在這項研究中，我們設計了一款可客製化問題的虛擬實境遊戲。老師端使用預先設計的問題模板，可以根據問題的主題打造屬於這問題的答題情境。透過360環景圖片及影片，使學生更能體會問題背後的原因。未來360∘串流服務應用在虛擬實境上的需求如雨後春筍。許多影片被暫存在內容傳遞網路或多接取邊緣運算。然而，串流一個360影片消耗了許多不被需要的頻寬和暫存資源。我們定義了這個問題並設計一套演算法最佳化暫存資源與使用者視窗覆蓋率根據Tile-based串流架構。與Least Recently Used (LRU) 和Least Frequently Used (LFU) 演算法比較下，我們的演算法在暫存資源與使用者視窗覆蓋率減低了25%的暫存資源及擴展了30%視窗覆蓋率。;Virtual Reality (VR) and Augment Reality (AR) has been provisioned as a killer-application for next generation. Especially in education, lots of VR application was be design for promoting learning effect with virtual environment and 360° video. With a numerous VR application, 360° video streaming may not achieve high quality of experience because of the strict requirement of end-to-end latency and the bandwidth of network. Rely on tile-based 360° video streaming and the caching capacity in MEC, Service predict the Field of View in Head-Mounted Device then deliver require tile in MEC. Prefetching tile in MEC can save the bandwidth of backend link and support multi-users. The smart caching decision may reduce the memory at the edge and make up the prediction error of FoV. For example, caching whole tile at each small cell cause higher storage cost compared to caching at one small which cover multi-users. In this paper, we define tile selection, caching and FoV coverage model as Tile Selection and Caching Problem. We propose a heuristic algorithm. Adopting the dataset of real users’ head movement, we compare our algorithm to Least Recently Used (LRU) and Least Frequently Used (LFU) caching policy. The result show that our proposed approach improves 30% of FoV coverage and reduce 10% caching cost than LFU and LRU.
