Statistical Learning Based Wireless Channel Parameter Estimation for Power Internet of Things

Abstract

With the rapid increase in the number of sensors in Power Internet of Things (PIOT), the data volume surges and efficient communications are needed. Precise estimation of delay spread is critical to improve the communication performance. In this paper, we propose a statistical learning (SL) based root mean square delay spread (RMSDS) estimation method, which consists of offline training and online estimation. During the offline training process, the instantaneous delay spread (IDS) for each data frame is firstly computed based on the known pilots. The acquired IDS is a random variable and its statistical parameter is related to the actual RMSDS, which is preset and known during offline training. Then, the functional relationship between the statistical parameter and the preset RMSDS is acquired through the SL algorithm. The acquired functional relationship can be used in the online estimation process to accurately estimate the RMSDS. Simulation results indicate that the proposed method can accurately estimate the RMSDS and greatly improve the bit error rate performance.