Impossible Open Spaces
Exploring the Effects of Occlusion on Noticeability of Self-Overlapping Virtual Environments
6th-semester bachelor's thesis project at Aalborg University Copenhagen. We continued our VR locomotion research, following our success in the Mission Impossible Spaces project. Our curiosity led us to explore the use of overlapping architecture in outdoor environments. The technique was originally built for indoor environments, but we reconed it can be used in structured outdoor environments.
The challenge of using overlapping architecture in outdoor environments is the different levels of visibility. Indoor environments perfectly occlude other spaces, like the joining rooms and the walkway. But for outdoor environments, these could be visible, therefore, making it harder to hide the impossible spaces.
We designed the outside environment as an old ruin structure, just like you would stumble upon on your hike. The experiment had 3 occlusion levels: total, only rooms, and only walkway. The only occluded element we kept in all the conditions was the wall between the rooms and the walkway, because without it, we could not hide the impossible spaces.
The biggest challenge in this project was designing the system that controls the overlap between the two rooms. Originally, overlapping architecture shifts the connecting wall to increase space. But if the next room would visibly shrink if not occluded.
We implemented our own overlapping architecture, making the adjacent room appear as large as the room the user is currently exploring.
Therefore, we needed to extend the corridor walls to accommodate the room expansion. So it is mandatory to hide the manipulation within a bigger structure.
We evaluated the 3 occlusion conditions using two metrics. First, detection thresholds: how much overlap is possible in each condition. Second, the distance participants report walking between the rooms, which reveals whether they simply saw larger spaces or actually experienced them. Overlaps of 0 to 75 percent were included in this study.
An open corridor did not affect detection thresholds, but it dropped significantly in open rooms condition. There was no obvious difference in distance estimation across conditions; in all conditions, greater overlap was associated with greater overestimation. We concluded that users judge the impossible spaces based on what they visually observe rather than walking distance. The open rooms method worked, but it was more difficult to implement and had a higher detection chance. Therefore, showing both rooms should be reserved for necessity and applied using lower overlap levels.