A structuring review on multi-stage optimization under uncertainty: Aligning concepts from theory and practice
A structuring review on multi-stage optimization under uncertainty: Aligning concepts from theory and practice
复制标题
DOI：
10.1016/j.omega.2019.06.006
复制
发表时间：
2020-10
影响因子：
6.9
通讯作者：
Hannah Bakker;Fabian Dunke;S. Nickel
中科院分区：
文献类型：
--
作者：
Hannah Bakker;Fabian Dunke;S. Nickel
虽然在过去的几十年中，不确定性下的优化方法已经得到了深入的研究，明确考虑不确定性和时间之间的相互作用，最近得到了越来越多的关注。需要对不确定性实现做出反应的一系列决策的问题在现实世界的应用中至关重要，例如，供应链计划、调度或财务。已经开发了几种强调这些问题的不同方面的方法，主要由特定的应用程序触发。虽然这些方法都打算解决一个类似的基本问题，他们有很大的不同方面的不确定性表示，规定的解决方案提供的信息和性能评估的手段。其结果是一个不确定的多阶段问题的支离破碎的图片-无论是从方法和面向应用的角度。它未能将不同学科的结果互连起来，甚至无法比较特定应用中各个方法的优缺点。本文旨在将解决不确定性造成的多阶段优化问题的不同方法整合到一个更广泛的画面中，从而为更全面的方法铺平道路，顺序决策下的不确定性。为此，首先对这些方法的历史发展作了沿着的描述。其次，概述了它们的主要应用领域。我们的结论是，解耦不确定性模型的解决方案的方法和开发标准化的性能指标的关键步骤，组织多阶段优化下的不确定性，并引发进一步的潜力，尚未开发的组合的不确定性模型和解决方案的方法。
While methods for optimization under uncertainty have been studied intensely over the past decades, the explicit consideration of the interplay between uncertainty and time has gained increasing attention rather recently. Problems requiring a sequence of decisions in reaction to uncertainty realizations are of crucial relevance in real-world applications, e.g., supply chain planning, scheduling, or finance. Several methods emphasizing varying aspects of these problems have been developed, mainly triggered by a particular application. Although these methods all intend to solve a similar underlying problem, they differ strongly with respect to the uncertainty representation, the prescriptive solution information they provide and the means of performance evaluation. The result is a fragmented picture of uncertain multi-stage problems – both from a methodological and an application-oriented perspective. It fails to interconnect results from different disciplines or even comparing strengths and weaknesses of individual methods in particular applications. This review aims at integrating the different methods for solving uncertainty inflicted multi-stage optimization problems into a broader picture, thereby paving the way for more comprehensive approaches to sequential decision making under uncertainty. For this purpose, a description of the methods along with their historic development is given first. Secondly, an overview on their main areas of application is provided. We conclude that decoupling uncertainty models from solution methods and developing standardized performance measures represent key steps for organizing multi-stage optimization under uncertainty and for eliciting further potentials of yet unexplored combinations of uncertainty models and solution methods.