Kernel Based Nonlinear Regression for Internet Round Trip Time-delay Prediction

Time delay degrades the performances of Internet-based control systems or teleoperation system, and even causes instability of closed-loop systems. If the Round Trip Time-delay (RTT) is acquired exactly previously, it is helpful to improve the performance of Internet-based control systems. Therefore, analysis or prediction for Internet time delay has become a hot research problem. This paper proposes the long-range nonlinear autocorrelation of RTT based on the real data of Internet delay measurements. Then, according to the characters, we present a sparse matrix based kernel regression (SMKR) scheme to predict RTT. Finally, simulation results show that the forecasting precision using this method is higher than that using sparse multivariate linear regressive (SMLR) method, which demonstrates the validity of the proposed approach.