Cost-Aware Dual Prediction Scheme for Reducing Transmissions at IoT Sensor Nodes

Abstract

This paper develops a method for deciding when to update the prediction model or transmit a set of measurements from the sensor to the fusion centre (FC) to achieve minimal data transmission in a dual prediction scheme (DPS). The proposed method chooses a transmission strategy that results in the lowest expected future transmission cost among a given set of strategies. In a practical IoT setting, statistical information of the measurements might be limited. Hence, without assuming any distribution for the measurements, the proposed method estimates the transmission cost for each strategy through bootstrap data where associated model residuals are resampled using the maximum entropy bootstrap algorithm, which preserves several stochastic properties of the empirical distribution. Numerical results with simulated and real world data shows that the proposed method results in significant reduction in the transmitted data.