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

This study presents a novel approach to mixed-integer linear programming (MILP) model for periodic inventory management that combines reinforcement learning algorithms. The rolling horizon method (RHM) is multi-period optimization applied handle new information in updated markets. RHM faces limitation easily determining prediction horizon; overcome this, dynamic developed which RL algorithms optimize the of RHM. state vector consisted order-up-to-level, real demand, total cost, holding and backorder whereas action included forecasting demand next time step. performance proposed was validated through two experiments conducted cases with stable uncertain patterns. results showed effectiveness management, particularly when proximal policy (PPO) algorithm used training compared other signifies important advancements both theoretical practical aspects multi-item management.