On scheduling a multiclass queue with abandonments under general delay costs (Q352983)

In this paper, the authors provide a new solution for a multiclass queuing system with abandonments under general delay costs. In particular, the paper considers dynamic scheduling decisions to minimise long-run average delays and abandonments costs such as i) linear, ii) convex and iii) convex-concave. A multi-class queuing system is modelled by approximating the Brownian Control Problem (BCP). This approach does not admit a pathwise solution due to abandonments. The solution was obtained by using the associated Bellman equation, which fields a dynamic index policy as the optimal control for approximating the BCP. Finally, a novel method for constructing a solution to the Bellman equation is proposed based on policies. To illustrate the effectiveness of the proposed policies, a simulation example is presented. In this analysis, the convex-concave delay cost cases are studied and are reviewed. A comparison with the generalised \(c \mu\) policy is also made. The proposed solution demonstrates the effectiveness of the proposed policies in terms of standard errors of average delay costs and abandonment costs.