Heuristic Short-term Path Prediction for Spontaneous Human Locomotion in Virtual Open Spaces

Abstract

Redirected Walking (RDW) shrinks large virtual environments to fit small physical tracking spaces while supporting natural locomotion. Particularly in predictive RDW, one of the core concepts of RDW, algorithms rely on predicting users' future paths to adjust the induced redirection, which manipulates users' perception to deviate their physical paths from the intended virtual paths. Current path predictions either assume drastic simplifications or build on complex human locomotion models, which are inappropriate for real-time planning and thus not usable for RDW. Further, adapting existing predictive RDW algorithms to unconstrained open space exponentially increases their computational complexity, so that they are not applicable in real-time. In this work-in-progress paper, we discuss the currently prevalent issues of path prediction in RDW and propose simple yet flexible path prediction models that support dynamic virtual open spaces. Our proposed prediction models consist of two shapes: a drop shape represented by the lemniscate of Bernoulli and a sector shape. They define an area, in which linear and clothoidic walking trajectories will be investigated.