Assessing the operational impact of tactical planning models for bike-sharing redistribution

Station-based bike-sharing systems provide users with inexpensive one-way bike rides. A major challenge for operators of BSSs lies in redistributing bikes so that users may take the bike rides they request. Existing research on tactical planning proposes optimization models for designing redistribution plans that a vehicle fleet implements on a daily basis. The purpose of this paper is to identify the value and limitations of stochastic programming for bike-sharing redistribution and to understand the efficacy of the obtained plans once they are implemented. To this end, we first analyze the variability in recorded ride data from three North American bike-sharing systems which mainly differ in the intensity of commuting. The results of the data analysis show that stations that are mainly used by commuters display less variability in demand than stations that are mainly used for other ride purposes like errands and leisure. To assess the effect of demand variability on the operational implementation of redistribution plans, we rely on agent-based simulation. In the simulation, vehicle tours are implemented as planned. However, since redistribution plans are designed based on demand forecasts of ride requests, guidance is needed about how to adjust the bike quantities to pick up from or deliver to each station when actual ride requests are observed. Therefore, we propose rule-based procedures to adjust redistribution decisions when the numbers of bikes at stations and vehicle loads differ from the setting considered by optimization. We show that demand variability is a leading indicator about whether redistribution plans perform well operationally.

Automated summary

This paper examines the redistribution issue in station-based bike-sharing systems and analyzes the demand variability of different types of stations. The impact of demand variability on the operational implementation of redistribution plans is evaluated through agent-based simulation.