Natural freehand grasping of virtual objects for augmented reality

Grasping is a primary form of interaction with the surrounding world, and is an intuitive interaction technique by nature due to the highly complex structure of the human hand. Translating this versatile interaction technique to Augmented Reality (AR) can provide interaction designers with more opportunities to implement more intuitive and realistic AR applications. The work presented in this thesis uses quantifiable measures to evaluate the accuracy and usability of natural grasping of virtual objects in AR environments, and presents methods for improving this natural form of interaction. Following a review of physical grasping parameters and current methods of mediating grasping interactions in AR, a comprehensive analysis of natural freehand grasping of virtual objects in AR is presented to assess the accuracy, usability and transferability of this natural form of grasping to AR environments. The analysis is presented in four independent user studies (120 participants, 30 participants for each study and 5760 grasping tasks in total), where natural freehand grasping performance is assessed for a range of virtual object sizes, positions and types in terms of accuracy of grasping, task completion time and overall system usability. Findings from the first user study in this work highlighted two key problems for natural grasping in AR; namely inaccurate depth estimation and inaccurate size estimation of virtual objects. Following the quantification of these errors, three different methods for mitigating user errors and assisting users during natural grasping were presented and analysed; namely dual view visual feedback, drop shadows and additional visual feedback when adding user based tolerances during interaction tasks. Dual view visual feedback was found to significantly improve user depth estimation, however this method also significantly increased task completion time. Drop shadows provided an alternative, and a more usable solution, to dual view visual feedback through significantly improving depth estimation, task completion time and the overall usability of natural grasping. User based tolerances negated the fundamental problem of inaccurate size estimation of virtual objects, through enabling users to perform natural grasping without the need of being highly accurate in their grasping performance, thus providing evidence that natural grasping can be usable in task based AR environments. Finally recommendations for allowing and further improving natural grasping interaction in AR environments are provided, along with guidelines for translating this form of natural grasping to other AR environments and user interfaces