US20260073552
2026-03-12
Physics
G06T7/70
The invention introduces a system and method for enhancing multi-user augmented reality (AR) by generating and updating a posegraph. This posegraph facilitates shared AR experiences by synchronizing spatial data from multiple client devices. When devices are in proximity, the system updates the posegraph with relative pose observations, assigning confidence values to improve accuracy and efficiency. This method enables seamless AR interactions in large environments without the need for complete map sharing or pre-mapped areas.
Maintaining a consistent shared coordinate frame across multiple devices is a significant challenge in AR applications. Conventional methods like visual-inertial odometry (VIO) and simultaneous localization and mapping (SLAM) have limitations such as drift and high bandwidth requirements. Existing solutions often rely on sharing complete maps or using markers, which are bandwidth-intensive and limit scalability. A scalable solution must efficiently fuse sensor data and operate without extensive pre-mapping or specialized infrastructure.
The Posegraph Optimization System (POS) addresses these challenges by using posegraphs, which are undirected graphs representing device poses connected by relative observations. This system allows for real-time, low-bandwidth sharing of localization data, supporting seamless multi-user AR experiences. Components of the POS include client devices with sensors, a central processing system for posegraph management, and a communication network for data exchange.
The system begins by collecting initial pose data from client devices, which may include VIO data, GNSS coordinates, WiFi signals, images, and IMU readings. An initial posegraph is generated to represent device positions and orientations. As users move, the system updates the posegraph when devices are in proximity, using observations from visual feature matching, device detection, UWB measurements, and more. Confidence values are assigned to different observations to optimize accuracy.
The POS employs continuous optimization techniques, such as Bundle Adjustment, to refine the posegraph and minimize errors in real-time as new data is received. By sharing only essential edge information rather than complete maps, the system reduces bandwidth usage while maintaining accurate multi-device localization. This approach supports large user numbers in dynamic environments without compromising performance or spatial consistency.