Data-driven methods for dockless bike infrastructure planning

In this work we address the problem of data-driven placement of critical bike infrastructure to address user route demand. The proposed framework uses trip origin-destination data that is commonly produced by bike share operators and included in standard data feeds. It does not require intermediate trip GPS data, which is not yet widely available. To compensate for lack of full path information in the data, the proposed method estimates the path of each trip using a shortest path routing algorithm, which runs on a bike-accessible network graph with user preferences. We apply this path estimation to each trip, which results in a volume estimate on each edge in the network graph. We can then assess the effectiveness of infrastructure upgrades from trip distance coverage and user impact perspectives. The framework is applied on a data set of tens of thousands of dockless bike share trips that occurred during the first month of a pilot program with ofo bikes at Vanderbilt University. Case study findings demonstrate that approximately 40% of observed trip distance can be covered by improving only 5% of the infrastructure and 75% of trips will travel some portion of their path on the upgraded infrastructure. This highlights the significant potential benefits for a modest infrastructure investment. While the capture rates may be different on each network, the methodological tools are applicable to data from any docked or dockless bike share system and infrastructure network. The tools will serve as the basis for a software platform that will help any city analyze data collected from shared use mobility devices including bikes and electric scooters and assess infrastructure investment.