Networked Architectures for Localization-Based Multi-User Augmented Reality

Multi-user augmented reality (AR) seeks to enhance the user experience in shared activities, whether collaboration for training, entertainment, or safety applications such as autonomous driving and driver assistance. AR involves a number of tasks that require both sensing and complex computations. A key task is localizing an AR device in the real world, in order to render the virtual holograms at the correct locations on the display. With cloud computing moving closer to the edge, a number of architectural options are available to aid in these complex computations, ranging from performing all of the computation at the end device to moving it all to the cloud, each with a different level of dependency on the communication link. In this work, we outline several architectural alternatives for localization in multi-user AR and their applicability to a number of important usage scenarios. The considered usage scenarios (entertainment, autonomous vehicles, and medicine) reflect a range of latency and spatial accuracy requirements. Within this context, we discuss our recent work on an edge-cloud-centric solution – SLAM-Share – and its applicability to various use cases. In addition, we evaluate its resilience to variations in communication delay. Overall, this article seeks to provide an overview of network architectures for localization-based AR to inform the design of future AR systems.