Regret-Minimizing Inventory Policies with Lost Sales and Stochastic Supply Constraints
Keywords:
Inventory control, regret minimization, lost sales, stochastic supply, online decision-making, supply chain optimization, adaptive policies, robust inventory management, learning algorithms, uncertain demand.Abstract
Inventory management under uncertainty is a central challenge in operations research, particularly when dealing with lost sales and stochastic supply constraints. In such settings, traditional optimization techniques that assume full knowledge of demand and supply distributions often fail to perform well in practice. This paper introduces a regret-minimizing approach to inventory control that evaluates performance relative to the best fixed policy in hindsight, rather than relying on expected cost minimization. The framework is designed to be adaptive, data-driven, and robust to real-time variability in both supply availability and customer demand. We explore the theoretical motivations for using regret as a performance metric, analyze structural characteristics of effective policies, and discuss real-world applications such as humanitarian logistics and e-commerce fulfillment. By focusing on learning-based strategies, this work bridges the gap between online decision theory and practical inventory systems in uncertain environments.
