Networked Control Over AWGN Channels: An Information-Theoretic Approach for Industrial IoT

Industrial Internet-of-Things (IoT) has attracted considerable recent attention owing to its potential of empowering smart grids, autonomous driving, and factory automation. Networked or remote control over communication links is expected to play a vital role in industrial IoT in which a plant should be stabilized based on information feedback from sensors to actuators. However, the information-theoretic performance limits of networked control in industrial IoT mostly remain open, especially when the event-triggered communication protocol is adopted. In this paper, we are interested in the minimum bandwidth and power required for stabilizing IoT-oriented remote-control systems exploiting classic and event-triggered feedback over AWGN channels. Specifically, a unified framework is presented by leveraging rate-distortion theory and stochastic differential equation jointly. Our framework not only enables a novel yet simple proof of data rate theorems for classic feedback with Riemann sampling, but also provides a systematic way for attaining the closed-form limits on the communication resources required by IoT with Lebesgue sampling. It is the first time that the resource efficiency gain of event-triggered communications is rigorously shown from an information-theoretic perspective. In particular, only a single bit per feedback can stabilize a plant in IoT with event-triggered communications. Simulations validate our theoretical analysis.