RII Track 2 FEC: Building Research Infrastructure and Workforce in Edge Artificial Intelligence

目前，使用人工智能(AI)需要访问互联网和非常大且复杂的远程计算机，以做出决策和预测。这会导致长时间的延迟以及隐私和安全方面的问题。人工智能领域的最新技术，也就是所谓的“边缘人工智能”，通过直接在相机、智能手机和可穿戴设备上收集和分析数据，避免了这些问题。然而，Edge AI仍处于初级阶段，有几个重要的技术问题需要解决。这个以研究基础设施改善Track-2为重点的EPSCoR协作(RII Track-2 FEC)奖是阿拉巴马州、阿肯色州和北达科他州的六所大学(包括两所为少数群体服务的机构)与几个私营部门合作伙伴的合作。作为对该项目新技术的测试，该项目团队将制造一种智能可穿戴设备，通过监测患者自己的呼吸来预测糖尿病的发病，而不需要医生解读结果。它将为高级大学生提供研究培训机会，还将在课程中培训高中教师，让他们自己的学生了解Edge AI的原理，为未来的世界播种这些基本概念的未来美国劳动力。RII Track-2 FEC奖的目标是发展Edge AI的综合研究基础设施和劳动力。该团队将开发的基本贡献和技术创新包括：(I)用于EDGE平台的轻量级AI支持的推理和机器学习算法；(Ii)新的专用集成电路(ASIC)设计方法，使AI ASIC具有超低功耗、可重构和开发周期短；(Iii)用于Edge AI的传感器设备平台，该平台基于新型功能化纳米级传感材料和纳米3D打印技术；以及(Iv)边缘AI设备平台，利用先前的进展来满足不同用例的要求。基于开发的基础设施，针对糖尿病护理的用例，该团队将设计、原型和测试一种用于个性化糖尿病管理的低成本智能可穿戴设备。与现有技术相比，开发的可穿戴糖尿病设备将显著降低成本并提高能效。领先的机构是南阿拉巴马大学；合作机构是北达科他州州立大学、阿肯色大学、北达科他州大学、阿拉巴马州农工大学和Nueta Hidatsa Sahnish学院。该团队将与多个行业合作伙伴密切合作，在不同的使用案例中采用和调整开发的Edge AI基础设施。该项目的研究成果将加速Edge AI的发展，并将增加美国在AI方面的竞争力。此外，该项目将整合研究、教育和劳动力发展，以便在多个层面提供有效的培训。该项目将开发从高中到本科生、研究生、博士后培训、初级教师和行业从业者的教育到劳动力的管道。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Using Artificial Intelligence (AI) currently requires access to the internet and very large and complex remote computers for making decisions and predictions. This causes long delays and privacy and security concerns. The latest techniques in AI, known as “Edge AI”, avoid these problems by collecting and analyzing data directly on cameras, smart phones, and wearable devices. However, Edge AI is still in its infancy and there are several important technical problems that need to be solved. This Research Infrastructure Improvement Track-2 Focused EPSCoR Collaborations (RII Track-2 FEC) award is a collaboration between six universities (including two minority-serving institutions) and several private-sector partners in Alabama, Arkansas, and North Dakota. As a test of the project's new technology, the project team will build a smart wearable device to predict the onset of diabetes by monitoring a patient's own breath without the need for a doctor to interpret the results. It will provide research training opportunities for advanced college students and will also train high-school teachers in lessons to educate their own students in the principles of Edge AI to seed the future US workforce in these essential concepts for tomorrow’s world.The goal of this RII Track-2 FEC award is to develop integrated research infrastructure and workforce in Edge AI. Fundamental contributions and technical innovations to be developed by the team include: (i) light-weight AI-empowered reasoning and machine learning algorithms for edge platforms; (ii) a new Application-Specific Integrated Circuits (ASIC) design methodology to enable AI ASICs with ultra-low power, reconfigurability, and short development cycles; (iii) a sensor device platform for Edge AI based on novel functionalized nano-scaled sensing materials with nano-3D printing techniques; and (iv) an Edge AI device platform exploiting the previous advances to meet the requirements of different use cases. Based on the developed infrastructure, targeting the use case of diabetes care, the team will design, prototype, and test a low-cost smart wearable device for personalized diabetes management. The developed wearable diabetes device will enable significant cost reduction and high power efficiency compared to existing techniques. The leading institution is the University of South Alabama; the collaborating institutions are North Dakota State University, the University of Arkansas, the University of North Dakota, Alabama A&M University, and Nueta Hidatsa Sahnish College. The team will work closely with multiple industry partners to adopt and adapt the developed Edge AI infrastructure in different use cases. Research outcomes of this project will accelerate the development of Edge AI and will increase the competitiveness of the United States in AI. Also, this project will integrate research, education, and workforce development in order to provide effective training at multiple levels. The project will develop an Education-to-Workforce Pipeline from high school to undergraduate, graduate, Post-Doctoral training, junior faculty, and industry practitioners.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.