Mapping bicyclist route choice using smartphone based crowdsourced data

Abstract

Bicycling has been identified as a critical component of livable communities, as it offers an environmentally friendly, cost-effective, congestion-reducing, and health-promoting mode of transportation for short trips. According to the National Household Travel Survey (NHTS 2009), nearly 40% of all personal trips in the U.S. are two miles or less, a reasonable bicycling distance. Although some of these trips may be secondary trips tied to the mode of primary trips (e.g. grocery shopping enroute to work or home), and are hence, non-convertible unless the primary commute mode is changed, the fact that only about 1.8 % of all those personal trips less than two miles are made on bicycles, is alarming. A major reason frequently cited for not adopting bicycling as a travel mode is a perceived lack of safety in facilities shared with high speed and volume traffic. To remedy the situation, ideally, all streets should be provided with separate bicycle facilities but agencies do not have enough funding nor enough right-of-way in many cases. Cyclists also differ widely in their perceptions of roadway safety and comfort and hence, possibly in their preference for infrastructure. To date, there are not enough data to understand cyclist preferences or how far the cyclists are willing to travel to access cycling facilities since bicyclists are a small and dispersed group and it is difficult to get data on their travel patterns through traditional traffic counts. Cycle Atlanta, a GPS based smartphone application (app) was developed at Georgia Tech in collaboration with the City of Atlanta to collect revealed preference route choice data of cyclists in Atlanta via crowdsourcing. This research used the collected data to (1) validate the popular classification of cyclists into different rider types based on their comfort and confidence while also modelling the underlying influence of socio-demographic attributes on self-perception of level of confidence and comfort; (2) develop a model to understand how far cyclists are willing to deviate from the shortest network distance based route and (3) design segmented route choice models for different types of cyclists based on their socio-demographics as well as their comfort and confidence level. In addition, a stated preference survey was analyzed to understand what factors influence the decision to use bicycle as mode of transportation. Finally, a data cleaning, curating, and map matching algorithm was developed as part of the research. This research, one of the first studies to use crowdsourced data to analyze cyclist behavior, is unique in its focus on the influence of socio-demographic and attitudinal makeup of cyclists on their decision to bicycle and their route choice. The results from this research provide valuable insight for future planning and policy decisions. First, female and senior cyclists are found to be in general low confidence, low comfort riders and they significantly differ in their route choice and infrastructure preference from their more confident counterparts. Second, the assumption that with more riding experience cyclists become confident enough to share the street with vehicular traffic, is not without its caveats. Although cyclists with more riding experience tend to see themselves as more confident riders, preference for separate infrastructure pervades all rider types, as does the negative influence of high speed and volume traffic. Third, cyclists are generally found to shy away from longer trips and hence, when faced with the trade-off between a significant detour and safety concerns, they may not make the trip itself. Therefore, having a connected network close to the shortest distance path is important in encouraging newer and low confidence bicyclists. This research provides a model that can be used to estimate acceptable deviation from any route based on road attributes and the cyclist characteristics.