Identification of Redirected Walking Thresholds in Immersive Virtual Environments

Performing real walking in a virtual environment has been consistently shown to invoke a strong sense of presence and allow better way-finding and distance estimation. The biggest problem with real walking in virtual reality is that the virtual environment is most of the time much larger than the available physical space. One possible solution is called redirected walking (RDW), where a mismatch between the real and virtual trajectories is introduced. The mismatch could be in terms of linear velocity, angular velocity and heading direction. The focus of this work is on the mismatch in heading direction (also called curvature gain), in which users walk on a different curvature in the virtual environment than in real life. There is a limit to how much curvature gain can be applied without users noticing it, so-called thresholds. The first part of this work focuses on establishing a threshold identification procedure that is more effi- cient than the currently widely used one. This proposed procedure consists of an adaptive Bayes method, in conjunction with a true two-alternative forced choice task. Using this procedure, an experiment was conducted to identify their individual thresholds. Results showed that curvature gain thresholds span a wide range, indicating a need for personalized thresholds. Results also showed that men are on average more sensitive to curvature gain than women. The next part of this work is dedicated to further improving the threshold identification procedure. Rely- ing on the fact that RDW works due to the dominance of the visual sense over other senses, the correlation between curvature gain thresholds and performance in different visual dominance tests was investigated. The selected visual dominance tests were rod-and-frame, sway, vection and blind walking. Results showed that the performance in the rod-and-frame test could be used to partially predict an individual’s curvature gain threshold. RDW is not only perceived differently by individuals, but also in different external conditions. Ex- periments were conducted to investigate the effect of speed, environment size, feeling of presence and cognitive load on curvature gain thresholds. Using a rhythm to regulate walking speed, a significant impact of speed on thresholds was found. Specifically, users are more sensitive to curvature gain when they walk faster. No significant impact of the environment size (2m wide corridor vs. 10m wide room) was found. It has also been found that there is a negative correlation between the feeling of presence and curvature gain thresholds: users tend to detect redirection better when they have stronger feeling of presence and control over their virtual representation. While users were under the cognitive load of performing a secondary task (seven subtraction task), thresholds were significantly higher. A new technique to improve RDW was also introduced: a discrete rotation is applied every time users blink. Results show that while a scene can only be rotated 2.4 degrees when eyes are open, it can be rotated significantly more during blinks (9.1 degrees). A computer simulation was carried out in order to investigate if blink redirection technique could be combined with existing RDW techniques to improve redirection efficiency. Simulation results showed that in a sufficiently large tracking space and high enough blink frequency, the reset count and the minimum space requirement for walking without encountering a stop could be significantly reduced by up to 29%. Finally, in order to improve path prediction accuracy for RDW algorithms, the last part of the thesis explored if humans have a tendency to alternate their walking direction in a maze-like environment, so- called spontaneous alternation behavior (SAB). It was found that the probability of a user switching their turning direction is 72%, and 90% if the previous two turns were in the same direction. When users had to perform a secondary task, the alternation rate still remains significantly higher than 50%. A last series of experiments were conducted to investigate whether applying curvature redirection has an effect on SAB rate in a T-maze. In the first scenario, the T-maze is visually curved but curvature gain is applied in the opposite direction such that users ended up walking straight. In the second scenario, the T-maze is visually straight but a curvature gain is applied such that users ended up walking on a curve. Results showed that there is no significant difference in alternation rate in both scenarios