Evaluation of AR Inconsistencies on AR Placement Tasks: A VR Simulation Study

One of the major challenges of Augmented Reality (AR) is the registration of virtual and real contents. When errors occur during the registration process, inconsistencies between real and virtual contents arise and can alter user interaction. In this paper, we assess the impact of registration errors on the user performance and behaviour during an AR pick-and-place task in a Virtual Reality (VR) simulation. The VR simulation ensured the repeatability and control over experimental conditions. The paper describes the VR simulation framework used and three experiments studying how registration errors (e.g., rotational errors, positional errors, shaking) and visualization modalities (e.g., transparency, occlusion) modify the user behaviour while performing a pick-and-place task. Our results show that users kept a constant behavior during the task, i.e., the interaction was driven either by the VR or the AR content, except if the registration errors did not enable to efficiently perform the task. Furthermore, users showed preference towards an half-transparent AR in which correct depth sorting is provided between AR and VR contents. Taken together, our results open perspectives for the design and evaluation of AR applications through VR simulation frameworks.