Xing Wei
China Telecom Research Institute, China
Dr. Xing Wei is a researcher at the China Telecom Research Institute. He received his Ph.D. degree in signal and information processing from Beihang University, China, in 2025. From 2022 to 2023, he was a guest researcher at the Fraunhofer Heinrich Hertz Institute, Germany. His research interests include immersive communications, generative AI-enabled communications, and AI-native networks. He received Best Paper Awards at the IEEE VTC 2026 Workshop and EMC² 2026. He currently serves as Lead Guest Editor for the collection “Multimodal Generative Intelligence for 6G Immersive Communications” in npj Wireless Technology, a Nature Portfolio journal.
Immersive communication requires coordinated communication and computing resources to deliver high-quality experiences under stringent latency constraints. Proactive streaming predicts users’ movements and prepares content in advance, but its performance depends on coupled decisions on rendering location, encoding rate, bandwidth, computing resources, and rendering and transmission durations. This talk presents a structure-guided learning framework for joint resource allocation across devices, edge servers, and the cloud. We first exploit analytical solutions to reduce the decision space, then embed user symmetry and constraint satisfaction into neural architectures to produce feasible allocation policies efficiently. Evaluations using measured visual quality demonstrate improved performance over numerical and learning-based baselines, generalization across user numbers, and millisecond-level online inference. As an extension, we discuss how privacy protection reshapes rendering-location choices and resource allocation, and how the framework accommodates the resulting tradeoffs between privacy and visual quality.