Cloud Computation of VR/AR Applications
Cloud computation has had an explosion of popularity due to its ease of use, low cost of entry, and ability to avoid in-house overhead of running a local server. The reintroduction of consumer VR is facing the same issue that stifled VR development in the 1990’s, a lack of affordable computation. VR headset costs have fallen dramatically since their reintroduction in 2012, yet the cost of a capable computation pushes consumers away from the product. VR has many interesting usages in health care, military, gaming and more. By reducing the need for a capable computer, VR becomes a more accessible medium. I intend to investigate the feasibility of cloud computation for VR applications and the required underlying network it would need to utilize.
​
VR cloud computation is currently an area of active research with Huawei leading an initiative to consolidate VR cloud computation research [1]. Current research includes utilization of the developing 5G network to provide the required data throughput to render a full VR system [2]. Other suggestions by Xueshi Hau et al. [3] include three different recommendations for cloud computing a VR system under our current network framework:
-
“Render the field of view (FOV) (current view) remotely and stream the corresponding video to the user's glasses”
-
“Render multiple views remotely, stitch to a 360-degree video, and stream the video to the user's glasses”
-
“Compress model as well as texture, stream them to the user's local edge device at first, then render at the local edge device and stream the video to the user's glasses.”
​
Additionally, there are all-in-one computation VR systems such as the Oculus Quest. However, these all-in-one devices are not able to produce the quality of content other VR systems are capable of. For this reason, all-in-one systems will be ignored except in the case of using an all-in-one to accept cloud computed resources as suggested earlier [2].
​
For this project, I would like to further the ideas of Hau et al. Specifically, their third idea, which involves a combination of cloud computation and local computation to display the VR environment. The proposal I would like to make involves a baseline rendering of an entire scene to be done in the cloud with final rendering details of the users FOV to be done by a local system. VR rendering done in this manner should greatly reduce the requirements of the local user’s system. One of the problems needing further investigation is required throughputs of a network for a seamless experience. An additional area of investigation is backup handlers in the case of a severed or hindered network to the cloud servers.