Applying Genetic Algorithm for Can-Order Policies in the Joint Replenishment Problem

Abstract

In this paper, we consider multi-item inventory management. When managing a multi-item inventory, we coordinate replenishment orders of items supplied by the same supplier. The associated problem is called the joint replenishment problem (JRP). One often-used approach to the JRP is to apply a can-order policy. Under a can-order policy, some items are re-ordered when their inventory level drops to or below their re-order level, and any other item with an inventory level at or below its can-order level can be included in this order. In the present paper, we propose a method for finding the optimal parameter of a can-order policy, the can-order level, for each item in a lost-sales model. The main objectives in our model are minimizing the number of ordering, inventory, and shortage (i.e., lost-sales) respectively, compared with the conventional JRP, in which the objective is to minimize total cost. In order to solve this multi-objective optimization problem, we apply a genetic algorithm. In a numerical experiment using actual shipment data, we simulate the proposed model and compare the results with those of other methods.

Reinforcement Learning for Optimizing Can-Order Policy with the Rolling Horizon Method

“…Kayiş, Bilgiç, and Karabulut [7] examined a two-item continuous-review inventory system with independent Poisson demand processes and economies of scale in joint replenishment and proposed and modeled a COP as a semi-Markov decision process using a straightforward enumeration algorithm. Nagasawa et al [8] suggested a method for determining the optimal parameter of a periodic COP for each item in a lost sales model using a genetic algorithm. The primary goal was to minimize the number of orders, inventory, and shortages compared to those in a conventional joint replenishment problem.…”

A Multi-Item Replenishment Problem with Carbon Cap-and-Trade under Uncertainty

“…To simulate their idea, they suggested a new method based on Markov decision theory to obtain a near-optimal solution. Nagasawa et al [21] presented a periodic can-order policy model that uses multi-objective programming to obtain the optimal can-order level. Most previous studies dealt with the continuous review system and focused on deriving the reorder level, can-order level, and order-up-to level based on various algorithms.…”