中央研究院 資訊科學研究所

Immersive cloud VR gaming offloads computationally intensive VR games to cloud servers to reduce the workload on the resource-constrained Head-Mounted Displays. Delivering a high-quality cloud gaming experience is challenging due to the complex human perception systems, heterogeneous and dynamic best-effort networks, and high expectations from VR gamers. In this talk, we first present the dynamic foveation techniques in cloud VR gaming, which significantly enhances the gamer Quality of Experience (QoE) by devoting more resources to the foveal region of HMD viewports, resulting in improved visual fidelity and bandwidth savings. This approach increases the Mean Opinion Score (MOS) and reduces latency, offering a more immersive and responsive gaming experience. Next, we address the intricacies of network conditions affecting cloud VR gaming QoE. By building a comprehensive testbed and conducting user studies, we develop prediction models that dynamically adapt encoding settings to improve gamer QoE in visual and interaction quality. In these two studies, cloud VR servers stream viewports encoded with traditional 2D video codecs to HMD clients. Modern 3D representations could be used in future cloud VR gaming for better interaction modes. In the last part of my talk, I share how we build a novel error concealment pipeline for dynamic 3D point clouds. This framework significantly enhances the quality of streamed content over unpredictable network conditions, ensuring a seamless and immersive user experience. 

Cheng-Hsin Hsu received M.S./B.S. degrees from National Chung-Cheng University, an M.Eng. degree from the University of Maryland, and a Ph.D. degree from Simon Fraser University. He joined the Department of Computer Science, National Tsing Hua University in 2011 as an Assistant Professor. He was promoted to an Associate Professor in 2014 and to a full Professor in 2021. Before accepting the teaching position, he was with the Deutsche Telekom Laboratory (California), Motorola Inc. (Illinois), and Lucent Technologies (Maryland) for more than six years. Cheng-Hsin has been a Visiting Scholar with the University of California Irvine (Summer 2013, Summer 2018-Summer 2019), the Qatar Computing Research Institute (Summer 2014), the University of Illinois Urbana-Champaign (Spring 2016), and the University of California Irvine (Fall 2018 and Spring 2019).

Cheng-Hsin’s research interests lie in multimedia networking, immersive videos, Internet-of-Things (IoT), smart cities, digital twins, and computer networks. He helped to organize international conferences in various capacities, such as Keynote/DEP Co-chair at IEEE NOMS’24, Publication Co-chair at IEEE SMARTCOMP’23, General Co-chair at ACM MMSys’21, Publicity Co-chair at ACM MMSys’18, Area Co-chair at ACM Multimedia’17, and TPC Co-chair at ACM MMSys’17. He has been an Associate Editor of ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM) since 2014.

He and his colleagues received the best paper awards at IEEE SMARTCOMP’20, IEEE CloudCom’17, APNOMS’16, IEEE RTAS’12, and IEEE Innovation’08, as well as the TAOS best paper award at IEEE GLOBECOM’12 and the Best Demo Award from ACM Multimedia’08. Cheng-Hsin was selected as one of the Multimedia Rising Stars by the ACM SIGMM in 2015. He won the Best Associate Editor Award from TOMM in 2016 and an Outstanding Scholar Award for Retaining Domestic Scholars from the Foundations for the Advancement of Outstanding Scholarship between 2018 and 2023.