Files
Abstract
This dissertation studies the dynamic scheduling of high-dimensional multiclass queueing systems in the Halfin–Whitt regime, using a deep learning-based computational framework. Chapter 1 focuses on a multiclass, single server pool queueing system, while Chapter 2 generalizes the study to multiclass, parallel-server systems. In Chapter 1, we consider a multiclass queueing model of a telephone call center, in which a system manager dynamically allocates available servers to customer calls. Calls can terminate through either service completion or customer abandonment, and the manager strives to minimize the expected total of holding costs plus abandonment costs over a finite horizon. Focusing on the Halfin-Whitt heavy traffic regime, we derive an approximating diffusion control problem and building on earlier work by Beck et al. (2021) develop a simulation-based computational method for solution of such problems, one that relies heavily on deep neural network technology. Using this computational method, we propose a policy for the original (pre-limit) call center scheduling problem. Finally, the performance of this policy is assessed using test problems based on publicly available call center data. For the test problems considered so far, our policy does as well as or better than the best benchmark we could find. Moreover, our method is computationally feasible at least up to dimension 500, that is, for call centers with 500 or more distinct customer classes. In Chapter 2, we extend this framework to multiclass, parallel-server queueing systems with heterogeneous service stations, each consisting of multiple identical servers. Since not all customer classes are served by all service stations, the system forms a bipartite network. We consider a discounted-cost formulation over an infinite time horizon, in which the system manager chooses a scheduling/routing policy to minimize the expected discounted cost. We study general parallel-server systems, allowing for non-work-conserving policies and incorporating both basic and nonbasic activities. Focusing on the Halfin–Whitt regime, we derive an approximating diffusion control problem and building on the work by Han et al. (2018) we develop a simulation-based computational method that leverages deep neural networks to solve such problems. Using this method, we propose a policy for the original (pre-limit) parallel-server queueing system. Across the test problems considered so far, which are calibrated using publicly available call center data, our proposed policy performs as well as or better than the best available benchmark.