Deep reinforcement learning-based spatio-temporal graph neural network for solving job shop scheduling problem

被引:0
作者
Gebreyesus, Goytom [1 ]
Fellek, Getu [1 ]
Farid, Ahmed [1 ]
Hou, Sicheng [1 ]
Fujimura, Shigeru [1 ]
Yoshie, Osamu [1 ]
机构
[1] Waseda Univ, Grad Sch Informat Prod & Syst, Fukuoka, Japan
关键词
Deep reinforcement learning; Spatio-temporal representation; Job shop scheduling; Graph neural network; MIGRATING BIRDS OPTIMIZATION; ALGORITHM; BENCHMARKS; TIME;
D O I
10.1007/s12065-024-00989-6
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The job shop scheduling problem (JSSP) is a well-known NP-hard combinatorial optimization problem that focuses on assigning tasks to limited resources while adhering to certain constraints. Currently, deep reinforcement learning (DRL)-based solutions are being widely used to solve the JSSP by defining the problem structure on disjunctive graphs. Some of the proposed approaches attempt to leverage the structural information of the JSSP to capture the dynamics of the environment without considering the time dependency within the JSSP. However, learning graph representations only from the structural relationship of nodes results in a weak and incomplete representation of these graphs which does not provide an expressive representation of the dynamics in the environment. In this study, unlike existing frameworks, we defined the JSSP as a dynamic graph to explicitly consider the time-varying aspect of the JSSP environment. To this end, we propose a novel DRL framework that captures both the spatial and temporal attributes of the JSSP to construct rich and complete graph representations. Our DRL framework introduces a novel attentive graph isomorphism network (Attentive-GIN)-based spatial block to learn the structural relationship and a temporal block to capture the time dependency. Additionally, we designed a gated fusion block that selectively combines the learned representations from the two blocks. We trained the model using the proximal policy optimization algorithm of reinforcement learning. Experimental results show that our trained model exhibits significant performance enhancement compared to heuristic dispatching rules and learning-based solutions for both randomly generated datasets and public benchmarks.
引用
收藏
页数:18
相关论文
共 55 条
[11]   Gated-Attention Model with Reinforcement Learning for Solving Dynamic Job Shop Scheduling Problem [J].
Gebreyesus, Goytom ;
Fellek, Getu ;
Farid, Ahmed ;
Fujimura, Shigeru ;
Yoshie, Osamu .
IEEJ TRANSACTIONS ON ELECTRICAL AND ELECTRONIC ENGINEERING, 2023, 18 (06) :932-944
[12]  
Gunarathna U., 2022, arXiv, DOI DOI 10.48550/ARXIV.2201.04895
[13]   Deep attention models with dimension-reduction and gate mechanisms for solving practical time-dependent vehicle routing problems [J].
Guo, Feng ;
Wei, Qu ;
Wang, Miao ;
Guo, Zhaoxia ;
Wallace, Stein W. .
TRANSPORTATION RESEARCH PART E-LOGISTICS AND TRANSPORTATION REVIEW, 2023, 173
[14]   Job Shop Scheduling: A Novel DRL approach for continuous schedule-generation facing real-time job arrivals [J].
Hammami, Nour El Houda ;
Lardeux, Benoit ;
Hadj-Alouane, Atidel B. ;
Jridi, Maher .
IFAC PAPERSONLINE, 2022, 55 (10) :2493-2498
[15]   A DEEP REINFORCEMENT LEARNING BASED SOLUTION FOR FLEXIBLE JOB SHOP SCHEDULING PROBLEM [J].
Han, B. A. ;
Yang, J. J. .
INTERNATIONAL JOURNAL OF SIMULATION MODELLING, 2021, 20 (02) :375-386
[16]   Research on Adaptive Job Shop Scheduling Problems Based on Dueling Double DQN [J].
Han, Bao-An ;
Yang, Jian-Jun .
IEEE ACCESS, 2020, 8 :186474-186495
[17]  
Holland JH., 1976, Adaptation in Natural and Artificial Systems, DOI 10.7551/mitpress/1090.001.0001
[18]   Day-ahead multi-modal demand side management in microgrid via two-stage improved ring-topology particle swarm optimization [J].
Hou, Sicheng ;
Gebreyesus, Goytom Desta ;
Fujimura, Shigeru .
EXPERT SYSTEMS WITH APPLICATIONS, 2024, 238
[19]   Deterministic job-shop scheduling: Past, present and future [J].
Jain, AS ;
Meeran, S .
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH, 1999, 113 (02) :390-434
[20]  
Kardos Csaba, 2021, Procedia CIRP, V97, P104, DOI 10.1016/j.procir.2020.05.210