利用邊緣運算最佳化360度全景影片之頭戴式虛擬實境串流

近年來，360度全景影片蔚為流行。諸多廠商也爭相發表各自的頭戴顯示器。與傳統的平面式螢幕相比，使用頭戴顯示器觀看360度全景影片，提供使用者更好的觀看體驗。然而，為了實現高品質的沈浸式觀看體驗，有諸多挑戰必須一一克服，例如：網路延遲、頻寬消耗等。在這篇論文中，我們提出利用邊緣運算360度全景影片之頭戴式虛擬實境串流系統。此外，我們更設計了一最佳化演算法，此演算法根據不同的使用者採取不同的串流策略，並能有效地降低網路延遲且提高使用者的觀看體驗。為了驗證本系統及演算法的效能，我們也蒐集並公開了一360度全景影片使用者觀看習慣資料庫，利用此資料庫，我們具體量化了利用邊端運緣360度全景影片之頭戴式虛擬實境串流系統的優缺點。與現有的串流系統相比下，我們的系統(一)降低頻寬消耗、(二)提供較好的使用者觀看體驗及(三)降低使用者端的運算資源消耗。此串流系統及使用者觀看習慣資料庫皆已在網路上開源，希望能幫助到產、學界對此領域有興趣的研究者。

關鍵字

360度全景影片；邊緣網路；影音串流；虛擬實境；頭戴式裝置

並列摘要

Over the past years, 360◦ video streaming is getting increasingly popular. Watching these videos with Head-Mounted Displays (HMDs), also known as VR headsets, gives a better immersive experience than using traditional planar monitors. However, several open challenges keep state-of-the-art technology away from the immersive viewing experience, including high bandwidth consumption, long turn around latency, and heterogeneous HMD devices. In this thesis, we propose an edge-assisted 360◦ video streaming system, which leverage edge networks to perform viewport rendering. We formulate the optimization problem to determine which HMD client should be served without overloading the edge devices. We design an algorithm to solve the problem as mentioned earlier, and a real testbed is implemented to prove the concept. The resulting edge-assisted 360◦ video streaming system is evaluated through extensive experiments with an open-sourced 360◦ viewing dataset. With the assistance of edge devices, we can reduce the bandwidth usage and computation workload on HMD devices when serving the viewers. Also, the lower network latency is guaranteed. We also conduct several extensive experiment. The results show that compared to current 360◦ video streaming platforms, like YouTube, our edge-assisted rendering platform can: (i) save up to 62% in bandwidth consumption, (ii) achieve higher viewing video quality at a given bitrate, (iii) reduce the computation workload for those lightweight HMDs. Our proposed system and the viewing dataset are open-sourced and can be leveraged by researchers and engineers to improve the 360◦ video streaming further.