Time-dependent queue length distribution in queues fed by $K$ customers in a finite interval

TL;DR

This paper analyzes the time-dependent queue length distribution in finite-interval queueing models with binomial arrivals, introducing an auxiliary model and a numerical method for Markovian cases with piecewise constant arrival densities.

Contribution

It introduces a novel auxiliary model with non-homogeneous Poisson arrivals and a numerical procedure for computing queue length distributions with controllable error bounds.

Findings

Derived the queue length distribution in terms of auxiliary model distributions

Developed a numerical method with adjustable truncation error bounds

Provided numerical examples demonstrating the method's effectiveness

Abstract

We consider queueing models, where customers arrive according to a continuous-time binomial process on a finite interval. In this arrival process, a total of $K$ customers arrive in the finite time interval $[0,T]$, where arrival times of those $K$ customers are independent and identically distributed according to an absolutely continuous distribution defined by its probability density function $f(t)$ on $(0,T]$. To analyze the time-dependent queue length distribution of this model, we introduce the auxiliary model with non-homogeneous Poisson arrivals and show that the time-dependent queue length distribution in the original model is given in terms of the time-dependent joint distribution of the numbers of arrivals and departures in the auxiliary model. Next, we consider a numerical procedure for computing the time-dependent queue length distribution in Markovian models with piecewise constant $f(t)$. A particular feature of our computational procedure is that the truncation error bound can be set as the input. Some numerical examples are also provided.