Exploring, walking, and interacting in virtual reality with simulated low vision: a living contextual dataset

Area of Science:

• Human-computer interaction
• Virtual reality research
• Robotics and autonomous systems

Background:

• Understanding human navigation and interaction at road crossings is crucial for developing safer autonomous systems and virtual reality environments.
• Existing datasets often lack the richness and scale required for nuanced behavioral analysis.

Purpose of the Study:

• Introduce the CREATTIVE3D dataset, a large-scale, multi-modal resource for studying human behavior at virtual road crossings.
• Investigate the impact of simulated low-vision conditions on human navigation and interaction.
• Provide a foundation for reproducible and comparable research in human-robot interaction and virtual reality.

Main Methods:

• Collected 40 hours of human motion data, encompassing 2.6 million poses, in dynamic 3D virtual reality scenarios.
• Integrated multivariate data including gaze, physiology, and motion capture.
• Implemented simulated low-vision conditions using dynamic eye-tracking during both real and simulated walking.

Main Results:

• Established the largest dataset of human motion in fully-annotated virtual reality road crossing scenarios.
• Captured rich, dynamic data capturing gaze, physiology, and motion.
• Demonstrated the feasibility of studying low-vision impacts within a complex 6-degrees-of-freedom VR environment.

Conclusions:

• The CREATTIVE3D dataset provides an unprecedented resource for analyzing human behavior in complex interactive scenarios.
• Enables fine-grained analysis of user nuances and promotes comparability across studies.
• Facilitates advancements in autonomous systems, virtual reality safety, and assistive technologies.