An End-to-End Deep Learning Method for Dynamic Job Shop Scheduling Problem

Job shop scheduling problem (JSSP) is essential in the production, which can significantly improve production efficiency. Dynamic events such as machine breakdown and job rework frequently occur smart manufacturing, making dynamic (DJSSP) methods urgently needed. Existing rule-based meta-heuristic cannot cope with DJSSPs of different sizes real time. This paper proposes an end-to-end transformer-based deep learning method named spatial pyramid pooling-based transformer (SPP-Transformer), shows strong generalizability be applied to different-sized DJSSPs. The feature extraction module extracts environment features that are further compressed into fixed-length vectors by compression module. Then, action selection selects simple priority rule experimental results show makespan SPP-Transformer 11.67% smaller than average dispatching rules, methods, RL proving realizes effective without training models for To best our knowledge, first application DJSSP, not only improves productivity industrial but also provides a paradigm future research on DJSSP.