CAREER: From the Cloud to the Crowd: An Enabling Solution for the Internet of Federated Things

这个教师早期职业发展计划（CAREER）奖将帮助美国保持制造业的领先地位，利用物联网，大量的计算和数据收集设备，如机床控制器，手机和远程传感器通过互联网连接，以实现改进的分析和决策能力，被称为物联网（IoFT）。这些互连设备的快速增长的计算能力和数据存储能力允许开发分散的、保护隐私的分析工具，这些工具可以通过并行化来增强公平性、隐私性、成本效益并减少计算负担，这与在云中的集中式数据中心收集和处理数据的流行模式形成了鲜明对比。随附的教育计划支持社区参与新兴技术，帮助未来的工程师准备与物联网技术互动和开发物联网技术。这是通过一个面向美国公共和中学生的外展计划来实现的，这些社区在STEM领域和大学校园中代表性不足。这项研究将为IoFT构建一个数据驱动的分散分析框架，能够在单个设备层面提供定制的决策和控制。预测框架允许设备保留其本地数据驱动模型，同时利用来自类似但不相同的设备的结果和数据。预测框架的一个关键特征是它能够考虑数据异质性，因为设备通常在不同的操作和环境条件下运行。贝叶斯理论和分布式推理，然后采用分散的不确定性量化的预测模型在个别设备的水平。然后在优化框架中使用预测，以通过优化可控系统输入来实现单个设备级的目标输出。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

This Faculty Early Career Development Program (CAREER) award will help the US to maintain a leading position in manufacturing by leveraging the Internet of Things, the multitude of computing and data collection devices such as machine tool controllers, cell phones and remote sensors connected by the Internet, to achieve improved analytics and decision-making capability, referred to as the Internet of Federated Things (IoFT). The rapidly increasing computational power and data storage capabilities of these interconnected devices allow the development of decentralized, privacy-preserving analytics tools that can enhance fairness, privacy, cost-effectiveness and reduced computational burden through parallelization, standing in contrast to the prevalent paradigm of collecting and processing data in centralized data centers in the cloud. The accompanying educational plan supports community-level involvement in emerging technologies, helping to prepare future engineers to interact with and develop IoFT technologies. This is achieved through an outreach program oriented towards the American public and middle-school students in the greater-Detroit communities underrepresented in STEM fields and on college campuses.This research will build a data-driven, decentralized analytics framework for the IoFT capable of providing customized decision-making and control at the individual device level. A predictive framework allows devices to retain their local data-driven models while leveraging results and data from similar, though non-identical, devices. A key feature of the predictive framework is its ability to account for data heterogeneity, as devices are often operated under different operating and environmental conditions. Bayesian theory and distributed inference are then adopted to enable decentralized uncertainty quantification for the predictive models at the individual device level. The predictions are then used in an optimization framework to realize target output at the individual device level by optimizing controllable system input. The techniques developed will be evaluated using an IoFT prototype for distributed 3D printing, and will provide public data for other researchers to further investigate IoFT within manufacturing.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.