Offset Estimation Based on ARIMA-LSTM for Time Synchronization in Single Twisted Pair Ethernet

Abstract

Single twisted pair Ethernet becomes popular in the industrial internet of thing (IIoT), since it can use only one twisted pair to provide high speed data transmission while the cables of the field bus can be reused. However, since its transmission medium is inferior to traditional Ethernet, it is easier to generate delay jitter that greatly impacts the accuracy of time synchronization. To solve this problem, in this paper, an offset estimation method based on AutoRegressive Integrated Moving Average (ARIMA) and Long Short-Term Memory (LSTM) is proposed to estimate the clock offset when the delay jitter appears. The offset estimation model is firstly obtained by training the ARIMA-LSTM with offline offset data. When the delay jitter is detected, the offset can be estimated by the model to replace the unreliable offset obtained by the time synchronization protocol. Experiments are executed in the testbed, and the results prove that the proposed method can improve the time synchronization accuracy in the single twisted pair Ethernet.