头盔式虚拟现实系统应用越来越广泛。在该类系统中，用户可通过自身在物理空间中走动的自然交互方式实现在虚拟环境中的漫游。这种交互方式需要解决的两个重要问题是：虚实空间的非等距映射问题和多用户实时位置跟踪中的长时间、完全遮挡问题。针对这两个问题，本项目拟结合心理学、生理计算、计算几何、机器学习、视听技术等方面的理论和技术，进行学科交叉研究：1）开展人体对虚拟空间的感知特点研究和影响因素；2）基于上述理论，设计新的重定向行走技术，实现更加有效的虚实空间映射算法，支持多用户通过在物理空间中自由走动的自然交互方式实现在虚拟环境中的漫游；3）融合用户的影子信息，基于机器学习方法实现基于单个RGB-D相机的多用户实时位置跟踪算法，解决长时间、完全遮挡问题。通过上述理论、算法的研究，支持更好地设计头盔式VR系统，使其支持用户通过在有限的物理空间中自由移动来实现在大规模场景中的漫游，

There are more and more applications of virtual reality system based on head-mounted display. Users can move through virtual environments by real walking. There are two important issues that need to be resolved for this kind of interactive method：non-equidistant mapping between the virtual space and physical space, and capturing occluded bodies for multi-person tracking with one camera. Consequently, we will research these issues based on computational geometry, psychology, physical computing and machine learning：1) research on the characteristics of human perception of virtual space when real walking in a physical space; 2) Based on the above results, design effective redirection walking method to resolved the problem of non-equidistant mapping between the virtual space and physical space, so that multi-users can move through virtual environments by real walking; 3) According to user’s shadow information, present a method of tracking multi-users under occlusion with a single RGB-D camera based on machine learning. The above results can be used to effectively design virtual reality systems based on head-mounted display, in which users can move through a large virtual environments by real walking in a limited physical space and get a better experience.