Deep reinforcement learning for solving hybrid flow shop scheduling problems with limited buffers

Publication Type

Journal Article

Publication Date

1-2026

Abstract

The Hybrid Flow Shop Scheduling Problem with Limited Buffers (LBHFSP) is a challenging issue commonly studied in computer science and operations research. In this paper, we propose a neural heuristic based on deep reinforcement learning (DRL) to solve the LBHFSP and its special variant, the Blocking Hybrid Flow Shop Scheduling Problem (BHFSP), using an encoder-decoder structured policy network trained to sequentially construct the processing scheme. To determine the processing relationship between the job, stage, and machine, we design a two-module attention-based architecture. Particularly, the encoder includes a memory module and a mixture module, allowing us to account for the impact of previous stages and integrate the processing time into the decision-making process. The decoder includes a memory module and a mask module, which exploits stored information to select jobs to be processed on the current machine and ensure the feasibility of the solution. Evaluations demonstrate that our neural heuristic performs favourably on randomly generated instances and exhibits strong generalisation capabilities. In real-world data of the excavator production line in Sany Heavy Machinery Ltd., it can improve over 3% compared to actual algorithms, proving that our method remains effective even when the number of machines varies across different stages.

Discipline

Artificial Intelligence and Robotics | Operations Research, Systems Engineering and Industrial Engineering

Research Areas

Intelligent Systems and Optimization

Publication

International Journal of Production Research

Volume

64

Issue

14

First Page

5735

Last Page

5756

ISSN

0020-7543

Identifier

10.1080/00207543.2026.2615080

Publisher

Taylor and Francis Group

Additional URL

https://doi.org/10.1080/00207543.2026.2615080

This document is currently not available here.

Share

COinS