On the simultaneous computation of target inventories and intervals for bimodal bike-sharing systems

The emerging demand for electric bicycles in recent years has prompted several Bike-Sharing Systems around the world to adapt their service to a new wave of commuters. Many of these systems have incorporated electric bikes into their network while still maintaining the use of regular mechanical bicycles. However, the presence of two types of bikes in a Bike-Sharing network may impact how rebalancing operations should be conducted in the system. Regular and electric bikes may exhibit distinct demand patterns throughout the day, which can hinder efficient planning of such operations. In this paper, we propose a new model that provides rebalancing recommendations based on the demand prediction for each type of bike. Additionally, we simulate the performance of our model under different scenarios, considering commuters’ varying inclination to substitute their preferred bike with one of a different type. Our empirical experiments indicate the potential of our model to improve user satisfaction, reducing the total lost demand by approximately 10%, while reducing the lost demand for electric bikes by around 30%, on average, when compared to the existing rebalancing strategy used by the real-world Bike-Sharing System under study. Remarkably, this was accomplished while maintaining an almost identical average hourly count of rebalancing operations.