Mobility for high-quality gaming and XR applications with real-time rendered graphics is challenging to achieve as the processing power of mobile devices is an order of magnitude weaker than PC setups. This problem can be overcome by offloading the graphics rendering to remote servers in the cloud. The main challenge in the remote rendering of graphics is the added latency. The goal of this project is to replace the traditional remote rendering pipeline which uses video codecs designed for non-interactive video and separate latency compensation steps decoupled from video encoding with a deep-learning based encoder/decoder pipeline with native built-in support for delayed presentation including pose prediction and view synthesis. We aim to completely hide the latency so that the user cannot perceive the difference between a locally and remotely executed application.

2023

2027