摘要：Abstract In a production/inventory system, invisible stock loss can lead to inventory inaccuracy that has negative impact on the replenishment process and will cause severe stock-out. In this paper, two robust replenishment control policies are designed for such a system. The first policy is developed based on a partially observed Markov decision process (POMDP) model of this system, and the second policy is constructed based on dynamic programming by replacing the random stock loss rate and demand rates with their mean values. Numerical experiments show that the POMDP-based policy generates more profit than the DP-based policy, and the DP policy performs better than the replenishment policy that simply ignores all the stock loss.
