William Millhiser. Admission Policies in Multiclass Tandem Queues with Blocking: New Approaches and Results due to Dynamic Programming Event Operators

Abstract. Consider one of the many service operations with relatively inflexible capacity for which dynamic pricing--that is, dynamic admission control--has been shown to improve profits and more accurately match demand to capacity. In many such systems, service is provided by tandem queueing networks, sometimes with blocking due to finite customer waiting room capacity. For example, doctors are blocked from performing elective surgeries when all beds are occupied in rehabilitative care facilities in certain US healthcare networks. An alternative viewpoint could be that of a service system with a gatekeeper who makes the admission decision and performs a preliminary service before the customer or job joins the queue for the specialist.

Relatively little is known about the optimal admission control policies in such networks. We will review existing literature and real-world problems. We use recent advances in dynamic programming "event operators" to obtain new results about optimal admission policies, and to perform sensitivity analysis on those policies with regard to various parameters such as arrival rates, services rates and the number of servers. Prior to the introduction of the event operators, this sensitivity analysis was largely intractable on general problems.