A Deep Reinforcement Learning (DRL)-Based Transformer Method for Solving the Open Shop Scheduling Problem

This study introduces a Transformer-based scheduling policy for the Open Shop Scheduling Problem (OSSP) utilizing an encoder-decoder architecture with multi-head attention. Trained on Taillard benchmark instances ranging from 4x4 to 10x10, the model generates feasible schedules with makespans within 15-30% of the best-known values and demonstrates competitive performance on larger instances (40x40 to 100x100) with average gaps of 12.89-15.12% relative to a standard lower bound. This approach offers a scalable, learning-based alternative to traditional dispatching heuristics, potentially simplifying the tuning process for practitioners in industrial scheduling applications.