The use of enhanced depth information in telepresence

Abstract

This thesis explored the potential performance benefits of enhancing depth cues in telepresence interfaces. A series of experiments addressed the role of binocular disparity and motion parallax in teleoperators’ performance. Experiments 1 and 2 demonstrated that the effects of enhancing binocular disparity and motion parallax depend upon the information demands of a given task. Enhanced depth cues helped observers to make more precise judgements in simple tasks that rely on judgement of depth differences or relative distance (alignment and depth matching tasks). However, systematic biases in performance were identified in metric tasks that rely on recovery of Euclidean geometry (shape judgements). Experiments 3 and 4 showed that teleoperators can quickly train to use enhanced depth information to perform metric tasks accurately, thus extending the range of tasks over which enhanced information can be used. Experiment 5 examined whether observers acquire transferable information about depth when learning to make depth judgements using binocular disparity or motion parallax. Participants showed no transfer of learning when training with altered binocular disparities and testing using motion parallax, or vice versa, suggesting that the learning demonstrated in Experiments 3 and 4 is cue-specific. Experiment 6 examined the use of depth cues for performing a task more typical of those performed under telepresence. The benefits of binocular and motion parallax cues, used in isolation or simultaneously, and the effects of enhanced motion parallax, were examined in a simulated “telesurgery” task, where other useful cues such as familiar size and perspective were already available. Observers’ performance vastly improved when binocular disparity was added as a cue; motion parallax, however, failed to improve performance, even when observers were encouraged to use it as a cue. These findings strongly suggest that telepresence performance may benefit from enhancing the information relevant to the specific task the system is intended for; contrary to the traditional approach in the design of telepresence, exact replication of the remote environment may not be crucial.