Natural User Interfaces for Conceptual Design Tools

该奖项的研究目标是开发和评估使用自然和流动的人机交互手段创建CAD几何图形的方法。第一个目标是使工程师能够通过草图、手势和交谈来描述概念性设计的结构，就像他们通常相互交流的方式一样。第二个目标是获取所需行为的多模式描述(草图、手势、话语)，并让系统推断产生该行为所需的几何约束。理解多通道设计描述的方法将依赖于黑板体系结构和相互消除歧义。每一种情态都会对当前话语/手势的含义做出假设，而跨情态推理将修剪可能的解释。这将为合并模式提供强大、灵活、通用的基础。将行为和粗略结构的描述转换为工作几何的技术将使对设备几何的约束足以确保设计的运动学行为成为可能。将这些技术与按特征设计方法相结合，将能够从草图生成3D几何模型。这里的建模过程和输出都是新颖的：该过程是对结构和行为的自然描述，并且输出包括从描述中自动推断的行为确保约束。如果成功，这项研究将导致技术超越传统的基于鼠标的界面施加的限制，允许设计者从非正式的设备描述中创建功能几何。这些技术将使设计师能够更直接地专注于BE-Havior，而不是结构，从而产生独特和新颖的产品设计。这项研究的结果将被广泛传播，在科学和技术教育伙伴关系年度会议上向4-8年级的学生展示，并纳入本科工程课程。

The research objective of this award is to develop and evaluate methods to create CAD geometry using natural and fluid means of human-computer interaction. The first goal is to en-able engineers to describe the structure of a conceptual design by sketching, gesturing, and talk-ing, in much the same way they ordinarily communicate with each other. The second goal is to take the multimodal description (sketch, gesture, utterance) of the desired behavior, and have the system infer the geometric constraints necessary to produce that behavior. The approach to un-derstanding multimodal design descriptions will rely on a blackboard architecture and mutual disambiguation. Each of the modalities will contribute hypotheses about the meaning of the cur-rent utterance/gesture/sketch, and cross-modal inference will prune the possible interpretations. This will provide a robust, flexible, general-purpose foundation for merging modalities. The techniques for transforming descriptions of behavior and rough structure into working geometry will make it possible to derive constraints on the geometry of a device sufficient to ensure a de-sired kinematic behavior. Combining these techniques with a design-by-features approach will enable generation of 3D geometric models from sketches. Here both the modeling process and the output are novel: the process is one of natural description of structure and behavior, and the output includes behavior-ensuring constraints that are automatically inferred from that descrip-tion.If successful, this research will result in techniques which go beyond the limits imposed by traditional mouse-based interfaces to allow designers to create functioning geometry from in-formal device descriptions. These techniques will enable designers to focus more directly on be-havior, rather than structure, resulting in unique and novel product designs. Results from this re-search will be widely disseminated, demonstrated to students in grades 4-8 at Science and Tech-nology Education Partnership annual conferences, and incorporated into undergraduate engineer-ing courses.