A Performance-oriented Network Component for Massive Multiplayer Virtual Reality Viewing Experiences

Closed Computer Science Master Thesis No Affiliation

Motivation and Goals

When watching any event, whether it is in a stadium or in a public viewing setting on a big screen, a big part of the experience is the crowd. The reactions and dynamics in large groups of people make the event much more exciting and contribute highly to the enjoyment of the viewing experience. The general atmosphere is the reason why people buy tickets to events or gather in public places rather than watching it from the comfort of their own homes. A social Virtual Reality (VR) viewing experience should allow for these factors to be present while removing difficulties that exist when attending the events physically, such as sold out tickets or travel expenses. It also keeps in check group dynamics and the anonymity felt by individuals in large groups that can lead to undesirable and even violent behaviour¹. Due to the people not being physically present and the possibility to allow users to ignore others or even just teleport away, these issues can be either kept in check or solved completely. However, the synchronization of a large number of user avatars and their actions over a network poses a challenge that has to be solved in order to allow the atmosphere of a public viewing experience to build in a virtual environment ². This is why this work aims to implement a network component which allows for synchronization of a large number player avatars over the internet using a client-server architecture. For this work only simple avatars will be used, however more complex player representations should be usable in the future without a large decline in performance. In combination with social interactions such as voice chat or gestures, allowing a large number of users in the same room enables positive group dynamics ³ and an atmosphere existent in the real world, thus massively improving the virtual viewing experience.

Background and Related Work

Social Virtual Reality

Social Virtual Reality is a term for multi-user VR applications that allow for social interactions between them. This includes multiplayer video games ⁴ as well as educational ² or therapeutical software ⁵ which feature client-server or peer-to-peer architectures to synchronize the environment between users. Due to the requirement of multiple users being present in these applications, in addition to the standard requirements of VR experiences such as presence, immersion and graphical quality there arises the issue of synchronization of the world as well as the avatars representing each user in the virtual environment². Scalability can provide an enormous challenge, especially when wanting to support enough people to allow for crowd dynamics to develop. The most common architectures for network communication currently are peer-to-peer and client-server architectures ⁶. Currently the leading Massive Multi-Player Games use client-server architectures to handle most of the load on their own servers, keeping the computational requirements relatively low for their users,

Crowd Behaviour

Due to dynamics that are present when people form a crowd or collective ⁷, the element of a large number of people being present is an essential part of recreating the experience of a real world stadium or public viewing in Virtual Reality. People in a crowd benefit from positive effects such as increased personal as well as collective self-esteem and more positive social beliefs, however possibly the biggest addition to the enjoyment of a public viewing experience are stronger emotional reactions ³. This leads to people feeling more involved and more invested in the event they are attending. However, crowds can also be a destructible force ¹, leading to vandalism and even death in the past. The hope is to be able to reproduce the positive effects of crowd behaviour while negating the negatives, with most of them being a non-issue due to the virtual aspect of the environment.

Planned Methods

The goal is to implement a client-server structured network component for a Social VR application that allows for a large amount of users (100+) to be present in the same environment without a large decline of performance. Aside from movements, actions, movable objects and possibly looks of avatars, everything else shall be done on the client systems themselves to limit the required network traffic.

As a framework the Unreal Engine 4 is chosen, which allows for a relatively simple implementation of Virtual Reality applications and can be extended with the use of plugins. The implementation will be done in C++, with minimal use of the Unreal Engine Blueprint scripting language when necessary or does not impact performance significantly.

For evaluation the frame rate of the application will be measured, the latency of actions performed by users with which they appear on screen for other users as well as the amount of network traffic produced by the application.

In addition, the impact of the networks improvements on the user experience can be tested. This can be done via a user study and could serve as an indicator as to how important performance optimization is in this aspect. If done with several stages of optimization, such a study could also yield target values for i.e. latency which should be hit by applications to appear as fluent to the users.

Work Plan

References

Contact Persons at the University Würzburg