Effective dynamic scheduling of twin Automated Stacking Cranes (ASCs) is essential for improving the efficiency of automated storage yards. While Deep Reinforcement Learning (DRL) has shown promise in a variety of scheduling problems, the dynamic twin ASCs scheduling problem is challenging owing to its unique attributes, including the dynamic arrival of containers, sequence-dependent setup and potential ASC interference. A novel DRL method is proposed in this paper to minimize the ASC run time and traffic congestion in the yard. Considering the information interference from ineligible containers, dynamic masked self-attention (DMA) is designed to capture the location-related relationship between containers. Additionally, we propose local information complementary attention (LICA) to supplement congestion-related information for decision making. The embeddings grasped by the LICA-DMA neural architecture can effectively represent the system state. Extensive experiments show that the agent can learn high-quality scheduling policies. Compared with rule-based heuristics, the learned policies have significantly better performance with reasonable time costs. The policies also exhibit impressive generalization ability in unseen scenarios with various scales or distributions.
25 February 2025 in IEEE Transactions on Instrumentation and Measurement
Deep learning technology based on large-scale labeled data is increasingly used to diagnose permanent magnet synchronous motor (PMSM) inter-turn short circuit (ITSC) faults. However, the presence of small fault samples context and the large amounts of unlabeled data are challenging problems in real-world industrial applications. To address these challenges, we propose an optimized Lipschitz generative network-gradient spectral generative adversarial network (GSGAN) within the semi-supervised mechanism. Then, we incorporate the graph attention mechanism to achieve accurate fault severity diagnosis. For the specific, firstly, the GSGAN utilizes gradient and parameter spectral information to strictly enforce the Lipschitz condition of the generative network for generating realisticlike ITSC fault data. Secondly, the semi-supervised mechanism enhanced by a temporal causal sparse fusion graph attention neural network (TCSFGAT) is proposed. The TCSFGAT obtains multi-sample level adjustable fault feature information which can clearly represent the graph features in temporal PMSM current fault information. In addition, TCSFGAT can effectively mitigate the impact of pattern confusion caused by mislabeled pseudo-labels in testset diagnostic process. Finally, based on the real industrial environment characterized by high temperatures and strong vibrations, experiments were conducted under 13 specific load and speed conditions to obtain raw fault data. Experimental results indicate the proposed framework achieves a remarkable 98.7% fault diagnosis accuracy.