CIF: Medium Projects: Event-Triggered Sampling: Application to Decentralized Detection and Estimation

连续时间信号的采样是现代数字技术时代的一个基本过程。本研究研究了一类新的事件触发采样器，其中采样时间由要采样的实际信号以动态方式决定。事件触发的采样方案的丰富性允许有意义的优化问题的数学公式，可以导致具有最优特征的采样策略，并导致潜在的显著性能增益。该项目将开发的理论和技术可能会对重要的应用领域做出重大贡献，例如用于事件监视和监测的传感器网络系统，以及用于基于智能电网的发电和配电的网络控制系统。本文主要研究分散统计信号处理一般领域中的三个主要问题，即序列检测（假设检验和变点检测）、参数估计和信号估计与控制。在这三个问题中，将研究可以优化相应检测或估计性能的最优事件触发机制。此外，还将分析一类特殊的事件触发采样器，并开发相应的数字信号处理（DSP）设计方法。这项研究将对统计信号处理和数字信号处理领域产生变革性的影响。特别是，在检测、估计和控制中使用事件触发采样可以显着提高分散设置中的性能，并在功率和带宽方面大大简化通信要求。

Sampling continuous-time signals constitutes a fundamental process in the modern digital technology era. This research investigates a novel class of event-triggered samplers, in which the sampling times are dictated in a dynamic way by the actual signals to be sampled. The richness of the event-triggered sampling schemes allows for the mathematical formulation of meaningful optimization problems that can lead to sampling strategies with optimal characteristics and resulting in potentially significant performance gains. The theories and techniques to be developed in this project can potentially make significant contributions to important application areas such as sensor network systems for event surveillance and monitoring, and networked control systems for smart-grid-based power generation and distribution.This research focuses on three major problems in the general areas of decentralized statistical signal processing, namely, sequential detection (hypothesis testing and changepoint detection), parameter estimation, and signal estimation and control. In all three problems, optimal event-triggered mechanisms will be investigated that can optimize the corresponding detection or estimation performance. In addition, a special class of event-triggered samplers will be analyzed and the corresponding design methods for digital signal processing (DSP) will be developed .This research will have transformative impacts on the fields of statistical signal processing and digital signal processing. In particular, the use of event-triggered sampling in detection, estimation and control can significantly improve the performance in a decentralized setup and simplify considerably the communication requirements in terms of power and bandwidth.