The Effect of Interaction-Display Offset on User Performance for a 3D-Widget Task in the Cave

Interaction in Immersive Virtual Environments (VEs) is fundamentally driven by the physical actions of the user’s body. The location of rendered virtual objects is, in contrast, the free choice of the VE application designer. The goal of Dmitri Lemmerman’s masters degree project is to explore the effect of the position of rendered graphical “widgets” on user performance in the use of those widgets. The particular task chosen was color-matching: using a 3D color picking widget, users were asked to match a target color. The independent variable was a scalar describing the offset between the interaction frame-of-reference (where the user moves his physical body) and the display frame-of-reference (where the graphical feedback appears). This relationship is not unique to the color picking widget, but generalizes to a larger set of 3D widgets in VEs. In the experiment, the relative position of the widget with respect to the user is tested in three conditions. While maintaining the same interaction frame-ofreference, the relative position of the display frame-of-reference is varied: to be the same as the interaction frame-of-reference (e.g. collocation), to be displayed at a 3 inch offset, and to be displayed at a 2 feet offset from the interaction frame-ofreference. Twenty-nine users were tested in a within-subjects design. Each subject was asked to match 15 colors in each of the 3 conditions. A distance scalar was computed based on a perceptual metric. A repeated measures ANOVA test on the data revealed a significant main effect of offset with a confidence of p < .05. Paired sample t-tests between the conditions did show a significant difference between the collocated case and each of the offset cases, but did not show a significant difference between the two offset cases. These results suggest that the presence of the user’s body appears to degrade performance when it is collocated with the virtual objects on which the user must focus (as it was in one of the conditions). However, even a short offset of 3 inches seems to alleviate this problem. Studies in HMD-based VEs have demonstrated significant performance advantages to collocation and the “direct manipulation” of virtual objects. Unlike HMDbased systems, a Cave-based VE platform allows the user to see both his body as well as the virtual objects. The results of this study should inform future VE user interface design for real-world applications and spur further experimental research in human/VE interaction principles.