Dynamic impossible spaces in VR : mapping of virtual and real obstacles

Abstract

This paper presents and implements an approach for mapping real-world obstacles into virtual reality (VR) environments. Physical objects are detected in real time using the Meta Quest 3's passthrough mode and room scanning, then dynamically mapped to virtual obstacles. To enable larger virtual worlds, procedurally generated impossible spaces are integrated via a graph-based depth-first search algorithm. A user study with 16 participants compared three scenarios: optimized obstacle placement, random placement, and no obstacles, evaluating presence, spatial presence, realism, involvement, and fear. Optimized placement outperformed random placement in all categories and did not perform worse than no-obstacle scenarios. Compared to a similar study without impossible spaces, presence and spatial presence improved further. Realism and involvement decreased only in the random condition. Overall, impossible spaces enhanced the VR experience when obstacles were placed optimally.