Eye tracking for locomotion prediction in redirected walking

Abstract

Model predictive control was shown to be a powerful tool for Redirected Walking when used to plan and select future redirection techniques. However, to use it effectively, a good prediction of the user's future actions is crucial. Traditionally, this prediction is made based on the user's position or current direction of movement. In the area of cognitive sciences however, it was shown that a person's gaze can also be highly indicative of his intention in both selection and navigation tasks. In this paper, this effect is used the first time to predict a user's locomotion target during goal-directed locomotion in an immersive virtual environment. After discussing the general implications and challenges of using eye tracking for prediction in a locomotion context, we propose a prediction method for a user's intended locomotion target. This approach is then compared with position based approaches in terms of prediction time and accuracy based on data gathered in an experiment. The results show that, in certain situations, eye tracking allows an earlier prediction compared approaches currently used for redirected walking. However, other recently published prediction methods that are based on the user's position perform almost as well as the eye tracking based approaches presented in this paper.