The effect of registration error on tracking distant augmented objects

Abstract

We conducted a user study of the effect of registration error on performance of tracking distant objects in augmented reality. Categorizing error by types that are often used as specifications, we hoped to derive some insight into the ability of users to tolerate noise, latency, and orientation error. We used measurements from actual systems to derive the parameter settings. We expected all three errors to influence userspsila ability to perform the task correctly and the precision with which they performed the task. We found that high latency had a negative impact on both performance and response time. While noise consistently interacted with the other variables, and orientation error increased user error, the differences between ldquohighrdquo and ldquolowrdquo amounts were smaller than we expected. Results of userspsila subjective rankings of these three categories of error were surprisingly mixed. Users believed noise was the most detrimental, though statistical analysis of performance refuted this belief. We interpret the results and draw insights for system design.