Anticipatory Service Network Design of Bike Sharing Systems

Abstract

Today‘s conurbations suffer from inefficiency in transportation systems. Bike sharing systems (BSS) combine the advantages of public and private transportation to better exploit the given transportation infrastructure. They provide bikes for short-term trips at automated rental stations. However, spatio-temporal variation of bike rentals leads to imbalances in the distribution of bikes, causing full or empty stations in the course of the day. Ensuring the reliable provision of bike and free bike racks is crucial for the viability of these systems. On the tactical planning level, target fill levels of bikes at stations are determined to provide reliability in service. On the operational planning level, the BSS operator relocates bikes in vehicles among stations based on target fill levels. A recent approach in tactical service network design (SND) anticipates relocation operations of BSS by means of a dynamic transportation model yielding the required demand of relocation services (RS). A RS is described by pickup and return station, time period, and the number of relocated bikes. RS represent the design decision for implementing a service between two stations in each period for each day of the system operation. The output of the SND model are the time-dependent target fill levels at stations and the set of cost-efficient RS to facilitate these target fill levels. However, the existing approach neglects the sequence of RS into tours, thus leading to a weak anticipation of operational decisions. We extend an existing SND approach by including the concept of service tours (ST). RS are sequenced in ST which start and end at the depot. Experiments shows that the ST obtained by the extended SND yield a stronger anticipation of operational decisions.