Tech Report CS-02-07

The Predictive Tracking Algorithm Testing Suite: A Tool for Developing and Analyzing Predictive Tracking Algorithms

Joseph J. LaViola Jr.

July 2002

Abstract:

We present The Predictive Tracking Algorithm Testing Suite, a software platform for developing and examining different predictive tracking algorithms across a variety of system and algorithmic parameters. The suite allows for the evaluation of predictors across a multidimensional parameter space which includes sampling rate, prediction time, noise variance, specific algorithmic parameters, and different types of user motion. By providing a general testing suite, VR and AR system designers can ask a variety of questions about how to maximize predictor performance in tracking systems and applications. We describe the features of the testing suite and it's underlying software architecture as well as our data preparation methodology. We also present some simple experiments that show the testing suite's utility by examining four Kalman filter-based predictors (two for position and two for orientation) on head and hand motion datasets.

(complete text in pdf or gzipped postscript)