Speed Perception in VR

A psychophysics experiment on how we perceive speed in 3D environments





January 2017 - May 2018


Vision Lab, Psychology Department, UIUC


Tools: Oculus Rift, Unity game engine, C#, R





Overview


As a research assistant in the Vision Lab, I helped develop and pilot a simulated roadway experiment in Unity using C#. We leveraged virtual reality technology to create a first-person driving simulation, in which participants assess their speed.


Since the existing research on VR perception is sparse, this project is an important step toward informed virtual reality design and understanding human perception in 3D spaces.


Research goals


The study was designed to address 2 primary research questions:

  • How accurate is a subject’s perception of their speed in a vehicle?
  • How does a participant’s viewpoint affect their speed perception?

The findings will demonstrate if speeds are underestimated in virtual reality systems, and if height manipulation can compensate for judgment discrepancies and lead to accurate speed estimates.


Trials


Subjects drive down a virtual road at a constant speed, and must gauge their speed relative to any two of the four test speeds: 35, 45, 55, and 70 mph. Throughout the trials, we manipulate the height of the participant’s viewpoint (calibrated based on the participant’s sitting height) along with their speed. We test the following heights:

  • 7 cm below baseline
  • baseline height
  • 7 cm above baseline

Subjects also answered questionnaires on demographics, prior experience with games (especially in VR), and if they experienced any simulator sickness.


Trial demo





Applications


The results of this experiment have applications for various training programs, such as military and novice driver training. Currently, many existing 3D simulations have difficulty replicating the mechanics of the real world. As a consequence, they are ineffective and can result in long-term cognitive or physiological harm to the user.


Results from this study will discover adjustments (such as certain depth and peripheral cues) that will make a simulation true to how human perception operates in the real world. The more realistic a simulation is, the more likely it can successfully serve as a low-cost, effective training option for mainstream users to adopt.