以同儕架構為基礎的3D網路虛擬環境架構是一個嶄新的研究議題，使用同儕架構來取代目前的主從式架構，除了中央伺服器外，使用者同時也可從其他參與者（或稱同儕）取得所需虛擬環境資料，不僅大幅減少中央伺服器的資源消耗，提升系統的擴充性，讓系統在使用者人數不斷增加時亦可正常運作。為了減少將完整虛擬環境資料從網路下載而花費的等待時間，3D技術讓使用者僅須漸進式的下載本身所處位置週遭一定範圍內的虛擬環境場景資料，即可在虛擬環境中遊覽並與其他使用者或與虛擬環境進行即時性的互動；而在使用者瀏覽虛擬環境的時候，3D串流技術也同時在下載未來即時性互動所需要的場景資料。在虛擬環境中，使用者會因為本身所在位置的不同，能夠提供資料的來源（同儕）也因此有所差異，如此一來，如何發現新的資料來源並由最適當的資料來源下載資料（統稱同儕選擇）將是影響3D串流效能的一個重要問題。 本論文針對同儕架構網路虛擬環境的3D串流方式，提出一個新的同儕選擇策略。透過使用者間主動交換新增資料訊息，加快找尋資料提供者的速度；而利用多層次請求範圍，減少特定區域內資料提供者接收過多請求而造成的資源競爭問題，以縮短資料請求完成時間。實驗模擬結果顯示出我們所提出的同儕選擇方法，的確可以比過去的相關方法達到更高的系統擴充性及效能。 The peer-to-peer networked virtual environment (P2P-NVE) is a new research direction. Different from client-server architecture, peer-to-peer architecture allows users to acquire desired 3D virtual environment data from not only the central server but also the other network participants. In order to enable immediate interactions between the users and the virtual environments, 3D streaming techniques fragment object data into several pieces for users to receive those pieces progressively. While acquiring object pieces, users can render the objects at different level of details without a full download. However, even though the desired data can be acquired from other peers, not all peers can be the suppliers as some may be too far from the objects to possess the object pieces. Some peer selection strategies thus have been proposed; however, they have some drawbacks in terms of both latency and message overhead. In this thesis, we propose new peer selection strategies of 3D streaming for P2P-NVEs. Peers periodically exchange incremental content availability information with neighbors, which helps requestors to be aware of suppliers quickly, saving both message overhead and time. We also adopt multi-level request areas to reduce resource contention. We perform simulation experiments for the proposed strategies. Simulation results show that our selection strategies perform better than related ones in terms of system scalability, request hit ratio, latency and data fill ratio of data.
