结合社交推荐和推拉策略的渐进式DVE预下载机制

Progressive DVE Prefetching Mechanism Using Social Recommendation and Pull-Push Strategy

为了提高基于对等网络的分布式虚拟环境(DVE)场景预下载的预测精度并降低场景数据传输延迟,将社交推荐和推拉混合策略应用到DVE场景预下载中.首先描述和量化节点化身在虚拟环境中的兴趣,并对化身间相似度进行计算,生成预推荐场景集;然后将化身的视野区域划分为预拉取和预推送区域,提出针对不同区域预下载的推拉策略.基于开源DVE平台FLo D进行的实验结果表明,该机制能提高场景的预测精度并减少信息交互次数,进而提升DVE场景传输效率和化身漫游流畅度.

In order to improve prediction precision of the scene prefetching and reduce the scene transmis- sion latency in the distributed virtual environment （DVE） based peer to peer network, we propose an effi- cient progressive prefetching mechanism, which combines the social recommendation with different Pull-Push strategies. First, we describe and quantify the avatar interests in the virtual environment to obtain the interest similarity between avatars and generate pre-recommend scene set. Then, we divide the avatar viewing area into the pull area and the push area, and provide different Pull-Push strategies for prefetching the scene in different area. Finally, we demonstrate the proposed prefetching mechanism using an open-source DVE platform called FLoD. Experiment results show that our approach improves the scene prediction precision, reduces the information exchange, and promotes the scene transmission efficiency and avatar roaming fluency.