[Next] [Previous] [Top] [Contents] [Index]
Interaction techniques for a virtual workspace
Many VR management schemes depend on gesture. Jacoby and Ellis [8][9] use gloved hand gestures to invoke 'virtual menus' to highlight an item, to select an item, and to move the menu, with ray casting as selection feedback. Fairchild et al. [4] implemented and evaluated a scheme know as Gesture Sequence Navigation, a major problem for users of which was the amount of gesticulation required. In many environments [5][1][19], a circular or radial menu of symbols serves as the selector from which the user chooses. Although this method has been shown to work well in 2D interfaces (Kurtenbach and Buxton's 'marking menus' [10]), the amount of physical effort required of 3D users in repeated selection led us to look for non-gestural approaches to VR interaction design.
We here address mainly approaches that involve precise, calibrated, 3D spatial work similar to our own task domain. We pass over 3D systems like [17] that do not use stereo; depth cues such as controllable level planes must then be added to let the user navigate, and the design problems are quite different from those of a shared space environment.
The closest (and earliest) analogous work to the Virtual WorkBench described below is the half-silvered mirror device of Schmandt [18], which pioneered the use of menus and icons in space shared by the user and the display. Textured icons of buildings could be moved around, and it allowed curve sketching on a plane coinciding with a 45û surface and a tablet. Ergonomic problems, and the complexity limitations of real-time graphics in the early 1980s, prevented any immediate move to uses such as the hoped-for application to VLSI design.
To date, the environment most seriously aimed at precise, calibrated 3D work is the directly viewed stereo HoloSketch of Deering [1]. The goal, supported by impressive precision engineering, is accurate 3D work (for instance, in CAD), but the examples offered mix simple stored forms like planks with shaky freehand 3D forms, so it is hard to judge the precision a user can actually achieve. (On a standard Sun workstation, some form of elbow support must clearly be added before most users could hope for adequate control, or use it in long sessions.) The elaborate menus (daisy wheels within wheels) are supplemented by keyboard input.
Shaw and Green [19] present the user with rotating menus, 'ray cast' for selection, memorize multi-button sequences for both hands, and still require keyboard and mouse input.
Hinckley et al. [6] avoid widgets, and attach virtual objects to real world props, moving parallel to the graphics but elsewhere. A new physical prop is needed for each new object type in the display; for example, a doll's head in the user's hand can represent a CT head volume better than a heart. A finite flat prop can represent an infinite cutting plane, to control a sectional view, but placing virtual therapeutic objects inside something represented by an impenetrable prop is difficult. Scaling--often vital for precision--breaks the metaphor.
The 'virtual tricorder' of [22] hangs from the user's virtual belt in a fully immersive HMD environment, where the technology cannot yet make normal text readable at a normal distance. When the calibration, latency, resolution and field of view problems are solved, such a tool will become a natural widget in such worlds. It is not applicable to precision work in the shorter term.
[Next] [Previous] [Top] [Contents] [Index]
Generated with Harlequin WebMaker