Depth cues for telerobotic tasks in virtual environments
Andrew Menching Liu, LAWRENCE W. STARK · 1992
The importance of a well-designed visual interface for optimum performance of telerobotics tasks had been well-documented for current systems. The technological and physical constraints of the current systems may lead to significant performance degradations. An alternative that is becoming feasible, is the virtual environment (VE) interface which, among other capabilities, displays a computer-generated reconstruction of the telerobotic environment. These quantitative studies of video and VE displays for a variety of display conditions will also illustrate some perceptual requirements of telerobotic manipulation relating to human depth perception. The first set of experiments examined the relationship between task performance and display image quality. Monocular blur is representative of the visual degradations that might occur in present systems. I found that the blur degraded task performance for three types of telerobotic tasks. The results show that blur of two diopters or more eliminated the advantage of stereoscopic display for the three-axis tracking and pick-and-place tasks. For axle manipulation, positioning error was not increased by the blur, but task completion time was prolonged. These three performance changes are probably due to the reduction of stereoacuity by monocular blur. Therefore, moderate levels of visual degradations that degrade the operator's depth perception and resolution should be avoided or eliminated. Virtual environment interfaces do not suffer from this type of image degradation since the image is not transmitted from the telerobotic working environment. The second set of experiments investigated the effect of rendering scenes with wireframe or solid objects on task performance. This is one way that a VE display can be designed to compensate for the computational limitations of the system. Occlusion cues, created by solid modelling, and stereo display of wireframe images provided enough depth information to perform tracking and pick-and-place tasks with nearly equivalent performance. Simultaneous presentation of the cues resulted in very small performance improvements over single cue presentation. These results suggest that a high degree of visual realism, which is computationally expensive, is not necessary to maintain good performance as long as sufficient depth information is available. The third and final set of experiments studied the advantages of adding object or observer motion to the VE interface. These enhancements can be implemented in the computer 3D model, thereby eliminating the need for specialized hardware at the telerobotic environment to implement motion. (Abstract shortened by UMI.)