NeSt-VR: Adaptive Bitrate Algorithm for Virtual Reality Streaming over Wi-Fi

This dataset contains results from streaming VR content over Wi-Fi 6 using our Air Light VR (ALVR) v20.6.0 fork. In particular, it comprises ALVR session logs with statistics in JSON format for each test in Sections VI and VII of our published paper, NeSt-VR: Adaptive Bitrate Algorithm for Virtual Reality Streaming over Wi-Fi. Additionally, for each test in Section VI, it includes tshark-processed traffic traces in space-separated CSV format, collected using Wireshark v4.0.3 at both the server and the network emulator’s Ethernet interface to the access point. Moreover, for each test in Section VI, validation result figures are included. For each test in Section VII, temporal evolution and/or boxplot figures for several Quality of Service metrics—such as delivery frame rate, bitrate, video frame round-trip time, and packet loss—are also included.

Section VI tests use a Constant BitRate (CBR) of 100 Mbps with several emulated network effects, including limited bandwidth (100 Mbps, 95 Mbps, 90 Mbps), packet loss (0.5%, 1%, 2%), duplicated packets (0.5%, 1%, 2%), and packet jitter (0–6 ms, 0–10 ms, 0–20 ms).

The dataset structure for Section VII includes a folder for each subsection (VII A: 7.1, VII B: 7.2, VII C: 7.3, VII D: 7.4). Section 7.1 folder includes tests on emulated limited network bandwidth (100 Mbps, 95 Mbps, 90 Mbps) using either CBR, ALVR's native Adaptive BitRate (ABR) algorithm, or our VR-tailored ABR, NeSt-VR (Network-aware Step-wise ABR algorithm for VR streaming). Section 7.2 folder contains a single-user (user A) mobility test using either CBR or NeSt-VR. Section 7.3 folder includes a multi-user test with two users (user A and user B) using either CBR or NeSt-VR, with results for both users streaming in isolation or concurrently. Section 7.4 folder contains tests with Overlapping Basic Service Set (OBSS) activity, where two access points operate on the same frequency channel with overlapping coverage areas, using either a fully overlapping channel bandwidth of 40 MHz or 80 MHz.

ALVR session logs contain several built-in ALVR statistics (event_type:{"id":"GraphStatistics", which includes total pipeline latency and its components) and additional statistics incorporated in our ALVR fork (event_type:{"id":"GraphNetworkStatistics", which records metrics such as frame span, frame interarrival, video frame round-trip time, packet loss, instantaneous video network throughput, peak network throughput, video frame jitter, video packet jitter, and filtered one-way delay; event_type:{"id":"HeuristicStats", which includes the decision-making statistics involved in each NeSt-VR bitrate adjustment interval). Please refer to our published paper or our ALVR fork for more details.

Tshark-processed traffic traces contain several packet-level details: the relative timestamp (frame.time_relative), source and destination IP addresses (ip.src, ip.dst), total packet length including headers and payload (frame.len), and the raw packet payload (data.data). The first 22 bytes of each packet’s payload contain ALVR’s application-specific prefix, which includes the associated frame’s payload size in bytes (4 bytes), a stream identifier (2 bytes), the frame index (4 bytes), the number of packets composing the frame (4 bytes), the packet index within the frame (4 bytes), and the packet’s relative departure time (4 bytes).