Abstract

Free-floating bike-sharing systems can have a positive influence on the mobility of urban centers and developing efficient localization strategies is crucial to avoid crowding at peak times and increase service availability. Our study aims to efficiently resolve the location of virtual bike stations in a Latin American city through a geospatial data wrangling methodology that allows us to respond opportunely to the potential demand forecasted for the city. This approach is imple-mented in Python, and it uses the Geopandas and LocalSolver libraries to determine locations for the virtual bike stations that maximize the system coverage. The decision-making process is supported by a binary integer mathematical programming model, and the instances are built from intercity travel surveys that provide realistic data based on travel demand. The developed deci-sion support system prototype provides a recommendation about where virtual bike stations should be located during peak hours and improve general availability by more than 37%. Moreover, when the system's users participate in the relocation of bicycles, the model can eliminate up to 80% of the CO2 emissions and nearly 50% of the operational costs associated with the relocation process.