As the need for efficient warehouse logistics has increased in manufacturing systems, the use of automated guided vehicles (AGVs) has also increased to reduce travel time. The AGVs are controlled by a system using laser sensors or floor-embedded wires to transport pallets and their loads. Because such control systems have only predefined palletizing strategies, AGVs may fail to engage incorrectly positioned pallets. In this study, we consider a vision sensor-based method to address this shortcoming by recognizing a pallet’s position. We propose a multi-task deep learning architecture that simultaneously predicts distances and rotation based on images obtained from a visionary sensor. These predictions complement each other in learning, allowing a multi-task model to learn and execute tasks impossible with single-task models. The proposed model can accurately predict the rotation and displacement of the pallets to derive information necessary for the control system. This information can be used to optimize a palletizing strategy. The superiority of the proposed model was verified by an experiment on images of stored pallets that were collected from a visionary sensor attached to an AGV.

Anomaly detection in multichannel sensor data is essential for ensuring safety and reliability in domains such as industrial monitoring, automotive systems, and healthcare. Traditional time-series models often focus solely on temporal patterns, struggling to capture the complex inter-variable relationships among highly correlated sensors. This limitation hinders their effectiveness in detecting anomalies and identifying the key contributing variables. To address this issue, we propose a transformer-based framework that integrates variable encoding, a method for embedding sensor type information, and a variable-guided attention mechanism that emphasizes interactions among variables rather than focusing only on temporal patterns. In contrast to prior transformer-based models, which typically apply self-attention across time steps, the proposed approach models intervariable dependencies directly, enhancing interpretability and robustness. By explicitly modeling sensor types and focusing on inter-variable dependencies, the proposed framework improves both anomaly detection accuracy and interpretability. Experiments on synthetic datasets demonstrate the model’s ability to detect diverse anomaly patterns, including subtle and complex group anomalies. Validation on real-world automotive battery data demonstrates the model’s high performance and robustness, achieving an AUROC of 0.96. Our approach offers a practical and efficient solution for anomaly detection in high-dimensional sensor data, with significant potential for real-time monitoring applications.