2D/3D Interaction in Tabletop Augmented Reality
2021.
Abstract
This thesis contributes to the research field of Human-Computer Interaction (HCI). The focus of the research is on tabletop Augmented Reality (AR) interaction. Allowing visualization of 3D virtual information linked to a table that acts as a physical support for interaction is one of the main advantages of tabletop AR.Tabletop AR systems thus define a relevant alternative to the often complex desktop software used to create and visualize 3D models (3D modeling, CAD, urban planning, architecture).Currently, these systems use mid-air interaction techniques inherited from classical AR that mimic the manipulation of physical objects. Although intuitive and direct, these techniques are often imprecise and tiring for long-term use.Tabletop AR systems also allow tactile interaction on the table which is more precise but less intuitive for 3D tasks than 3D mid-air interaction. In this context, the research question of our work concerns the design of 3D object selection techniques in tabletop AR that are precise, not tiring, and easy to handle and use.The design space to be explored is vast and includes selection techniques based on tactile interaction on the table, mid-air interaction above the table, and hybrid interaction combining the two interaction spaces i.e. the table and the space above the table. The exploration of this vast design space gave rise to three contributions.Our first contribution is to identify which AR device, between a head-mounted display (HMD) and a handheld device (smartphone/tablet), is the most efficient for the visualization and selection of 3D objects. The experimental results show better performance and user feedback with an HMD. We have therefore focused on tabletop AR with an HMD in our subsequent work.Our second contribution concerns the selection of distant 3D objects. We propose a new bi-manual technique, namely textit{RayLens}, to improve the accuracy of distant object pointing. textit{RayLens} combines a ray and a 2D magnifying lens that can be moved in 3D. This technique has proven to be efficient, not tiring, and particularly appreciated by the users thanks to its ease of use and intuitiveness.Finally, our third contribution focuses on the precise selection of objects at hand and further develops the use of the table as an input space for interaction. By experimentally comparing the three interaction spaces provided in tabletop AR (2D table space, 3D space above the table, hybrid space), we highlight the importance of the table support in improving accuracy and reducing fatigue. We also show the effectiveness of hybrid techniques that successfully combine the accuracy of tactile interaction with the speed of mid-air interaction.