CAREER: Uncertainty-Aware and Data-Driven Methods for Electronic and Photonic Design Automation

职业：电子和光子设计自动化的不确定性感知和数据驱动方法

• 批准号: 1846476
• 负责人: Zheng Zhang
• 金额: $ 51.03万
• 依托单位: University of California-Santa Barbara
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-04-01 至 2025-03-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1846476&HistoricalAwards=false
• 关键词: CAREER; Uncertainty; Aware; Data; Driven; CAREER; Uncertainty; Aware; Data; Driven

Designing complex engineering systems such as self-driving cars, electronic and photonic integrated circuits, requires design automation software to complete many challenging tasks that are impossible or too time-consuming if done manually. In practice, almost all engineering designs are subject to such unavoidable uncertainties as noise, fabrication process variations and insufficient knowledge about external environments. These uncertainties often cause performance degradations, system failures and sometimes fatal accidents. However, existing design automation software requires massive data samples from time-consuming computer simulations or non-trivial measurement when uncertainties are involved. This project uses electronic and photonic integrated circuits as driving examples, and will develop novel design automation algorithms to improve the performance and reliability under various uncertainties. The education components of this project include creating two graduate courses of uncertainty and data analysis, training future workforce through undergraduate and graduate research. The outreach education and training through the awardee institution and through academic conferences will enable technology and knowledge transfer to a broad community. Although this project targets on applications in electronics and photonics, the developed algorithms and theory will be applicable to many other domains such as autonomous driving, renewable energy systems, and medical imaging. Since uncertainty-aware photonic design automation is still at its early stage, this project will enable a new field of important research. The resulting algorithms and tools will support the foreseeable large-scale photonic integration which will boost the performance of future computing and communication systems. The technical goal of this project is to develop novel uncertainty-aware electronic and photonic design automation algorithms that require only a small data set and a very low computational cost in the design flow. This project will span three research topics: uncertainty-aware simulation, optimization and data-driven variation modeling. Firstly, novel algorithms will be developed to address several long-standing challenges in the forward uncertainty quantification of electronic and photonic circuits, such as the coupled impact of fundamentally different types of uncertainties and long-term probabilistic simulation errors. Secondly, leveraging the developed forward uncertainty simulator, this project will further develop ultra-fast optimization tools to improve the yield of electronic and photonic circuits. The main focus will be investigating large-scale "non-sampling" stochastic optimization algorithms. The developed algorithms will enable rigorous yield optimization with "small" simulation data sets and thus significantly reduce the software runtime on a computer. Finally, rigorous statistical estimation algorithms will be developed to calibrate critical device model parameters and to extract statistical variability distributions based on limited and noisy indirect circuit-level measurement data. The designed algorithms and prototyping software will be validated by practical design cases.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.

设计复杂的工程系统，例如自动驾驶汽车，电子和光子集成电路，需要设计自动化软件来完成许多不可能或过时的任务，如果手动完成。实际上，几乎所有工程设计都受到不可避免的不确定性，例如噪声，制造过程变化以及对外部环境的知识不足。这些不确定性通常会导致性能降解，系统失败以及有时是致命的事故。但是，现有的设计自动化软件需要涉及不确定性时耗时的计算机模拟或非平凡测量来大规模的数据示例。该项目使用电子和光子集成电路作为驾驶示例，并将开发新颖的设计自动化算法，以提高各种不确定性下的性能和可靠性。该项目的教育组成部分包括创建两个不确定性和数据分析的研究生课程，通过本科和研究生研究培训未来的劳动力。通过获奖者机构和学术会议的推广教育和培训将使技术和知识转移到广泛的社区。尽管该项目针对电子和光子学的应用，但开发的算法和理论将适用于许多其他领域，例如自动驾驶，可再生能源系统和医学成像。由于不确定性感知的光子设计自动化仍处于早期阶段，因此该项目将为重要的重要研究提供新的领域。由此产生的算法和工具将支持可预见的大规模光子集成，这将提高未来计算和通信系统的性能。 该项目的技术目标是开发新颖的不确定性感知的电子和光子设计自动化算法，该算法仅需要一个小的数据集和设计流中的计算成本非常低。 该项目将涵盖三个研究主题：不确定性感知模拟，优化和数据驱动的变化建模。首先，将开发新的算法，以应对电子和光子电路的正向不确定性量化的几个长期挑战，例如根本不同类型的不确定性和长期概率模拟误差的耦合影响。其次，该项目利用开发的正向不确定性模拟器，将进一步开发超快速优化工具，以提高电子和光子电路的产量。主要重点是研究大规模的“非采样”随机优化算法。开发的算法将通过“小”仿真数据集实现严格的收益率优化，从而大大降低了计算机上的软件运行时。最后，将开发严格的统计估计算法来校准关键设备模型参数，并根据有限和嘈杂的间接电路级测量数据提取统计变异性分布。设计的算法和原型软件将通过实用设计案例来验证。该奖项反映了NSF的法定任务，并且使用基金会的知识分子优点和更广泛的影响评估标准，被认为值得通过评估来获得支持。