Investigation of eye movements during walking in real and simulated environments

• Main Author: Zhao, Ming
• Other Authors: Pelah, Adar ; Hunt, Andy
• Published: University of York 2012
• Subjects: 612.7
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.566321

This thesis reports experimental research on locomotion simulators, improving design methods and investigating human performance in simulator environments (SE) in comparison with real-world environments (RE). Locomotion simulators are used in gaming, training and medical applications, but in cases where motion capture of human movement is required they can be difficult to apply effectively without a great deal of trial-and-error. A new design tool is introduced, tested and successfully applied to this problem. In addition, eye movement performance is investigated and compared for a natural locomotion task requiring counting and reporting on ball targets of different sizes and colours while walking along a 38m corridor in both RE and its simulated counterpart. It is found that the number of eye saccades and other measures in the real environment are different and consistently better (lower) in performance compared with a carefully replicated simulated environment. Experiments that isolate the possible factors for these differences find that neither walking itself, neither walking speed nor visual optic flow speed alone can explain the differences. Rather, it is found that the underlying factor is a mismatched combination of walking and visual speed during locomotion that can occur in SE, revealing also that eye movements are involved in the complex perceptual interactions taking place during locomotion and are not due to mechanical effects of physical movement. An efficiency metric of eye movement performance during locomotion is devised as a model metric, and used to compare and tune the different combinations of parameters in locomotion simulators. Using this metric, results show that SE conditions can be tuned to achieve levels of performance comparable to the real world by matching optic flow and walking speeds at middle speed levels. Overall, the research contributes to a better understanding of human interactions in locomotion, and introduces new methods and approaches for investigating, tuning and optimizing the design and development of virtual reality locomotion environments.