Emerging applications such as autonomous driving, AR/VR and intelligent manufacturing will deploy sensors to collect status information. To improve the quality of service in such autonomous systems, awareness of the semantic information of the state needs to be as timely as possible. To enhance decision-making by capturing crucial semantic data, we need a robust automated semantic networking framework to guide the design of data sensing, scheduling and routing. In this talk, I will introduce our research on timely information sensing, scheduling and information retrieval, including: (1) An online algorithm for improving information timeliness in semantic data sensing (2) A provable cross-layer control for semantic information scheduling (3) Semantic data retrieval with diffusion models. Our findings highlight the importance of opportunistic network resource allocation that leverages data semantic importance and dynamic channel conditions to optimize information timeliness in wireless networks. Finally, I will discuss my research vision on opportunistic networking algorithms that exploit channel propagation characteristics and traffic fluctuations in next-generation wireless networks, and opportunistic learning for network parameter configurations.

Haoyue Tang currently a postdoctoral associate at Meta AI. Prior to that, she was a postdoc in the Institute of Network Science, Yale University working with Prof. Leandros Tassiulas. She received her PhD and Bachelor degrees in the Department of Electronic Engineering, Tsinghua University. During her PhD studies, she was a visiting student in Télécom Paris, Microsoft Research Asia and Tencent Inc. She received the excellent graduate of Beijing award, a selected participant in Rising Stars of EECS 2022 workshop, National Scholarship and Stars of Tomorrow Excellent Intern from Microsoft. Her research interests include stochastic optimization and information theory, with applications in wireless networking and generative models.