RESUMO
The creation and management of content are among the main open issues for the spread of Augmented Reality. In Augmented Reality interfaces for procedural tasks, a key authoring strategy is chunking instructions and using optimized visual cues, i.e., tailored to the specific information to convey. Nevertheless, research works rarely present rationales behind their choice. This work aims to provide design guidelines for the localization of in-view and not occluded components, which is recurrent information in technical documentation. Previous studies revealed that the most suited visual cues to convey this information are auxiliary models, i.e., abstract shapes that highlight the space region where the component is located. Among them, 3D arrows are widely used, but they may produce ambiguity of information. Furthermore, from the literature, it is unclear how to design auxiliary model shapes and if they are affected by the component shapes. To fill this gap, we conducted two user studies. In the first study, we collected the preference of 45 users regarding the shape, color, and animation of auxiliary models for the localization of various component shapes. According to the results of this study, we defined guidelines for designing optimized auxiliary models based on the component shapes. In the second user study, we validated these guidelines by evaluating the performance (localization time and recognition accuracy) and user experience of 24 users. The results of this study allowed us to confirm that designing auxiliary models following our guidelines leads to a higher recognition accuracy and user experience than using 3D arrows.
RESUMO
This work investigates the possibility of using a novel "minimal AR" authoring approach to optimize the visual assets used in augmented reality (AR) interfaces to convey work instructions in manufacturing. In the literature, there are no widely supported guidelines for the optimal choice of visual assets (e.g., CAD models, drawings, and videos). Therefore, to avoid the risk of having AR technical documentation based only on the author's preference, our work proposes a novel authoring approach that enforces the minimal amount of information to accomplish a task. Minimal AR was tested through a simulated AR LEGO-based assembly task. The performance (completion time, mental workload, errors) of 40 users was evaluated with 4 combinations of visual assets in 4 tasks with an increasing amount of information needed. The main result is that visual assets with an excess of information do not significantly increase performance. Therefore, the location of a specified object should be "minimally" authored by an auxiliary model (e.g., a circle and an arrow). For identifying an object within a couple, color coding is preferred to using additional visual assets. If more than two objects must be identified, a drawing visual asset is also needed. Only when the orientation of a selected object must be conveyed, animated product models are required. These insights could be helpful for an optimal design of AR work instructions in a wide range of industrial fields. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s00170-021-08449-6.
