Upgrades, Reneging and Retrials: Advanced Dynamic Control of Service Systems

一些人估计，美国在呼叫中心工作的人数高达300万。这与众所周知的美国经济从制造业基地向服务业基地的转变是一致的。本文的研究内容是研究服务系统的动态控制问题，主要创新有三个方面：升级、放弃和重试。虽然每种情况的动机都来自呼叫中心，但所获得的结果将更普遍地适用。应考虑以下问题：？在市场细分后，是否对经典的资源分配策略进行了简单的修改，以减少低级别客户的等待时间，同时保证高级别客户的高质量服务？从经典分配政策中得出的见解在今天的先进系统中仍然有用吗？在录取率受控制的模型中，在当今新的决策模式下，有没有简单、可实施的政策？为了解决这些问题，我们提出了具体的服务系统模型。每一种模式都从理论和管理的角度提出了新的挑战。如果成功，将为服务系统中的最优资源分配提供见解，并为服务系统控制的新探索打开大门。我们所提出的研究的智力价值包括检验一些经典的调度和准入控制模型的扩展。这些扩展将更新最先进的技术，以涵盖更多最近的服务系统创新，并回答美国当前服务经济中引入的新范式。在某些情况下，我们预计，随着模型变得更加复杂，所获得的洞察力可以作为对盲目使用经典结果的警告。事实上，拟议的模型甚至没有提出最笼统的建议，但即使到目前为止，其见解也是深刻的。作者继续努力争取少数族裔顶尖研究生参与到他的工作中来。此外，研究结果将加强作者在康奈尔大学教授的应用概率和运筹学研究/管理高级课程的课堂教学。

Some estimates put the number of people in the U.S. that work in call centers as high as 3 million. This is consistent with the well-known shift of the U.S. economy from a manufacturing base to a service base. The proposed research is to study the dynamic control of service systems with three major innovations: upgrades, reneging, and retrials. Although the motivation in each case stems from call centers, the results obtained will be applicable more generally. The following questions are to be considered:? After market segmentation, are there simple modifications to classic resource allocation policies that can reduce waiting times for lower class customers, while guaranteeing high quality service for higher class customers?? Do the insights drawn from classic allocation policies continue to be useful in the advanced systems of today?? In models with controlled the admission rates, are there simple, implementable policies given today's new decision-making paradigms?To tackle each of these questions specific models of service systems are presented. Each model presents new challenges from both a theoretical and managerial standpoint. If successful, insights into optimal resource allocation in service systems will be provided as well as opening the door for new exploration of service system control.The intellectual merits of our proposed research include examining extensions of some of the classic models in scheduling and admission control. These extensions stand to update the state of the art to cover more recent innovations in service systems and answer new paradigms introduced in the current service economy of the United States. In some cases we expect that the insights gained can serve as a warning against blindlyusing classic results as the models become more intricate. Indeed, the proposed models are not even proposed the most general possible, yet the insights even so far are profound. The author continues to work vigorously to pursue top Minority graduate students to be involved in his work. Moreover, the results will enhance classroom instruction for advanced courses in applied probability and operations research/management that the author teaches at Cornell University.