Virtual reality (VR) streaming is a new tool to deliver remotely rendered content over the network to an HMD from a PC, while relying heavily on adaptive bitrate (ABR) algorithms in order to try to maximize the user's Quality of Experience (QoE). ABR algorithms work by dynamically adjusting the bitrate of the stream, striving to provide an optimal immersive experience tailored to the available network capacity.

Nevertheless, the dynamic nature of network conditions —including bandwidth fluctuations, latency variations and random packet losses— makes this a challenging task. Due to this and the ever increasing popularity of commercial streaming platforms, the field of improving and designing new ABR algorithms has been gaining traction in the literature. 

Currently, UPF-WN is actively engaged in the development of novel ABR algorithms within ALVR, an open-source solution that enables streaming VR content from a PC to an untethered head-mounted display (HMD). This allows users to play more graphically demanding VR games through wireless streaming, where otherwise they would be limited to running less taxing games directly on the HMD (as the Meta Quest does); or using a data cable from the PC to the HMD, which can be straight up uncomfortable. 

As a preliminary step, ALVR has been enhanced with additional network performance metrics —proven to be informative about the network state— for later use as inputs into the ABR algorithm's decision-making process. These new statistics include classical ones such as packet loss and jitter, but also extend to metrics specific to the delivery timing of video packets and the stream performance. In addition, existing metrics such as the network throughput have undergone redefinition and validation to be more accurate and reflective of the traffic received by the HMD. Watch the demo! 

The next step involves the design, implementation, and evaluation of effective VR-centric ABR algorithms, optimizing the bitrate selection process by leveraging the insights gained from the network performance metrics integrated into ALVR.