Deep Learning-Based Dynamic Scheduling for Semiconductor Manufacturing with High Uncertainty of Automated Material Handling System Capability

To improve the dynamic responsiveness and production efficiency of manufacturing workshops to personalized orders, a multi-agent manufacturing system. However, the previous methods cannot work well in a changeable workshop environment when encountering a large number of stochastic disturbances of orders and resources. Develop a robust scheduling approach for dynamic workshop environments, effectively managing stochastic disturbances in orders and resources to optimize production efficiency and minimize disruptions. Dynamic scheduling approach for semiconductor manufacturing using Proximal Policy Optimization (PPO) and Autoencoder (AE) for feature extraction, defining state space to capture task and production attributes. Action space design employs dispatching rules for operation selection, while the reward function adapts production to demand, trained in action and state phases. Autoencoders compress high-dimensional input data into a lower-dimensional latent space while preserving important features. The obtained results demonstrate that the proposed method achieves better performance Delivery Data Rate (DDR) of 41.20 %, Mean Cycle Time (MCT) of 51.66%, Mean of 12.85%, and Execution time of 5.07s. These results ensure superior dynamic scheduling performance compared to other existing methods, such as Dynamic Scheduling Using Flexible Job Shop Scheduling problem (FJSP), and Multiple-DynamicScheduling-Rule (MDSR).