Using Locomotion Models for Estimating Walking Targets in Immersive Virtual Environments

Abstract

Redirected Walking allows a person to explore unlimited virtual environments in a limited physical tracking space. To prevent the user from colliding with the physical boundaries of the tracking space, so-called redirection techniques are used. These techniques introduce a subtle mismatch between the user's real and virtual movement and therefore keep him inside the tracking space while at the same time they allow him to explore an unlimited virtual environment. In most cases, there is more than one redirection technique available, and steering algorithms are used to select the best one at any given time. These algorithms use an optimal control scheme to select the optimal redirection action based on a prediction of the user's future path. In this paper, we present a novel approach for predicting a person's locomotion target. Using a set of known possible targets and models of human locomotion, this approach creates a set of expected paths and compares them to the path already traveled by the user in order to estimate the probability of the user heading for a certain target. We present a new approach for comparing two paths and evaluate its performance against three other approaches. We also compare four different ways of modeling a human's path to a target. To gather data for the comparison, a user study is conducted and the prediction performance of the different proposed approaches is discussed.