EAGER: Control Theory for Real-time Privacy-preserving Consensus Control of Engineering Networks

这个早期概念探索性研究（EAGER）项目考虑了如何在工程网络隐私保护的背景下修改已被充分理解的控制架构，以产生新的共识控制策略。通信和计算能力正越来越多地集成到甚至最普通的物品中，形成一个能够独立和合作行动的相互连接的设备网络。这些连接的对象可以协调它们的行动，以提高系统的效率和性能，如智能电网和自动交通控制系统。然而，数据的广泛共享也有泄露私人信息的风险。例如，在网络运营商和调度单位之间交换非常详细的消费者使用概况，可以优化智能电网中的电力预测、发电和分配，但也允许其他人推断出家庭住户的存在、不存在甚至特定活动的知识。在自动驾驶汽车等时间紧迫的工程应用中，经典加密速度太慢，需要经过精心调整的控制器每秒生成数十或数百个命令。在这个项目中开发的洞察力是，交换的信息，甚至控制命令，可能会被现有控制体系结构的创新应用所掩盖。本项目将论证这种方法的可行性，并概述其主要特点。该项目的成果将在不违反机密性的情况下，提供沟通和协调的经济优势，从而促进国家繁荣。研究成果将纳入研究生课程，并纳入由首席研究员指导的本科生研究项目。本项目将探讨分散协调算法在随机时变耦合权重下的稳定性，该算法将用于模糊信息。随着无线通信和网络技术的发展，分散协调算法在网络机器人、传感器网络和智能交通系统中得到了广泛应用。尽管其固有的灵活性和可扩展性使分散协调算法对大型系统具有吸引力，但它们也对隐私保护设计提出了重大挑战。这是因为传统的隐私保护机制依赖于中央数据聚合器或受信任的第三方的帮助，这些机制在分散的实现中被排除在外。该项目建立在调查员最近的研究成果之上，特别是通过不确定的时变控制来模糊信息的隐私保护机制。该方法不需要任何第三方的帮助，并且在灵活性、可伸缩性、准确性和计算开销方面优于现有方法。然而，这种保护隐私的方法导致耦合权随机时变，对分散协调的影响尚不清楚。本项目将通过严格分析随机时变耦合权重下分散协调的收敛条件和速度来解决这一问题。本课题的重点是首先刻画耦合权值时变且在一定区间内随机选取的情况下能够实现非线性一致性的条件，然后分析随机时变耦合权值下分散协调的收敛速度，最后利用多机器人试验台对得到的结果进行系统验证。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。

This EArly-concept Grant for Exploratory Research (EAGER) project considers how well-understood control architectures can be modified to produce novel consensus control strategies in the context of privacy protection for engineering networks. Communication and computing capabilities are increasingly being integrated into even the most commonplace items, forming an interconnected network of devices capable of independent and cooperative action. These connected objects can coordinate their actions to improve the efficiency and performance of systems such as smart power grids and automated traffic control systems. However, the widespread sharing of data also risks divulging private information. For example, the exchange of very detailed consumer usage profiles between network operators and dispatch units enables optimized power forecast, generation, and distribution in a smart grid, but also allows others to infer knowledge about the presence, absence, or even the specific activities of a home's occupants. Classical encryption is too slow to be deployed in time-critical engineering applications such as self-driving cars, where carefully tuned controllers generate tens or hundreds of commands per second. The insight developed in this project is that exchanged information, and even control commands, can be obscured by the innovative application of existing control architectures. This project will demonstrate the feasibility of this approach and outline its major features. The results of this project will advance national prosperity by offering the economic advantages of communication and coordination, without violating confidentiality. The research results will be integrated into the graduate curriculum, and into undergraduate research projects supervised by the Principal Investigator.This project will explore the stabilization of decentralized coordination algorithms under random time-varying coupling weights which will be used for obscuring information. With recent advances in wireless communications and networking, decentralized coordination algorithms have become widespread in networked robots, sensor networks, and intelligent transportation systems. Although their inherent flexibility and scalability make decentralized coordination algorithms appealing for large-scale systems, they also pose significant challenges to privacy protection design. This is because conventional privacy-preserving mechanisms rely on the assistance of a central data aggregator or a trusted third party -- mechanisms that are ruled out in a decentralized implementation. This project builds upon recent results by the Investigator, specifically a privacy-preserving mechanism which obscures information through uncertain time-varying controls. The approach does not need the assistance of any third party and is superior to existing approaches in terms of flexibility, scalability, accuracy, and computation overhead. However, this privacy-preserving approach leads to random time-varying coupling weights, whose influence on decentralized coordination is unclear. This project will address this problem through rigorous analysis of the convergence conditions and speed of decentralized coordination under random time-varying coupling weights. The main thrusts of the project are first to characterize the condition under which nonlinear consensus can be achieved when coupling weights are time-varying and randomly chosen from a certain interval, next to analyze the convergence speed of decentralized coordination under random time-varying coupling weights, and finally to systematically verify obtained results using a multi-robot test bed.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.