無線混合現實(MR)系統，需要提供使用者及時且準確的與虛擬物件互動 的沉浸式體驗。為了在有限的硬體計算資源下達成此目標，現有的方式 為將計算轉移至邊緣伺服器，並藉由預測的方式減緩MTP延遲所造成看 到的畫面不是當下畫面的情況發生，提前渲染視野的影像進行傳輸。由 於預測不是無錯誤的，因此可能出現物件顯示位置錯誤的問題，而導致 使用者無法準確與物件進行互動。為了解決這個問題，我們提出了使用 本地端的物件精準位置做為定位輔助的串流架構並且開發了基於所提出 架構的串流資料優化算法，避免多餘的資料傳輸。實驗結果顯示，在相 同的MTP延遲下，所提出的串流架構在虛擬物件定位和網路資料量傳輸 方面都優於現有的方法。;Wireless mixed reality (MR) systems must provide users with an immersive experience of interacting with virtual objects in real-time and accurately. In order to achieve this goal under limited hardware computing resources. The existing method is to transfer the calculation to the edge server and use the field of view (FoV) prediction method to reduce the situation that the image seen is not the current image caused by the Motion-to-Photon (MTP) latency and render the relative video in advance for transmission. Since predictions are not error-free, objects may be displayed in the wrong position, preventing users from interacting with objects inaccurately. To solve this problem, we propose a streaming architecture that uses the absolute position of objects on the MR devices as a positioning aid. Then, we develop streaming data optimization algorithms based on the proposed architecture to avoid redundant data transmission. Experimental results show that under the same MTP latency, the proposed streaming architecture outperforms existing methods in both virtual object location and network data volume transmission.
