This paper summarizes the detailed paths of participants in redirected walking (RDW) curvature gain experiments. The experiments were carried out in a wave field synthesis (WFS) system of 5×6 meters. Some users were blindfolded and had to control their walking by acoustical cues only, others wore an Oculus Rift DK2 which presented them a virtual scenery in addition. A marker at the participant's head allowed us to record the paths with our high-precision tracking system. The naive assumption of RDW with curvature gains would be that the test persons walk on the circumference of a circle, but the observed walking patterns were much more complex. Test persons showed very individual walking patterns while exploring the virtual environment. Many of these patterns may be explained as a sequence: 1. walk a few steps toward the assumed target position, 2. check for deviations, 3. adjust path to new assumed target position, which results in different patterns of various path curvature. The consequences for the application of RDW techniques are: Curvature gain tries to guide the users on a circular arc: the “ideal path”, whereas the real paths are mostly outside of the circle of the ideal path. The deviations in the audio-only case are much larger than in the audio-visual case. The measured curvature gain thresholds systematically under-estimate the required walking space, as they do not account for the required extra space for walking outside the circular path.

We present an experiment to identify detection thresholds for acoustic redirected walking by means of wave field synthesis. The most natural way to navigate an avatar through an immersive virtual environment (IVE) is by copying the tracked physical movements of a user. Redirected walking offers an approach to tackle the discrepancy between the potentially infinite IVE and the generally limited available physical space or tracking area, by applying manipulations, such as rotations or translations, to the IVE in form of gains to the users movements. 39 blindfolded test subjects performed a total of 2777 constant stimulus trials with various amounts of rotation and curvature gains. The test subjects were divided into four groups with different knowledge of the experiment, and one group performed two-alternative-forced-choice tasks, while the others could give feedback freely. The detection thresholds were greatly dependent on the groups i.e., the knowledge of the experiment. The 25% detection threshold was reached by the most relevant test group at gains that up-scaled rotations by 5%, down-scaled them by 37.5%, and bend a straight path into a circle with a radius of 5.71 meters. Almost no signs of simulator sickness could be observed.