CAREER: Nonsmooth Control Systems for Societal Networks with Data-Assisted Feedback Loops: Theory and Algorithms

本CAREER提案的总体目标是正式推进混合和非光滑数据辅助控制器的分析和综合，这是一种算法：(a)在闭环系统中纳入数据驱动机制，以实现实时估计、学习和适应；和(b)的特点是混合和非光滑的动力系统能够满足严格的鲁棒性，稳定性和瞬态要求，这些要求在数学上是使用光滑控制技术无法实现的。这项研究是由技术进步推动的，这些技术进步使得驱动、传感、计算和通信设备越来越便携、廉价，并且在网络工程系统中越来越普遍，包括机器人网络、电网和连接的交通系统。在这些应用中，底层（混合）动力系统及其相应的决策问题日益复杂，暴露了传统平滑反馈控制和优化方法的根本局限性。该研究计划将通过为部署在网络物理基础设施上的多智能体系统开发一种基于混合控制理论的数据辅助网络控制新范式来克服这些限制。该项目包含了一个强有力的教育和推广计划，包括通过夏季充实营积极招募和指导来自不同背景的学生，以及为初高中学生提供的课后项目。外联计划还包括在控制和自主系统的广泛领域举办区域讲习班，以及与工业和国家实验室积极合作，为研究提供信息和指导。该研究项目将利用混合控制理论中开发的最新数学工具，进一步整合和发展为三个连贯的研究重点：1)开发鲁棒数据辅助切换和非光滑控制器，能够通过在多个基于反馈的算法之间切换来克服平滑稳定、跟踪和优化的障碍，这些算法同时使用受控系统产生的实时和记录数据；2)多智能体数据辅助控制器的鲁棒性协调，以协同利用其个人能力以获得所需的网络范围性能，同时保留与网络大小和数据需求相关的适当可伸缩性属性；3)多智能体系统策略数据辅助控制器的分析与综合，多智能体系统中单个智能体系统地、动态地操纵其数据以达到欺骗的目的，而不引起闭环系统的不稳定行为。在项目中发现的理论原理，以及提出的算法，将在现实的数值和实验工程系统中进行测试和验证。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。

The overarching goal of this CAREER proposal is to formally advance the analysis and synthesis of hybrid and non-smooth data-assisted controllers, which are algorithms that: (a) incorporate data-driven mechanisms in the closed-loop system for the purpose of real-time estimation, learning, and adaptation; and (b) are characterized by hybrid and non-smooth dynamical systems able to meet stringent robustness, stability and transient demands that are mathematically unachievable using smooth control techniques. The research is motivated by technological advances that have made devices for actuation, sensing, computation, and communication increasingly portable, inexpensive, and prevalent in networked engineering systems, including robotic networks, the power grid, and connected transportation systems. In these applications, the increasing complexity of the underlying (hybrid) dynamical systems and their corresponding decision-making problems have exposed the fundamental limitations of traditional smooth feedback control and optimization methods. The research plan will seek to overcome these limitations by developing a new paradigm of data-assisted network control based on hybrid control theory for multi-agent systems deployed over cyber-physical infrastructure. The project incorporates a strong educational and outreach plan that will involve active recruitment and mentorship of students from diverse backgrounds via summer enrichment camps, as well as after-school programs for middle and high-school students. The outreach plan also includes the development of a regional workshop in the broad areas of control and autonomous systems, as well as active collaborations with industry and national laboratories to inform and guide the research. The research project will leverage recent mathematical tools developed in hybrid control theory, further integrated and developed in three cohesive research thrusts: 1) The development of robust data-assisted switched and non-smooth controllers able to overcome obstructions to smooth stabilization, tracking, and optimization by switching between multiple feedback-based algorithms that use concurrently real-time and recorded data generated by the system under control; 2) The robust coordination of multi-agent data-assisted controllers to synergistically exploit their individual capabilities to obtain a desired network-wide performance, while preserving suitable scalability properties with respect to the size of the network and their data requirements; 3) The analysis and synthesis of strategic data-assisted controllers for multi-agent systems where certain individual agents systematically and dynamically manipulate their data for the purpose of deception without inducing unstable behaviors in the closed-loop system. The theoretical principles uncovered in the project, as well as the proposed algorithms, will be tested and validated in realistic numerical and experimental engineering systems.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.