Abstract
Rising cyclist fatalities in the US necessitate innovative methods to study cyclist safety. Virtual reality (VR) technology has advanced significantly in recent decades, leading to commercially available, highly realistic, and cost-effective VR head mounted displays (HMD). Virtual reality bicycle simulators allow researchers to collect stated and revealed preference data using easily modifiable virtual environments, while keeping participants in a controlled and safe environment. This is a potential solution to drawbacks of traditional research methods and existing data sources (such as crash and survey data). Crash data, which is largely geared towards motor vehicles, excludes information that would be useful for studying cyclist safety (e.g., bicycle infrastructure at crash site, safety equipment such as lights and reflectors, helmet use, etc.). Bicycle crashes are also highly underreported in crash databases. While surveys are a cost-effective method to reach large populations, particularly when distributed online, they are subject to hypothetical bias and individual interpretation, and cannot provide the kind of immersive visualization afforded by VR.
Bicycle simulators exist in labs worldwide, and range from basic (i.e., using a keypad to move a bicyclist forward on a screen) to technologically advanced (i.e., physical stationary bicycle used by the research participant that can record steering, braking, and speed, with visualization across multiple screens or via a VR HMD). This dissertation discusses the design and implementation of a state-of-the-art VR bicycle simulator in a new lab (the Omni-Reality and Cognition Lab [ORCL] at UVA). This dissertation also describes the validation of the simulator against real world cyclist behavior and using the simulator to assess cyclists’ perceived safety under different bicycle infrastructure scenarios.
While VR allows for a controlled, low-risk environment for repeatable experimentation, established methods for validating bicycle simulators for transportation research currently do not exist. This study validates a bicycle simulator which allows users to pedal, steer, and brake on a stationary bicycle trainer while wearing a VR headset. A replica virtual environment is created from the Water Street corridor in Charlottesville, Virginia. Then, simulator user behavior (N=50) is benchmarked against real world cyclist behavior (N=90) collected from video footage of Water Street. Absolute validity of speeds between the participants in the VR environment and cyclists in the real-world corridor was verified. VR research participants also reported via stated preference surveys their perceived realism about various aspects of the simulator. Results from the survey show that 94% of participants felt the simulator was immersive and 70% felt the VR environment was consistent with their real-world cycling experiences. This research presents a framework for validating a VR bicycle simulator, a critical first step to confirm the capabilities of VR simulation for bicycle transportation research.
Perceived safety of vulnerable road users can be studied using VR simulators to reproduce real-world-like behaviors, while concurrently collecting SP data. This study uses the VR bicycle simulator and instructs participants to ride through three different immersive virtual environments where all settings are identical except the bicycle infrastructure type (sharrows, bike lane, and protected bike lane). The sharrows environment is a replica of the as-built conditions of the real-world corridor against which the simulator has been benchmarked. Using data from post-experiment surveys, it was found that overall, participants (N=50) felt significantly safer in bike lanes and protected bike lanes compared to the sharrows, but the effect is nuanced based on gender. Female cyclists found both the bike lane and protected bike lane to significantly increase safety compared to the sharrows, while for male cyclists, only the bike lane was reported to feel significantly safer than the sharrows. When examining cyclist perceptions across two vehicle volume levels, no statistically significant difference in perceived safety was found. Some behavioral differences were observed across the three environments; cyclist speeds were lower in the protected bike lane than in the bike lane or as-built environments and standard deviation of distance to the curb is significantly lower in the protected bike lane and bike lane environments than in the as-built environment. These results indicate that bicycle infrastructure can meaningfully impact cyclist’s physical location and movement. This study demonstrates the potential for using VR simulation for understanding cyclist perceived safety on various bicycle infrastructure types, which may be especially valuable when evaluating new and unfamiliar infrastructure designs.
