An Analysis of the Coordination Gap between Joint and Modular Learning for Job Shop Scheduling with Transportation Resources

The study presents an analysis of job-shop scheduling with transportation resources, focusing on the effectiveness of joint versus modular training approaches in multi-agent reinforcement learning. It quantifies the "coordination gap," demonstrating that joint training generally yields better performance, though its advantages diminish in bottleneck scenarios with high constraints. This research offers practical insights for practitioners on selecting appropriate training modalities based on specific environmental conditions, enhancing scheduling efficiency in decentralized manufacturing systems.