Geometry Extraction for Ad Hoc Redirected Walking Using a SLAM Device<bold> </bold>

Redirected walking applications allow a user to explore large virtual environments in a smaller physical space by employing so-called redirection techniques. To further improve the immersion of a virtual experience, path planning algorithms were developed which choose redirection techniques based on the current position and orientation of the user. Furthermore, additional algorithms were developed to guarantee the user's safety at all times. In order to ensure a reliable performance, both safety and planning algorithms depend on accurate position tracking which is commonly provided by an external tracking system. The disadvantage of this kind of tracking is the time-consuming preparation of the physical environment, which renders the system immobile. A possible solution to eliminate this dependency is to replace the external tracking system with a state-of-the-art inside-out tracker based on the concept of Simultaneous Localization and Mapping (SLAM). SLAM algorithms track the position and orientation of a sensor by fusing various measurements and using a continuously generated and expanding map of its surroundings. However, the information contained in this map is not suitable for existing safety and planning algorithms due to its feature-based localization properties. In this paper, we present an approach in which we attach a commercially available SLAM device to a head-mounted display to track the head motion of a user. From sensor recordings of the device, we extract environmental information specifically suitable for existing path planner and safety algorithms of a redirected walking application using a sequence of algorithms. Accordingly, the resulting spatial data consists of 2D coordinates describing closed polygons which are used to approximate the geometry of the walkable area. This geometrical data can then be streamed to the redirected walking application.<bold> </bold>