Exécution de notre outil de synthèse sur un nouvel article

Résumé : Accurate network latency prediction is essential for Ultra-Reliable and Low-Latency Communication (URLLC) in 5G networks, enabling mission-critical applications such as the Industrial Internet of Things (IIoT) and the Internet of Vehicles (IoV). Existing statistical and deep learning methods, including RNNs, LSTMs, and Transformer-based architectures, often struggle to capture the complex, non-linear dependencies in multivariate network data, while incurring high computational and memory costs. In this paper, we investigate the application of TimeXer to the specific task of network latency forecasting, and we propose a practical adaptation workflow for incorporating TimeXer into latency prediction pipelines. To validate its practical effectiveness, we collected a real-world 5G network latency dataset under operational URLLC scenarios that captures diverse network conditions and transmission behaviors. Extensive experiments show that applying TimeXer with our adaptation achieves consistent improvement over competing methods, yielding a mean squared error (MSE) reduction of 9.3% compared with PatchTST and 13.8% compared with TS2Vec, while maintaining manageable computational overhead. These results demonstrate that TimeXer, when properly adapted, is a robust and effective option for real-world 5G latency prediction tasks.