Nonstochastic information flows in networked dynamical systems

点击翻译按钮获取中文摘要

Feedback control is a crucial element of manufacturing, vehicular and energy systems, and is needed to guarantee hard performance bounds in safety- and mission-critical environments. When these control systems are implemented over communication networks, the amount of information flowing through them becomes a critical determinant of performance. However, the nonprobabilistic control objectives make standard information theory inapplicable. This project aims to develop a novel, nonstochastic theory of information in order to analyse and design networked dynamical systems that obey worst-case performance limits. This will yield robust, probability-free algorithms for distributed control, filtering and causality inference.