[Next] [Previous] [Top] [Contents] [Index]
Interaction techniques for a virtual workspace
The user should be free to focus on problems of engineering or artistic design, diagnosis, etc., with a minimum of conscious attention to details of selection, rotation, and so on. The actions required should thus echo natural ones as far as possible (reach for something rather than recall a button sequence), allow both visual and motor memory to assist, and lead to error as rarely as possible. A Workbench tool always has a handle, always located where it is seen and felt to be, through the eyes and through the user's grip on a material sensor, matching human need for hand-eye coordinated location (emphasized by the finger marks on every monitor) becomes much more acute in 3D.
Similarly, we do not select or manipulate objects by remote pointing such as 'ray casting' [5][12]. This is partly because we do not need to, since we do not use the work volume to display objects that are out of reach, and partly because we need a tool steadier than a laser pointer. Moreover, pressing a button usually disturbs the hand a little; a tool tip near the fingers moves much less than the end of a ray. Selecting 'delete' when 'move' is intended, or selecting the wrong small item for deletion, can waste a great deal of previous work. Choosing one out of (for instance) fifty standard landmarks on a skull requires little concentration if one thinks 'tweezers', a great deal if one is pointing from twenty centimetres away--particularly with a device like the DataGlove. (A virtual hand allows a very literal 'finger and thumb' grasping, but displaying it is costly and obscures the workpiece; it is better to model a handle than the hand.)
The 3D motions possible for a menu scheme--spinning, unfolding, retreating,...--are seductive to the designer, but all dangerously interesting. If one needs to choose as quickly and unthinkingly as the left little finger finds 'a' on a keyboard, unstable locations are a menace.
Although the Workbench immerses the workpiece in your space, rather than you in a larger volume, the tools in your hands are an integral part of the virtual world. This gives the sense of co-presence characteristic of immersion and reproduces most of their constraints: no access to the keyboard or other 'external world' devices than the material handles of the virtual tools.
The display must be used efficiently, in both space and time; it must appear uncluttered, and respond fast to minimize time-lag problems, though medical applications make great demands on rendering speed. Both gain by concentrating display effort on the focus of interaction. While the user interacts with one object, others can be displayed in a simplified form. While the user is dealing with menu buttons, data such as an MRI volume may be rendered at lower resolution; conversely, when dealing with the volume, the menu is rendered as a simplified box.
Effective work demands ease in transitions. In "pick 3D point, rotate volume, pick next 3D point," the implied tool changes from selector to rotator to selector should need a minimum of time and conscious attention. Feedback is vital in the user's sense of the state of the interaction. We currently combine visual feedback via the display with the user's neuromuscular sense of the tool's location, and will shortly add auditory and haptic feedback.
We limit interactions to a natural volume within which reach and stereo fusion are comfortable. This places objects within 15 to 50 cm from the eyes, in a cone perhaps 50û wide. Objects too close to the eye are hard to fuse, and conflict with the eyes' focus on the display. Those too far to reach can be retrieved by shrinking the scene towards one. This control of scale is a key to precise control; one should move easily between designing a cathedral and detailing its pews.
[Next] [Previous] [Top] [Contents] [Index]
Generated with Harlequin WebMaker