Policy-Robust Processing Networks: Characterization and Design

点击翻译按钮获取中文摘要

This award will contribute to the nation's economic competitiveness by improving the design and operation of stochastic processing networks. Stochastic processing networks, consisting of interconnected processing nodes with some flexibility to manage their own schedules, arise in a variety of settings, including many service systems and some manufacturing systems. The fact that arrival times, processing times, and routes through the network may not be predictable with certainty makes these networks particularly difficult to analyze. This award supports research into conditions for global stability of these networks where each node in the network has some autonomy and independent scheduling authority; that is, the larger network does not operate under a centralized scheduler. Stability of the network ensures that any entity entering the network will eventually be processed and leave the network. The project studies stochastic processing networks incorporating renewal arrival processes, service times with general distributions, Bernoulli routing, and multiple classes of jobs at each station. A primary interest is the global stability of these networks, i.e., stability of the network as a whole when each individual node operates under its own non-idling scheduling policy. In these networks, control is distributed, rather than being determined by a central scheduler. The major goal of this research is the development of a new framework for global stability that can support the design of operating rules that allow fairly general decentralized control. The project will lead to a better understanding of the design of robust complex networks. The techniques employed involve ideas from fluid and diffusion approximations to queueing networks, along with notions from robust optimization.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.