Unravelling bicyclists' perceived safety using a bicycle simulator combined with immersive virtual reality and a physiological sensor

Abstract

This dissertation investigates the capabilities and limitations of immersive 360-degree virtual reality (VR) for bicycle research to study bicycle facility preferences, perception of safety of bicyclists, and bicycling behaviour. There is a need for methods that provide a better understanding of bicyclists’ perceived safety and preferences on currently unavailable and/or unknown facilities. Several barriers to bicycling in Singapore were identified including safety concerns, unavailability of appropriate bicycle facilities, weather conditions, and trip distance, which were in line with previous findings in other countries. However, a major difference between the research at hand and research from previous studies is that due to the scarcity of bicycle infrastructure and bicycling experience in Singapore, the application of commonly used survey methods might lead to unreliable and inadequate results. Different survey methods have been used to study bicyclists’ behaviour, experiences, and preferences; ranging from verbally described facilities to surveys including images and videos. VR experiments blur the boundaries between stated preference (SP) surveys and revealed preference (RP) surveys and provide a realistic sense of presence in any design, as compared to imagery and video-based surveys. Yet, little research has fully explored the potential of VR for mobility applications, in particular, bicycle research. A deeper understanding of the efficacy of VR in bicycle research is critical for knowing how to use it in experiments as a research method, and ultimately, to reach more accurate results. Especially because VR experiments can be difficult to implement and require meticulous planning. To shed some light on the application of VR for bicycle research, this dissertation discusses the results of an experiment using a bicycle simulator combined with immersive VR. In total, 150 Singaporeans participated in this experiment and were asked about their demographics, bicycling attitudes, and perceptions and preferences after bicycling in five different environments with an instrumented bicycle in VR. A $5\times2$ mixed design was used with bicycling environment as the within-subject factor and pedestrian / traffic volume as the between-subject factor. Three data sources were available: 1. questionnaire that was filled out before, during, and after the experiment, 2. bicycle simulator measurements such as speed, pedalling, and braking activity, and 3. physiological sensor that collected the electrodermal activity (EDA) of each participant indicating their arousal level. Since speed and passing distance of adjacent motorised traffic affect bicyclists’ perceived safety and comfort, a part of this research was dedicated to the perception of speed and perception of space in VR by varying vehicle speeds and lane widths in immersive virtual reality. The results identified thresholds to which the majority of the participants could recognise speed and distance variations between two subsequent scenes in VR. These thresholds can be used in future research investigating bicycle level of service, stated preference designs, and studies involving passing distance. Other findings of this VR experiment confirmed how bicycling environment and ambient pedestrian / traffic volume affects perceived level of safety (PLOS) and willingness to bicycle (WTB). The highest ratings for PLOS and WTB were given to the segregated bicycle path. Overall, participants' preferences of bicycle facility were found to be in line with previous studies using classic bicycle survey methods. In addition, the choice of riding speed was found to be greater on segregated bicycle path and painted bicycle path on the road. Results of a regression model showed how braking activity along with head movement increased right before arriving at an intersection to reduce the speed and check for turning vehicles. Participants' arousal level also significantly increased right before the intersection showing the induced stress when encountering an intersection. Besides intersections, arousals happened at other locations on each facility due to different sources such as conflict with pedestrians or passing of motorised vehicles. The incurred arousals were quantified and they were associated to bicycling stress or excitement in each facility. Based on these findings, a bicycle simulator combined with immersive 360-degree VR found to be a promising research method to evaluate existing and non-existing bicycle facilities which provides valuable inputs for researchers as well as planners to develop local bikeway network. Furthermore, the controlled laboratory environment allowed for the application of physiological sensors which enables researchers to identify hazardous bicycling events and have a deeper understanding of experienced bicycling stress at a very fine level. Finally, the limitations of VR framework for bicycle research are presented.