Making virtual reality "more real" and the perception of potential collisions

Abstract

Updated) Purpose: To see if making the experience in virtual reality closer to the “real world” experience (e.g. actually walking, rather than standing or sitting, in a walking simulator) affects task performance. Improved experience of “presence” might make performance in the virtual reality similar to real-world performance, whereas poor presence or an incorrect rendition might impair performance. Methods: We measured perception of a potential collision with stationary obstacles using four experimental situations to compare: standing or walking; walking with or without participant speed control; and correct or incorrect viewpoint. Participants stood or walked on a treadmill 75cm in front of a 95-degree-wide screen that displayed a “shopping mall” corridor with textured floor and shop fronts. Adult-man-size obstacles appeared for 1 second and participants indicated whether they would collide if they continued on the same path. Data for 14 participants were analyzed to find the participant’s perceived safe passing distance and decision quality. Results: When standing, participants had a slightly smaller perceived safe passing distance (p=0.07) and made better decisions (p=0.01) than when walking. Walking with and without participant speed control provided equivalent performance. The incorrect viewpoint biased the results to one side (p=0.08). Conclusions: Our attempts to increase realism did not alter perception of potential collisions. Our with-participant-speedcontrol walking condition required that the participant exert effort to propel the treadmill (i.e. not motorized), which might reduce task performance compared to a feedback-controlled motorized system. An incorrect viewpoint (rendition) caused a bias so obstacle side should be considered in data analysis. Other issues that might affect the experience of presence, including head-tracking and binocular view (stereo cue of flat screen), are under investigation. Supported in part by NIH grant EY12890