Modeling and Predicting for Stochastic Traffic Based on Uncertainty Set in Time-Sensitive Networking

The uncertainty, aroused from the burst of stochastic traffic, will increase the delay of periodic traffic in Time-Sensitive Networking (TSN). To confront this challenge, a novel stochastic flow model based on the budget constraint uncertainty set is proposed to obtain interval prediction of future time points. Considering the coupling relationship between multiple periods, a mixed integer linear programming (MILP) model is first established to cut the redundant part of the box set, and the budget constraints are gradually generated. In addition, with the consideration of the strong correlation between adjacent periods, another kind of budget constraint is further introduced into the mentioned uncertainty set such that the accuracy of the uncertainty set can be further improved. Based on the uncertainty set with budget constraints, a group of linear inequalities covering the traffic data of future first-time points is established and thus prediction interval can be given by solving the inequalities. Finally, the stochastic traffic prediction interval results of several time points in the future can be obtained via rolling prediction technology. Numerical tests are implemented with the actual data to verify the efficiency of the budget uncertainty set.