The development of optimal control strategies for many stochastic models relies on the observed traffic intensity. However, implementation of such control strategies is often infeasible because of high operating costs induced by the fluctuations of traffic flows. In this study, we propose a framework for estimating and monitoring the traffic intensities of stochastic systems. The framework does not require knowledge of any input traffic statistics, and it allows us to adaptively estimate the intensity function over time and simultaneously detect its significant changes so that the control strategy can be adjusted accordingly without requiring high operating costs. Finally, a canonical queueing system with various types of input traffic is used to evaluate the effectiveness of the proposed framework.
Relation:
Applied Stochastic Models in Business and Industry, 30(2), 200-217