基于强化学习算法的供应链自适应随机库存控制研究

针对非平稳随机需求的多级多周期供应链库存控制,本文建立了一种包括一个供应商和多个零售商的供应链分散式自适应库存控制模型,以满足给定的零售商服务水平。同时,运用强化学习算法,并根据需求变化情况,供应商和零售商分别自适应地调整库存控制参量。仿真试验表明,当相对需求分布已知,而需求未知时,订货量和服务水平都相对不稳定;安全因子范围大的,订货量和服务水平的波动相对较大,且能够更快的把服务水平调整到目标服务水平区间内。该模型是合理和有效的。

A Reinforcement Learning-Based Adaptive Supply Chain Stochastic Inventory Control

In this paper, we propose a decentralized adaptive inventory control model for a multi-level multi-cycle supply chain consisting of one supplier and multiple retailers with non-stationary random demand. The objective is to satisfy a given interval target service level predefined for each retailers. Using reinforcement learning algorithm, the supplier and the retailer can adaptively adjust their inventory control parameters according to the change of demand respectively. Simulation experiments show that, contrast to stationary demand case, in nonstationary demand situation, order and service level are relatively unstable for a large range of safety factor, and faster service levels adjusted to within the range of the target service levels. Therefore, the model is reasonable and effective.