LEAP-HI: Hybrid Intelligence for Design: Bridging Human and Machine Intelligences for Collaborative Design of Engineering Systems and Infrastructure

大型工程系统和基础设施，如建筑物和飞机，对共同设计它们的跨学科团队构成了挑战。今天的建筑师、工程师和设计师使用计算工具来帮助产生想法，模拟和分析大型系统行为的特定方面。然而，这些系统越庞大、越复杂，就越有必要将设计的所有方面集成到一个统一的设计工具中。这项美国繁荣、健康和基础设施（LEAP-HI）领先工程奖支持混合方法的基础研究，该方法通过使用机器学习、凸优化、图像处理和人工智能等新兴计算技术，使计算机和人类能够在系统级设计中进行协作。这项工作有可能通过加强设计创造力和探索，从而提高产品性能和安全性，从而有利于美国的工业创新和竞争力。该项目整合了产品设计、建筑和结构系统以及机械工程等学科的专业知识，并包括扩大代表性不足群体的参与和改善工程和建筑教育的活动。将研究一种新的基于机器学习的创新设计空间探索模型，使建筑师、设计师和工程师能够创造性地、有效地合作。该方法将设计决策与实时性能预测联系起来，而不需要严格地预先指定决策参数，将模拟设计方法的自由与计算能力相结合。对于大型、复杂的工程系统，研究工作将集中在利用人工智能和机器学习方法来帮助系统架构师提取设计师？S意图，并将该意图公式化为一组分层结构的功能需求，这些需求可以作为系统的子系统进行维护。深度学习工具和算法将用于表示和嵌入作为方法基础的设计功能。工程团队将对凸工程中计算工具的使用进行研究，以了解混合智能设计方法如何影响复杂系统设计中的设计意图概念。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Large engineered systems and infrastructure, such as buildings and airplanes, pose a challenge to the interdisciplinary teams who work together to design them. Today's architects, engineers and designers use computational tools to help generate ideas and simulate and analyze specific aspects of the behaviors of large systems. However, the larger and more complex those systems become, the more it becomes necessary to integrate all aspects of design into a unified design tool. This Leading Engineering for America's Prosperity, Health, and Infrastructure (LEAP-HI) award supports fundamental research on a hybrid approach that enables computers and humans to collaborate in system-level design through the use of emerging computational techniques from machine learning, convex optimization, image processing and artificial intelligence. The work has the potential to benefit US industrial innovation and competitiveness by enhancing design creativity and exploration, thereby improving product performance and safety. The project integrates expertise in the disciplines of product design, architectural and structural systems, and mechanical engineering and includes activities to broaden the participation of underrepresented groups and improve engineering and architecture education. A new, machine-learning-based model for creative design space exploration that enables architects, designers, and engineers to collaborate creatively and effectively will be researched. The method links design decisions to real-time performance predictions without the need for rigidly pre-specifying decision parameters, combining the freedom of analog design methods with the power of computation. For large, complex, engineered systems, the research effort will focus on leveraging artificial intelligence and machine learning approaches to assist system architects in extracting a designer?s intent and formulating that intent as a set of hierarchically structured functional requirements that can be maintained as subsystems of systems. Deep learning tools and algorithms will be used to represent and embed the design functions that underlie the methods. Studies of the use of computational tools for convex engineering by engineering teams will be conducted to understand how hybrid intelligent approaches to design impact notions of design intent in complex system design.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.