A Simulation-optimization Approach to Design Efficient Systems of Bike-sharing

Abstract

This article presents a methodology to simultaneously modelize private car and public bicycle transport modes, considering their interactions through the modeling of the modal split and the assignment of each mode's trips to the network. Then the model is used to optimize the location of docking stations in a public sharing bicycle system, to achieve a transport system as efficient and sustainable as economically and socially possible. To model the car-bike system, it was necessary to perform a user behavior analysis, emphasizing the characterization of the potential bicycle user. Our objectives were: to discern which variables does the user take into account to choose, in an urban environment, between public bicycle and car; and to modelize bicycle users’ route choice. The proposed methodology is divided into three parts. In the first one, a transport network was created, over which car and bicycle users can be simulated at the same time. The second part is a type of combined modal split-assignment model, with its input coming from a global matrix (car and bicycle trips). A Multinomial Logit model for modal split and network assignment models will follow an iterative process to provide the final OD matrices and service variables for each mode of transport. After the definition and validation of the multimodal model, the third part, an optimization model, has been developed, based on bi-level mathematical programming. The objective is to optimize the location of a system of public bicycle docking stations. A combination of the different optimization criteria and problem reduction strategies has been applied to the real scenario of Santander city (Spain). It has been possible to perform a comparative analysis of the current situation, versus the model's proposal, getting results helpful to the relevant decision–making entities.