Solving Tactical Logistics Planning Problems Under Uncertainty

这项拨款支持在不确定需求下战术后勤规划问题的新理论和模型的发展。本研究的核心决策问题焦点在于生产和物流系统中下游需求与上游资源能力的最优匹配。建模方法隐含地说明了实际物流系统的一个重要结构特征，它需要确定一个唯一的上游源设施（例如，工厂，仓库）来服务于每个下游阶段（例如，仓库，零售商）。这种在生产和物流系统中普遍存在的结构既降低了下游订购和交付接收的复杂性，又减少了上游跨设施协调的需求。作为这些中期决策的结果，短期运营计划和执行问题分解为单独的子问题（每个上游设施一个子问题），从而降低了系统范围内运营计划的复杂性。本研究采用优化物流网络性能的综合观点，其中运输决策与驱动整体性能的其他物流网络决策（如生产和库存存储决策）联合考虑，同时明确考虑需求不确定性。该项目将为有效解决战术物流规划问题提供综合方法，从而为简化配送系统成本提供机会。特别是，新的和有效的算法将被开发出来，可以用有限的计算量来解决中型到大型问题到接近最优，并且还享有正式的性能保证。相关的算法将在一般模型设置中开发，允许专业化到广泛的特定问题上下文。此外，通过与优化软件开发公司的合作，它将提供一种解决方案方法，该方法将允许在解决其他问题时重用为一个问题开发的大部分软件，从而加快软件开发时间。

This grant supports the development of new theory and models for tactical logistics planning problems under uncertain demand. The core decision problem focus of this research lies in optimally matching downstream requirements with upstream resource capabilities in production and logistics systems. The modeling approach implicitly accounts for an important structural feature of practical logistics systems, which requires identifying a unique upstream source facility (e.g., plant, warehouse) for serving each downstream stage (e.g., warehouse, retailer). This pervasive structure found in production and logistics systems both reduces downstream ordering and delivery receiving complexity, and decreases the need for upstream cross-facility coordination. As a result of these medium-term decisions, short-term operations planning and execution problems decompose into separate subproblems (one for each upstream facility), which reduces system-wide operations planning complexity. This research takes an integrated view of optimizing logistics network performance, where transportation decisions are considered jointly with other logistics network decisions that drive overall performance, such as production and inventory storage decisions, while explicitly accounting for demand uncertainty.This project will provide an integrated approach for solving tactical logistics planning problems effectively, thus providing opportunities for streamlining distribution system costs. In particular, new and effective algorithms will be developed that can solve medium- to large-size problems to near-optimality with limited computational effort, and also enjoy a formal performance guarantee. The associated algorithms will be developed in a general model setting, allowing for specialization to a broad range of specific problem contexts. In addition, through collaboration with an optimization software development firm, it will provide a solution methodology that will allow for the reuse of much of the software developed for one problem in the solution of other problems, thereby speeding up software development time.