Deep Reinforcement Learning-Driven Scheduling in Multijob Serial Lines: A Case Study in Automotive Parts Assembly

Multijob production (MJP) is a class of flexible manufacturing systems, which produces different products within the same system. MJP widely used in product assembly, and efficient scheduling crucial for productivity. Most existing methods are inefficient multijob serial lines with practical constraints. We propose deep reinforcement learning (DRL)-driven framework by properly considering constraints identical machines, finite buffers, machine breakdown, delayed reward. analyze starvation blockage time, derive DRL-driven strategy to reduce time balance loads. validate proposed using real-world factory data collected over six months from tier-one vendor world top-three automobile company. Our case study shows that improves average throughput 24.2% compared conventional approach.