Managing Perishable Inventory Systems: New Algorithms and Approximations

这笔赠款为开发简单、高效和接近最优的易腐随机库存系统的近似算法提供了资金。生鲜食品、药品和血库等易腐烂产品无处不在，是我们社会不可或缺的一部分，变质和过期是对杂货零售商等公司盈利的主要威胁。因此，寻找有效的易腐产品库存管理策略具有重要意义。然而，由于动态易腐烂库存系统的高维特性，其分析在理论和计算上都是出了名的困难。事实上，即使在需求独立且分布相同的情况下，最优控制策略也是非常复杂的，而使用动态规划来计算最优策略通常是困难的，因为存在“维度诅咒”。所研究的模型允许一般的非平稳和相关的需求过程，捕捉到经济的季节性本质，开发的近似算法承认理论上最坏情况下的性能保证。如果研究成功，本研究结果将为库存管理者提供有效的工具，以有效地结合需求预测，如提前需求信息(ADI)、预测演化的鞅模型(MMFE)、自回归移动平均(ARMA)需求模型和马尔可夫调制需求过程(MMDP)，以制定易逝性库存系统的库存补充决策。它将使库存管理人员能够利用现有和可靠的数据，如预测更新，进行库存规划。这项研究将促进随机库存系统的科学研究，并加深对该学科领域的理解。该项目的成果将帮助企业减少浪费，增加收入，甚至拯救生命(例如，在血库应用方面)。

This grant provides funding for the development of simple, efficient, and near-optimal approximation algorithms for perishable stochastic inventory systems. Perishable products, such as fresh food, pharmaceuticals, and blood banks are ubiquitous and an indispensable part of our society, and spoilage and outdating represent a major threat to the profitability of companies such as grocery retailers. Thus, finding effective inventory management policies for perishable products is of significant importance. However, the analysis of dynamic perishable inventory systems is notoriously difficult in both theory and computation due to the high-dimensional nature. Indeed, the optimal control policies are very complex even in the case of independent and identically distributed demands, and the computation of optimal policies using dynamic program is in general intractable due to the "curse-of-dimensionality." The models studied allow general non-stationary and correlated demand processes, capturing the seasonality nature of the economy, and the approximation algorithms developed admit theoretical worst-case performance guarantees. If successful, the results of this research will lead to efficient tools for inventory managers to effectively incorporate demand forecast, such as advance demand information (ADI), martingale models of forecast evolution (MMFE), autoregressive moving average (ARMA) demand models, and Markov modulated demand process (MMDP), in making inventory replenishment decisions for perishable inventory systems. It will allow inventory managers to make use of available and reliable data, such as forecast updating, for inventory planning. The research will advance the science of stochastic inventory systems and provide deeper understanding of the subject area. The outcome of the project will help firms reduce waste, increase revenue, and even save lives (e.g., in blood bank applications).