineers are not tailored to the early stages of design. Most designs start as pencil and paper sketches, and are entered into CAD systems only when nearly complete. Our goal is to create a kind of “magic paper ” capable of bridging the gap between these two stages. We want to create a computer-based sketching environment that feels as natural as sketching on paper, but unlike paper, understands a mechanical engineer’s sketch as it is drawn. One important step toward realizing this goal is resolving ambiguities in the sketch — determining, for example, whether a circle is intended to indicate a wheel or a pin joint—and doing this as the user draws, so that it doesn’t interfere with the design process. We present a method and an implemented program that does this for freehand sketches of simple 2-D mechanical devices. 1 Sketching Conceptual Designs Engineers typically make several drawings in the course of a design, ranging from informal sketches to the formal manufacturing drawings created with drafting tools. Drawing is far more than an artifact of the design process; it has been shown to be essential at all stages of the design process [Ullman et al., 1990]. Yet almost all early drawings are still done using pencil and paper. Only after a design is relatively stable do engineers take the time to use computer aided design or drafting tools, typically because existing tools are too difficult to use for the meager payoff they provide at this early stage. Our aim is to allow designers to sketch just as they would on paper, e.g., without specifying in advance what component they are drawing, yet have the system understand what has been sketched. We want to have the input be as unconstrained as possible, in order to make interaction easy and natural; our route to accomplishing this is to build a sufficiently powerful sketch recognizer. It is not yet obvious that a freehand sketching interface will be more effective in real use than a carefully designed menubased system. In order to do the comparison experiments, however, we must first build powerful sketch-based systems. | |||||||||||||||
Publication details | |||||||||||||||
| |||||||||||||||