TY - GEN
T1 - Unified DRL for Enhanced Flexible Job-Shop Scheduling with Transportation Constraints
AU - Wang, Yijie
AU - Wang, Runqing
AU - Sun, Jian
AU - Wang, Gang
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - In the domain of smart manufacturing systems (SMSs),the flexible job-shop scheduling problem with transportation constraints (FJSPT) represents a crucial challenge and holds the potential to significantly enhance production efficiency. Extending beyond the traditional job-shop scheduling problem (JSP), FJSPT incorporates the scheduling of automated guided vehicles (AGVs) to transport intermediate products between machines. Despite the promise of deep reinforcement learning (DRL) in solving combinatorial optimization problems, there remains a notable scarcity of research employing DRL to tackle FJSPT. To address this gap, this paper introduces an end-to-end DRL approach for simultaneous scheduling of machines and AGVs in FJSPT. The proposed method amalgamates operation selection, machine assignment, and AGV planning into a unified decision-making process, leveraging a graph attention network (GAT) and the proximal policy optimization (PPO) algorithm for efficient and stable training. Experimental evaluations conducted on synthetic data and public instances demonstrate that the proposed method outperforms existing approaches in terms of scheduling performance and computational efficiency.
AB - In the domain of smart manufacturing systems (SMSs),the flexible job-shop scheduling problem with transportation constraints (FJSPT) represents a crucial challenge and holds the potential to significantly enhance production efficiency. Extending beyond the traditional job-shop scheduling problem (JSP), FJSPT incorporates the scheduling of automated guided vehicles (AGVs) to transport intermediate products between machines. Despite the promise of deep reinforcement learning (DRL) in solving combinatorial optimization problems, there remains a notable scarcity of research employing DRL to tackle FJSPT. To address this gap, this paper introduces an end-to-end DRL approach for simultaneous scheduling of machines and AGVs in FJSPT. The proposed method amalgamates operation selection, machine assignment, and AGV planning into a unified decision-making process, leveraging a graph attention network (GAT) and the proximal policy optimization (PPO) algorithm for efficient and stable training. Experimental evaluations conducted on synthetic data and public instances demonstrate that the proposed method outperforms existing approaches in terms of scheduling performance and computational efficiency.
UR - http://www.scopus.com/inward/record.url?scp=86000759203&partnerID=8YFLogxK
U2 - 10.1109/CAC63892.2024.10864691
DO - 10.1109/CAC63892.2024.10864691
M3 - Conference contribution
AN - SCOPUS:86000759203
T3 - Proceedings - 2024 China Automation Congress, CAC 2024
SP - 877
EP - 882
BT - Proceedings - 2024 China Automation Congress, CAC 2024
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2024 China Automation Congress, CAC 2024
Y2 - 1 November 2024 through 3 November 2024
ER -