BRIGE: Minimum-Energy Bio-Inspired Analogic Computing Devices with Stochastic Switching Transistors under Ultra-Low VDD

BRIGE：超低 VDD 下具有随机开关晶体管的最低能耗仿生模拟计算设备

• 批准号: 1342225
• 负责人: Mingjie Lin
• 金额: $ 14.84万
• 依托单位: The University of Central Florida Board of Trustees
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2013
• 资助国家: 美国
• 起止时间: 2013-09-15 至 2016-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1342225&HistoricalAwards=false
• 关键词: BRIGE; Minimum; Energy; Bio; Inspired

ECCS-1342225Lin, MingjieUniversity of Central FloridaBRIGE: Minimum-Energy Bio-Inspired AnaLogic Computing Devices with Stochastic Switching Transistors under Ultra-Low VDDABSTRACTIntellectual Merit: This BRIGE project aims at achieving robust ultra-low power computation by exploiting the stochastic switching behavior observed within a CMOS digital circuit that is driven by an ultra-low supply voltage approaching the digital switching limit. This objective is motivated by two observations. First, probabilistic inference and stochastic learning are fundamental in sensor data processing. Second, emerging devices will exhibit sophisticated physical property that may natively compute probabilistic algorithms. This proposed effort will first model and analyze the stochastic switching behavior in minimum-energy CMOS transistors under ultra-low VDD (¡Ö 50mV) both analytically and experimentally. Subsequently, it will develop a field-theoretic methodology to optimize a large-scale logic circuit built with such stochastic switching devices in order to improve its robustness. Finally, it will exploit the stochastic switching behavior natively to design and implement AnaLogic circuits (between analog and logic circuits) that emulate a robust self-motion algorithm inspired by fly eye based on optical flow extraction.Broader Impacts: Leveraging the physics of field-effect devices to perform computational tasks, this proposed research could potentially inspire a totally unconventional design paradigm for emerging nanoscale device technology with severe device variability and switching uncertainty. Furthermore, the proposed field-theoretic approach offers a rich mathematical structure, therefore can broaden the current digital circuit design theory. Finally, the proposed methodology can enable more accurate understanding of existing logic circuit design methods, especially on their limitations when directly applied to future device technologies driven by ultra-low VDD. Besides disseminating its research findings through new curricula and hardware-based stochastic logic circuit emulations, this project will approach the challenge of broadening the engineering participation from underrepresented minority groups in both bottom-up (public STEM education) and top-down (PhD students recruiting) directions. The PI will create mentoring and outreach programs specifically designed to attract female, African-American, Latino, and first-generation college students to join his group, thus preparing a new diverse work force for the computing industry. Additionally, the Orlando Science Center will be used as the main platform to stimulate public interests in STEM education of computing. The success of this educational effort, through innovative exhibits and engaging mini-lectures, will be judged by the PhD enrollment of computer engineering from underrepresented groups at UCF and the size of public audience to its collaborative exhibit efforts with the Orlando Science Center.

ECCS-1342225 Lin，MingjieUniversity of Central FloridaBRIGE：具有超低VDD下随机开关晶体管的最小能量生物启发AnaLogic计算设备摘要智力优势：该BRIGE项目旨在通过利用在CMOS数字电路中观察到的随机开关行为来实现鲁棒的超低功耗计算，该CMOS数字电路由接近数字开关极限的超低电源电压驱动。这一目标是由两个观察所促成的。首先，概率推理和随机学习是传感器数据处理的基础。第二，新兴设备将展现出复杂的物理特性，可以原生地计算概率算法。本文首先对最小能量CMOS晶体管在超低VDD（~ 50 mV）下的随机开关行为进行了理论和实验分析。随后，它将开发一个场论的方法来优化一个大规模的逻辑电路，建立这样的随机开关器件，以提高其鲁棒性。最后，本论文将利用随机开关行为来设计和实现AnaLogic电路（介于模拟电路和逻辑电路之间），仿真基于光流提取的蝇眼启发的强大自运动算法。利用场效应器件的物理特性来执行计算任务，这项研究可能会激发出一种全新的非传统设计范式，用于具有严重器件可变性和开关的新兴纳米器件技术。不确定性此外，所提出的场论方法提供了一个丰富的数学结构，因此可以拓宽目前的数字电路设计理论。最后，所提出的方法可以更准确地理解现有的逻辑电路设计方法，特别是在直接应用于未来的器件技术时，由超低VDD驱动的限制。除了通过新课程和基于硬件的随机逻辑电路仿真传播其研究成果外，该项目还将在自下而上（公共STEM教育）和自上而下（博士生招聘）方向上扩大代表性不足的少数群体的工程参与。PI将创建指导和推广计划，专门用于吸引女性，非洲裔美国人，拉丁美洲人和第一代大学生加入他的团队，从而为计算机行业准备新的多元化劳动力。此外，奥兰多科学中心将被用作激发公众对STEM计算教育兴趣的主要平台。这一教育工作的成功，通过创新的展览和吸引人的小型讲座，将通过UCF代表性不足的群体的计算机工程博士入学人数和公众观众的规模来判断其与奥兰多科学中心的合作展览工作。