面向视图一致性的虚拟角色划分方法

摘要

In large-scale distributed virtual environments (DVE) adopting multiple-server architecture, to manage mobile avatars effectively, the servers vere needed to response to avatars' queries in a real-time manner and reduce the view inconsistency among virtual avatars efficiently. By considering the influences of network communication latency and avatars' moving behaviour on view inconsistent events in DVE environments, the moving area of interest (MAOI) holding by avatars in which they may perform querying was analyzed, and then a method to partition avatars into clusters based spatial density-based clustering was proposed. According to the distances between avatars and the gravitational centre of each cluster, a repartitioning algorithm was also designed to refine the avatar groups so as to preserve load balancing among the servers and reduce the view inconsistent events. Finally, extensive experiments were conducted to evaluate the features of our approach.%在基于多服务器架构的大规模分布式虚拟环境中,为有效管理大量的移动虚拟角色,需要服务器能及时响应虚拟角色的各类请求,并能有效减少虚拟角色之间的视图不一致性事件.综合考虑网络通信时延、虚拟角色的移动行为对视图不一致性的影响,分析了虚拟角色可能进行信息查询的移动兴趣区域,然后基于空间密度聚类方法来实现虚拟环境中虚拟角色间的聚类；根据聚类的大小和服务器负载能力给出了基于聚类几何重心的划分方法,从而达到减少虚拟环境中视图不一致性事件并保持服务器间负载均衡的目标.通过系统性实验评估了所提方法的特点.