Curvature-based distribution algorithm: rebalancing bike sharing system with agent-based simulation

Abstract

The selected rebalancing strategy determines the service results for bike sharing system (BSS). All cities have different operators, such as the number of stations, bikes, and rebalancing trucks, traffic congestion patterns, terrain maps, and maintenance costs. Therefore, each city needs a rebalancing strategy of its own. However, identifying an appropriate rebalancing strategy requires a method that can simulate a large volume of rebalancing operators. In this paper, we propose a novel method to dynamically rebalance BSS with agent-based simulation. We developed a curvature-based distribution algorithm which allows simulations with a large number of rebalancing trucks and stations (40 trucks and 581 stations), compared to previous studies on bike rebalancing. The algorithm uses a three-dimensional (3D) terrain map created with a large volume of BSS usage data (3,595,383 observations) to generate dynamic truck routes for agent-based simulation. In addition, traffic congestion data were used as weighting values to improve the accuracy of the proposed method for truck route generation. Thus, the proposed algorithm can test different rebalancing strategies that are suitable for a specific city with various spatiotemporal conditions. Specifically, the system analyst can simulate different truck numbers, working ranges, and working hours to identify a suitable strategy for various cities to estimate yearly budget. The algorithm is simple to implement and adaptive to various bike usage data. The research also proposes a visual analysis method based on simulated results of rebalance imbalance metrics, service failure rate, number of daily operations, daily truck travel distance, and stochastic bike usage behavior.