Trading off SNR and the Number of Observations to Improve the Value of Information in IoT Networks

Abstract

The freshness and usefulness of information play an important role in offering ubiquitous connectivity for time-critical control applications. A concept named value of information (VoI) is proposed based on the field of information theory to quantity the usefulness of data for sensor-assisted Internet of Things (IoT) networks in the presence of transmission noise. In this work, we focus on general Gaussian random process models and study the rate of change of the VoI when generating more data samples and increasing the signal-to-noise ratio (SNR). We further look at Gauss-Markov random process models, and investigate the impact of the number of observations and the SNR on the VoI performance. It is interesting to find that using more data samples is effective to improve the VoI only in the low SNR regime, while it yields zero rate of change of the VoI in the high SNR regime. Moreover, the VoI can be improved by increasing the SNR in both high and low SNR regimes regardless of how many samples are used. We also find a trade-off between the SNR and the number of observations, and scale back SNR to achieve the same VoI improvement by adding one extra observation. The results illustrated in this work can be used in the design of practical real-time IoT networks.