CAREER: Learning, Estimation, and Control of Networked Epidemic Processes

目前控制疾病在人群中传播的方法采用的技术仅依赖于数学模型或仅依赖于数据，这两种极端之间没有明确的联系。拟议的研究将建立一套基本理论、工具和算法，通过利用多个实时数据流来建模、学习和控制实时流行病传播过程，同时评估基于模型和数据驱动方法之间的权衡。拟议的研究包括三个重点。第一个重点将包括设计一套新模型，以描述不同环境下的流行病传播特征，为连接和比较不同分辨率的模型提供工具。下一个重点将包括算法设计和开发，旨在通过利用多个实时数据流选择合适的模型和识别模型特征。最后一个推力将在数据驱动的预测控制框架内整合前两个推力的多分辨率模型，研究基于模型的方法与数据驱动的方法。这项研究将有助于为政治家、公共卫生官员、行政人员和商业领袖等决策者提供见解，以更好地减轻未来的疾病爆发。拟议的研究将开发一类多分辨率模型，使非线性控制设计能够跨越和适应基于模型和数据驱动的频谱，并专注于网络流行病过程的应用。该项目将确定在存在损坏数据集的情况下可实现性能的基本界限，并提供性能保证的理论和算法。所提出的研究将导致对影响具有多个在线数据流的网络系统的建模、学习和控制的基本因素的更好理解；建立系统的模型选择程序，从不确定和有偏差的数据中估计参数和结构；并针对不同模型分辨率开发可实现的实时控制策略。该研究计划将通过软件平台整合到教育中，作为教育工具，为学生提供数据、指数增长、动态建模和控制方面的实践经验。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。

Current approaches for controlling diseases spreading through populations employ techniques that rely solely on mathematical models or only depend on data, with no clear connection between these two extremes. The proposed research will establish a set of fundamental theories, tools, and algorithms to model, learn, and control real-time epidemic spreading processes by leveraging multiple live data streams while evaluating the trade-off between model-based and data-driven approaches. The proposed research consists of three thrusts. The first thrust will encompass the design of a set of novel models to characterize epidemic spreading under different settings, providing tools for connecting and comparing models at different resolutions. The next thrust will include algorithm design and development aimed at selecting the appropriate models and identification of model features by leveraging multiple live data streams. The last thrust will incorporate the multi-resolution models from the first two thrusts within a data-driven predictive control framework, studying model-based vs. data-driven approaches. This research will help provide insights for decision makers, such as politicians, public health officials, administrators, and business leaders, to better mitigate future disease outbreaks. The proposed research will develop a class of multi-resolution models that enable nonlinear control design that spans and adapts along the model-based vs. data-driven spectrum and is focused on the application of networked epidemic processes. The project will identify fundamental bounds on achievable performance in the presence of corrupted data sets and provide theories and algorithms with performance guarantees. The proposed research will lead to a greater understanding of the fundamental factors that affect the modeling, learning, and control of networked systems with multiple online data streams; establishing systematic procedures for model selection, estimating parameters and structure from uncertain and biased data; and developing realizable real-time control strategies for and across different model resolutions. The research plan will be integrated into education through a software platform to serve as an educational tool, giving students hands-on experience with data, exponential growth, dynamic modeling, and control.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.