Keynote Speaker

Title: Key Technologies and Prototype Design for Integrated Sensing and Communications System

Abstract: In the future, millions of base stations (BSs) and billions of users (UEs) will natively build an integrated sensing and communications (ISAC) system, which can utilize intelligent ubiquitous methods to realize the ultimate goal of sensing, i.e., constructing the global mapping from real physical world to digital twin world, while providing communications services at the same time. For this purpose, we conduct a series of theoretical and technical researches on ISAC, in which we decompose the real physical world into static environment, dynamic targets, and various object materials. The ubiquitous static environment occupies the vast majority of the physical world, for which we design static environment reconstruction (SER) scheme to obtain the layout and point cloud information of static buildings. The dynamic targets floating in static environments create the spatiotemporal transition of the physical world, for which we design comprehensive dynamic target sensing (DTS) scheme to detect, estimate, track, image and recognize the dynamic targets in real-time. The object materials enrich the electromagnetic laws of the physical world, for which we develop object material recognition (OMR) scheme to estimate the electromagnetic coefficient of the objects. Finally, based on these theoretical researches, we build an ISAC hardware prototype platform working in millimeter wave frequency band, realizing high-precision SER, DTS, and basic OMR, which provides preliminary verification for building the digital twin for communications networks.

Biography

Prof. Feifei Gao
IEEE Fellow
Tsinghua University, China

Biography

Prof. Yudong Zhang
IET/EAI/Fellow, IEEE/ACM Senior Member
Southeast University, China

Biography

Prof. Yoshinori Dobashi
Hokkaido University, Japan

Title: Efficient Visual Content Generation

Abstract: To address the issues of high energy consumption and carbon emissions caused by the high computational complexity of visual content generation models, this report investigates neural network compression and stable training methods for image generation models from three perspectives: architectural design, inference steps, and parameter quantization, with the goal of enabling green, low‑carbon model lightweighting technologies. Specifically, the work includes: (1) Optimizing network architectures to reduce the number of model parameters or the amount of parameter activations; (2) Designing single-step algorithms for the inference phase of generative models to reduce inference latency; (3) Applying low-bit quantization to models to further shrink model size while maintaining stable training. These three lines of research are mutually reinforcing and are expected to form a comprehensive lightweighting solution for large models targeting edge-side deployment needs, achieving an excellent balance between resource consumption and model performance.

Biography

Prof. Nannan Wang
Xidian University, China