Handling Error in Ubiquitous Tracking Setups

Augmented reality is a technology that aims at combining real and virtual reality by integrating virtual objects into the user’s view of the real world. In order to place these augmentations at the right locations, different tracking technologies are employed for finding out where the user is and in what direction he is looking. This requires high accuracy, high update rates and low delay to give a convincing impression. Ubiquitous tracking is a new research project focused at dynamically integrating userworn and stationary tracking systems by combining the concepts of augmented reality and ubiquitous computing in order to enable mobile wide-area augmented reality applications. When many different and previously unknown tracking technologies are to be combined dynamically at runtime, statistics about the sensor accuracy are necessary and all transformations of measurements require updating the associated error descriptions. In order to allow automatic conversions between different coordinate systems, the product of multiple tracker measurements can be computed. In heterogeneous tracking setups however, measurements by different sensors generally are not made simultaneously and therefore require pre-processing before such a combination is possible This thesis proposes a mathematical model as well as a software architecture based on the DWARF framework for describing sensor errors in position and orientation in Ubitrack systems. Using a Gaussian error model, errors are propagated over multiple chained coordinate system transformations. A Kalman filter-based approach with separate motion models is described for dynamic sensor fusion, measurement simultaneity, prediction, as well as the estimation of static transformations.