Machine Learning Enabled Sleep Time Estimation (MLE-STE) Architecture for Indoor Positioning in Energy-Efficient Mobile Internet of Things

Abstract

Indoor positioning and tracking systems require not only accurate position estimates of mobile IoT devices but also energy efficiency in order to maximize the battery life of the mobile IoT device. The contribution of this paper is the design of a machine learning enabled indoor positioning and tracking system in which artificial intelligence is utilized for the estimation of the duration for which a mobile IoT device needs to sleep in order to conserve energy. Our Machine Learning Enabled Sleep Time Estimation (MLE-STE) architecture is comprised of the following stages: First it forms the forecast of the near-future trajectory of the mobile IoT device. Second based on these forecasts it determines the optimal sleep duration subject to the constraint of a maximum tolerable forecasting error. We demonstrate that our MLE-STE architecture outperforms both of the following state-of-the-art algorithms in this area: Positioning Interval based on Displacement (PID) and Dynamic Positioning Interval Based on Reciprocal Forecasting Error (DPI-RFE). This work represents a significant advance in the development of accurate indoor positioning and tracking algorithms that target the energy efficiency of mobile IoT devices.