POSTER
Lingzhi Zhao, Yongqiang Gui, Yanyan Suo, Sandesh Dhawaskar Sathyanarayana, Mingyuan Wu, Bo Chen, Ruixiao Zhang, Klara Nahrstedt, Shu Shi
In this poster, we present MARS, a practical rate adaptation approach for real-time VR streaming. Delivering VR content over wireless networks costs much more bandwidth than traditional video streaming. Prior work primarily reduces the cost by exploiting users’ spatial viewing patterns by prioritizing the visible region or streaming only the current viewport. MARS reduces the bandwidth consumption from a different perspective by leveraging users’ temporal viewing patterns, inspired by the observation that users’ dynamic visual acuity decreases during fast head motion. To this end, MARS dynamically adjusts the bitrate of VR frames based on viewers’ head motions. However, realizing this is non-trivial and has two key challenges: (1) the head motion dynamics and bitrate adaptation should be precisely synchronized; (2) the bitrate adaptation should strike a careful balance between the bandwidth reduction and visual quality degradation. To address the challenges, MARS first proposes a kinematics-based synchronization mechanism to capture the viewers’ motion phases for timely bitrate adaptation. It further employed a perceptually-guided bitrate adaptation algorithm to wisely reduce the bitrate by exploiting viewers’ visual acuity. MARS provides a plug-and-play solution that can be readily integrated into existing VR streaming platforms. We prototype MARS and evaluate it through both real-world user studies and benchmarking. Results show that MARS delivers user experience comparable to a fixed-bitrate baseline while reducing bandwidth consumption by 10%-19%.