Unbiased gradient estimation in queueing networks with parameter-dependent routing

TL;DR

This paper introduces an unbiased gradient estimation method for queueing networks with parameter-dependent routing, providing theoretical conditions for unbiasedness and an algorithm for practical implementation.

Contribution

It proposes a new unbiased gradient estimation technique for queueing networks with parameter-dependent routing, along with sufficient conditions and an algorithm.

Findings

The gradient estimate is unbiased under general conditions.

A practical algorithm for gradient estimation is developed.

The method applies to queueing networks with complex routing mechanisms.

Abstract

A stochastic queueing network model with parameter-dependent service times and routing mechanism, and its related performance measures are considered. An estimate of performance measure gradient is proposed, and rather general sufficient conditions for the estimate to be unbiased are given. A gradient estimation algorithm is also presented, and its validity is briefly discussed.