Modulating the performance of VR navigation tasks using different methods of presenting visual information

Abstract

Spatial navigation is an essential ability in our daily lives that we use to move through different locations. In Virtual Reality (VR), the environments that users navigate may be large and similar to real world places. It is usually desirable to guide users in order to prevent them from getting lost and to make it easier for them to reach the goal or discover important spots in the environment. However, doing so in a way that the guidance is not intrusive, breaking the immersion and sense of presence, nor too hard to notice, therefore not being useful, can be a challenge. In this work we conducted an experiment in which we adapted a probabilistic learning paradigm: the Weather Prediction task to spatial navigation in VR. Subjects navigated one of the two versions of procedurally generated T-junction mazes in Virtual Reality. In one version, the environment contained visual cues in the form of street signs whose presence predicted the correct turning direction. In the other version the cues were present, but were not predictive. Results showed that when subjects navigated the mazes with the predictive cues they made less mistakes, and therefore the cues helped them navigate the environments. A comparison with previous Neuroscience literature revealed that the strategies used by subjects to solve the task were different than in the original 2D experiment. This work is intended to be used as a basis to further improve spatial navigation in VR with more immersive and implicit methods, and as another example of how the Cognitive Neurosicence and Virtual Reality research fields can greatly benefit each other.