Solving Multi-objective Production Inventory Control Problems with Reinforcement Learning aided by Differential Evolution

doi:10.1016/j.asoc.2026.115534

Solving Multi-objective Production Inventory Control Problems with Reinforcement Learning aided by Differential Evolution

2026-05-21

0

PRE

AI

DOI：10.1016/j.asoc.2026.115534

Original

Original request for help

Save

Abstract

En 中文

• The production inventory control problem in cold rolling process of silicon steel is formulated as an LQT model that incorporates stock-dependent demand and uncertain parameters. • A novel algorithm is proposed that integrates learning and optimization methods, where Q-learning addresses system uncertainties and an evolutionary algorithm jointly optimizes parameters in both the model and learning algorithm. • Compared to other advanced inventory control and parameter optimization methods, the proposed algorithm not only reduces operational costs but also generates control policies with greater preference diversity. • Managerial insights are derived through sensitivity analysis.

Keywords:

Production inventory control

Silicon steel

Q-learning

Differential evolution

Multi-objective optimization

Solving Multi-objective Production Inventory Control Problems with Reinforcement Learning aided by Differential Evolution

Abstract

Journal

Organization