Adapting to dynamic registration errors using level of error (LOE) filtering

We describe our initial work on generating augmented reality (AR) displays in the face of dynamically changing errors in the pose (position and orientation) of both the user and objects in the world. Dealing with this problem is particularly important in mobile AR environments, where the tracking accuracy of the user's head can change frequently and dramatically as she moves between areas with radically different tracking systems, such as in and out of buildings. We introduce the notion of level of error filtering, analogous to level of detail culling in 3D graphics systems, to help programmers build interfaces that automatically adapt to changing registration errors.