A Testbed for Studying and Choosing Predictive Tracking Algorithms in Virtual Environments

Abstract

We present a testbed for comparing predictive tracking algorithms that allows virtual environment system developers to make better choices about which predictors to use in their environments and aids researchers in determining how predictors work across various virtual environment configurations. Our testbed saves the virtual environment developer and researcher both time and effort with the important task of reducing dynamic tracking error and masking latency. The testbed consists of three components: a prediction algorithm library, a motion data repository, and a graphical testing application which provides users with the ability to test different predictive tracking algorithms across a variety of user motion sequences. The testbed provides enough generality for testing across different algorithmic and system parameters such as sampling rate, prediction time, and noise variance. The paper describes the contents of the predictor library and how to extend it, the types of motion data sets collected thus far, the motion data preparation methodology, and the graphical testing application's functionality and architecture. A simple testing scenario showing output from the testbed is also presented.