Public defence in Computer Science, M.Sc. (Tech) Gazi Illahi

Title of the thesis: On Improving QoE of Remote Rendered Graphics

Doctoral student: Gazi Illahi
Opponent: Professor Cheng-Hsin Hsu, National Tsing Hua University, Taiwan
Custos: Professor Antti Ylä-Jääski, Aalto University School of Science, Department of Computer Science

Remote rendering refers to offloading rendering tasks of an application to remote, cloud or edge-based server and streaming the results to a client, for example, as a video stream. Remote rendering can enable complex graphical applications to be made available on resource constrained devices which otherwise would not be able to run such complex applications. Remote rendering, however, comes with its own constraints. Rendering high quality complex graphics produces high bitrate video which needs considerable downstream bandwidth to be streamed to a client. Further, the latency from the client to the remote rendering service has to be low, especially for interactive applications like cloud gaming and especially so for immersive applications like cloud XR.

In this thesis, avenues to improve the end user Quality of Experience (QoE) of remote rendered graphics, for example as used in Cloud Gaming and Cloud XR are investigated. Recent developments in fields like computer graphics, psychophysics and machine learning are utilized towards that end. To reduce the downstream bandwidth requirements of applications like cloud gaming and cloud XR, the thesis proposes and implements optimizations based on human visual system for rendering graphics and encoding the rendered graphics. Specifically, gaze based foveated rendering and foveated encoding is studied. To mitigate latency requirements of interactive remote rendered graphics applications, machine learning based solutions are presented to predict future user gaze and pose. Gaze and pose are used during the rendering process and their accurate prediction may help in reducing latency induced reduction in drop of end user quality of experience.
The thesis provides evidence that such optimizations are feasible to be used in real time and may be deployed in remote rendered graphics to improve the QoE of end users. 

Keywords: Remote Rendering, Cloud Gaming, Cloud XR, Cloud VR, Foveated Rendering, Foveated Encoding, Gaze Prediction in XR, Pose Prediction in XR

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/ 

Contact information:

Doctoral theses of the School of Science: https://aaltodoc.aalto.fi/handle/123456789/52