CAREER: Robust Modeling and Analysis of Discrete-Event Dynamic Systems

这项拨款为开发新的方法和支持理论提供资金，用于使用数学规划对大规模随机离散事件动态系统进行鲁棒建模和分析。该研究将涉及离散事件系统的鲁棒建模和灵敏度分析以及系统动力学的结构特性分析。这项工作将结合仿真和数学规划方法，以获得更有效的仿真模型，预测仿真模型中的逻辑错误，并获得更稳健的仿真结果。将对离散事件系统动力学进行解析建模，以便在传统方法失效时对仿真进行灵敏度分析。这将使建模者能够从单次模拟运行中提取更多信息，有效地利用非模拟实验的部分信息，并支持大规模系统的实时灵敏度信息应用。大规模的、真实世界的系统将为通过项目开发的概念和方法的实现提供一个测试平台。为本科生和研究生提供多学科研究机会将是该项目的重要成果，该项目以综合教育计划为特色，包括课程重组、指导活动和K-12外展。预计这项研究将带来显著的好处。首先，将开发新的基于数学规划的方法来实现模型效率，减少误差，以及在存在输入不确定性和/或模型误差的情况下进行离散事件模拟的鲁棒实验。其次，这项工作将导致对仿真系统动力学和分析它们的分析方法的更好理解。第三，实验人员将能够从单次模拟运行中提取更多信息，并应用稳健的灵敏度分析程序。最后，新的基于数学规划的模拟方法，用于离散事件模拟的建模和分析，将提高系统建模过程的有效性和效率，以实现广泛的实际应用。

This grant provides funding for the development of new methodologies and supporting theory for robust modeling and analysis of large-scale stochastic discrete-event dynamic systems using mathematical programming. The research will involve robust modeling and sensitivity analysis of discrete-event systems as well as structural property analyses of system dynamics. The work will combine simulation and mathematical programming approaches to get more efficient simulation models, predict logical errors in simulation models, and obtain more robust simulation results. Discrete-event system dynamics will be modeled analytically to support sensitivity analyses for simulation when traditional methods fail. This will enable modelers to extract more information from a single simulation run, make effective use of partial information for non-simulated experiments and support the application of real-time sensitivity information for large-scale systems. Large-scale, real-world systems will provide a test bed for implementation of the concepts and methodologies developed through the project. Multidisciplinary research opportunities for undergraduate and graduate students will be an important outcome of this project which features an integrated educational plan leading to course restructuring, mentorship activities, and K-12 outreach.Significant benefits are anticipated from this research. First, new math-programming-based approaches to achieving model efficiency, error reduction, and robust experiments for discrete-event simulation in the presence of input uncertainties and/or model errors will be developed. Second, the work will lead to an improved understanding of simulation system dynamics and analytical methodologies for analyzing them. Third, experimenters will be able to extract more information from a single simulation run and apply robust sensitivity analysis procedures. Ultimately, new math-programming-based simulation methodologies for modeling and analysis of discrete-event simulation will improve the effectiveness and efficiency of the systems modeling process for a wide range of practical applications.