CIF: Small: Limitations of communication and control and over noisy feedback links

通信网络的广泛可用性促进了节点之间的相互作用，这些节点通过在可用的通信链路上交换信息而不断地相互影响。从社会网络到生物网络，这种动态的、互动的交流方式在许多网络系统中都很常见。它的因果性质不同于经典通信系统的批量编码和处理，传统的信息论结果不能很好地处理它。这些环境更适合于控制和调节过程的运作和理解，更一般地说，更适合于学习和适应过程。提出的研究的主要目的是表征在噪声通信链路上相互作用的动态系统的基本限制。基于一种新的信息论量，我们开始分析点对点交互通信中编码器获取接收端噪声反馈的问题，该信息论量捕获了反馈信道有噪声时前向信道中的实际信息流。由此，我们建议推导出公共信道的噪声反馈容量的可计算界限，开发出对反馈链路中的噪声具有鲁棒性的控制启发的实用编码方案，并描述使用噪声反馈的成本和收益。此外，我们建议将理论扩展到其他由远程制导和虚拟现实应用驱动的新的噪声反馈结构。由此产生的动力系统同时是通信系统和控制系统。我们的目的是研究这些相互作用的动力系统的性质和局限性，以一种综合的方式，通过推广相关的结果在网络控制系统。

The widespread availability of communication networks fosters interactions among nodes that constantly affect each other by exchanging information over the available communication links. This dynamic, interactive, aspect of communication is common to many networked systems, from social networks to biological networks. Its causal nature differs from the batch coding and processing of classical communication systems and is not handled well by the traditional information theory results. These settings are more appropriate for the functioning and the understanding of control and regulation processes and, more generally, processes of learning and adaptation.The main objective of the proposed research is to characterize the fundamental limitations of interacting dynamical systems over noisy communication links. We start analyzing the problem of interactive point-to-point communication where the encoder has access to noisy feedback from the receiver based on a new information theoretic quantity that captures the actual information flow in the forward channel when the feedback channel is noisy. From it, we propose to derive computable bounds on the noisy feedback capacity for common channels, to develop control-inspired practical coding schemes that are robust to noise in the feedback link, and to characterize costs and benefits of using noisy feedback. Further, we propose extensions of the theory to other new noisy feedback structures motivated by remote guidance and virtual reality applications. The resulting dynamical systems are both communication and control systems at the same time. We aim to study properties and limitations of these interacting dynamical systems, in an integrated way, by extending the relevant results in networked control systems.