基于加权多元线性回归的两步式时间同步算法

时间同步在通信、计算机、测控等网络中是一项重要技术。针对目标时钟在多同步源时的时间同步问题,从统计学中参数估计的角度建模,提出了一种基于WMLR (Weighted Multiple Linear Regression,加权多元线性回归)的两步式时间同步算法。此算法分为粗同步和精同步2个阶段,通过给不同的时间信息加权的方式提高同步精度,并且可以计算不同同步源之间存在的系统误差。结果表明,这一算法通过合理应用来自不同同步源的时间信息,可以提高时间同步的精度及稳定性。由于模型的普遍性,这一基于WMLR的算法也可以应用于其他类似的多源参数估计的场景中。

Two-Step Time Synchronization Based on Weighted Multiple Linear Regression

Synchronization is a key challenge in many communication systems. This paper proposes a two-step WMLR (Weighted Multiple Linear Regression) based synchronization scheme for synchronizatio nscenarios with dual synchronization sources. In the proposed scheme, the clock offset and clock skew are first estimated with NLR (Normal Linear Regression) using timing information from two sources separately. Unlike most synchronization methods that select estimates from one source, the proposed scheme efficiently uses timing information from both sources with WMLR, where the weights were also conducted in the first step, and gives better estimates for clock offset and clock skew. Simulation results show that the proposed scheme provides better synchronization accuracy and better stability than any of individual sources.