Stated preference analysis of autonomous vehicle among bicyclists and pedestrians in Pittsburgh using Bayesian Networks

Abstract

Presently, technology innovations are disrupting the status quo and changing the way people travel. In an effort to enhance safety, ease driving tasks, and attract car buyers, automobile manufacturers are offering new vehicle automation technologies. As these vehicle technologies become more automated, navigation around and interactions with pedestrians and bicyclists in complex travel environments becomes more challenging. With people being less predictable and less identifiable than other machines, these technologies can pose safety concerns for all users. In light of this, there is a need to further study the interaction between cyclists, pedestrians, and automated vehicles. In 2019, Bike Pittsburgh (BikePGH) conducted a survey of autonomous vehicles (AVs) in Pittsburgh, Pennsylvania to understand the perception of bicyclists and pedestrians when sharing the road with AVs. This study used the data collected by BikePGH to understand various factors associated with bicyclists' and pedestrians’ perception of safety when sharing the road with AVs. Bayesian Networks (BNs) were used to learn the probabilistic interrelationship among AVs' aspects. BN results revealed that familiarity with the technology behind AVs, feeling safe while sharing the road with AVs, and using Pittsburgh's public streets as a proving ground for AVs were associated with higher likelihood of AVs’ safety potential to reduce traffic injuries and fatalities. On the other hand, feeling safe while sharing the road with human-driven cars was associated with lower likelihood of AVs’ safety potential to reduce traffic injuries and fatalities. Furthermore, the BN model predicted that the experience of sharing the road with AVs while riding a bicycle or walking, familiarity with the technology behind AVs, and using Pittsburgh's public streets as a proving ground for AVs were associated with higher likelihood of feeling safe sharing the road with AVs. The joint analysis of the variable showed the highest predicted probabilities of 95% and 86%, respectively for AVs’ potential to reduce traffic injuries and fatalities and for feeling safe sharing the road with AVs. The practical application of this study is presented along with recommendations to operators, city engineers, and planner.