Queueing Systems with Server Synchronization

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

This project is concerned with improving business operations through the development of enhanced workforce management techniques. The goal is to determine how workers should be assigned dynamically to tasks in the presence of uncertain demand and processing requirements for various products and services. The main novelty is the incorporation of synchronization constraints on worker assignment. For example, the completion of certain tasks may require a team of workers, whereas other tasks require one worker at a time. In both cases, workers must synchronize their schedules in that they must be simultaneously at either the same or different tasks. This project involves the development of easily implementable, robust, and dynamic policies for assigning workers to tasks in real time, taking into account such synchronization constraints on worker assignment. Potential application areas include health care delivery, job shop scheduling, call centers, project management, and parallel computing. If successful, this award will yield valuable insights and concrete guidelines for how the performance of a large class of production and service systems can be improved at low cost by better utilizing the available workers and other resources. The planned research involves modeling businesses as queueing networks and workers as servers. The research objectives will be achieved through sample-path analysis, Markov decision processes, fluid limit analysis, computer simulation, and linear programming. In particular, for networks with general topologies and infinite buffers, fluid limit analysis will be used to identify what fraction of time the servers should spend in the various possible server configurations. Then dynamic server allocation policies that come arbitrarily close to being optimal will be developed. On the other hand, systems with finite buffers will be analyzed using Markov decision process techniques, where there will be constraints on server assignments. The proposed research involves the rigorous analysis and control of queueing systems with flexible, heterogeneous servers when the schedules of different servers must be synchronized (e.g., to account for needed collaboration or non-collaboration). The focus is on maximizing system capacity (throughput), but other performance measures will also be considered, including holding costs.