Diffusion approximation for open state-dependent queueing networks in the heavy traffic situation (Q1916479)

The author considers an open queueing network in which arrival and service rates depend on the state of the network, i.e. on queue length. They are modeled as multidimensional birth and death processes. It is shown that a properly normalized sequence of queue lengths converges to a reflecting diffusion process under a heavy traffic condition. For a \(K\)-station queueing network, this diffusion takes values in the \(K\)-dimensional nonnegative orthant with a fixed direction of reflection for each boundary hyperplane and has nonsingular drift and diffusion coefficients. This drift and the diffusion coefficient depend on the state of the process because the arrival and service rates depend on the state of the queueing system. This contrasts with the result obtained by \textit{M. T. Reiman} for networks, where the arrival streams are renewal processes and service times have general probability distribution not depending on the state of the network [Math. Oper. Res. 9, 441-458 (1981; Zbl 0549.90043)].