Adaptive Robust Nonlinear Filter Algorithm for Time Synchronization

This paper adopts a fusion algorithm to deal with different noise estimates and outlier processing in nonlinear time-synchronized system. In the framework of the unscented Kalman filter, the statistical characteristic parameters of measurement noise are estimated based on the variational Bayesian process. Combined with the maximum a posterior(MAP) estimation method, we use the measured noise parameters obtained by the variational Bayesian method to estimate the process noise and simultaneously realize the real-time adaptive estimation of the two kinds of noise statistical properties. A judgment method based on the likelihood function of innovation is embodied in the algorithm to make real-time judgments on the time deviation outlier generated in the synchronization process, and smooth the outlier through the window sliding average. In the nonlinear time synchronization process where the noise is unknown and the measurement process contains outliers, the algorithm can adaptively complete the synchronization estimation of the covariance matrix of noise, and can effectively judge and process the clock deviation outliers in the synchronization process.