Simulation-Optimization Methods and Modelling-to-Generate Alternatives for Planning Under Uncertainty

当众多的系统组件表现出相当程度的不确定性时，综合规划问题就显得特别复杂。这种复杂性常常与许多事先无法确定的随机因素混合在一起。通常不存在精确的分析公式，即使存在，也包含大量高度随机的非线性成分。为如此大的随机问题确定好的解决方案，同时考虑到固有的不确定性，是非常困难的。不管这些困难如何，这种规划必须在“现实世界”决策、规划和政策制定的几乎每一个领域中执行。模拟优化（SO）为构建复杂规划问题的最佳解提供了一种包含以概率分布表示的不确定性的优化方法。在SO中，所有未知的目标、约束和参数都被仿真模型所取代，其中决策变量提供了每个仿真实验运行的设置。一个有效的优化组件通过可行域引导解决方案探索，只执行有限数量的模拟。虽然我之前的一些研究已经证明了SO可以在大规模的规划应用程序中产生令人印象深刻的结果，但是不同的实现过程的解决时间/质量可能会有很大的不同——从而限制了它对所有规划目的的普遍适用性。将寻求改进搜查程序。我的研究将调查通过减少解决时间和/或为上述大规模决策和计划问题提供更好质量的解决方案来提高SO性能的各种方法。

Comprehensive planning problems prove particularly complicated when their numerous system components exhibit considerable degrees of uncertainty. This complexity is frequently compounded by numerous stochastic elements that cannot be definitively ascertained beforehand. Often precise analytical formulations do not exist and, even when they do, contain a multitude of highly stochastic, non-linear components. Determining good solutions to such large stochastic problems, while simultaneously incorporating the inherent uncertainties, can prove exceedingly difficult. Irrespective of such difficulties, this planning must be performed in virtually every sphere of "real world" decision-making, planning, and policy-formulation. Simulation-optimization (SO) provides an optimization approach that incorporates uncertainties expressed as probability distributions for constructing best solutions to complex planning problems. In SO all unknown objective(s), constraints, and parameters are replaced by simulation models in which the decision variables provide the settings under which each simulation experiment is run. An efficient optimization component guides the solution exploration through the feasible domain performing only a limited number of simulations. While some of my previous research has demonstrated that SO can produce impressive results in large-scale planning applications, the solution time/quality of the procedure can vary considerably from one implementation to another - thereby limiting its universal applicability for all planning purposes. Improved search procedures will be sought. My research will investigate a variety of approaches for improving the performance of SO by decreasing the solution time and/or producing better quality solutions for the aforementioned large-scale decision making and planning problems. Numerous "real world" applications of the procedure will be considered and extensive testing will be employed to ascertain SO's suitability (and, just as importantly, its non-suitability) in disparate planning environments. In efforts to gauge and demonstrate the efficiency improvements from the new approaches, I will continue with data from several earlier case studies that have used this technique in such diverse settings as municipal solid waste planning, reverse logistics & remanufacturing of leaded waste products, planning & management of the extended supply chain, and environmental sustainability. These applications all possess considerable public and environmental importance. Several novel applications for SO will also be sought and the benefits of its application will be demonstrated. Additional research will extend the method (and all its newly-found efficiencies) into multi-objective problem situations that contain considerable uncertainty. This will provide significant theoretical and practical advances.