Incentives and redistribution in homogeneous bike-sharing systems with stations of finite capacity

Christine Fricker, Nicolas GastEURO J. Transp. Logist 2016 · DOI · OpenReview

TL;DR

This paper models homogeneous bike-sharing systems with finite station capacities to understand how user behavior and operational strategies impact service quality. They find that even with optimal fleet sizing, random user choices lead to significant service disruptions, but simple incentives like directing users to less-loaded stations or truck-based redistribution can dramatically improve performance.

What Problem Does This Paper Address?

Bike-sharing systems (BSS) are increasingly popular for urban mobility, but their effectiveness hinges on managing the availability of bikes and parking spots. A core challenge is the "resource allocation problem": users arriving at a station might find no bikes available, or users trying to return a bike might find no empty spots. This is exacerbated by the inherent stochasticity of user arrivals, trip durations, and destination choices. Furthermore, stations have finite capacities, leading to saturation effects that are often ignored in simpler models. This paper aims to quantify the impact of these factors on system performance and explore how incentives and redistribution mechanisms can mitigate these issues in a simplified, homogeneous setting.

Methods

The authors employ a mean-field approximation to model the BSS. This technique allows them to analyze large systems by treating the stations as statistically identical and their interactions as averaged out. This reduces the complex, discrete stochastic process of individual stations to a deterministic system of ordinary differential equations (ODEs). They then analyze the steady-state behavior of these ODEs to understand the long-term performance of the system. For scenarios involving user incentives or redistribution, they extend this framework to include differential inclusions, which handle the non-smooth dynamics that arise from these mechanisms. The core performance metric used is the proportion of problematic stations, defined as stations with either no available bikes or no available parking spots.

Datasets

Not discussed in the paper. The paper focuses on theoretical modeling and simulation-based validation of the proposed models.

Key Insights and Ideas

  • Randomness is a major performance killer: Even in a perfectly balanced system with optimal fleet size, random user choices (where to pick up/drop off bikes) lead to a significant proportion of problematic stations, inversely proportional to station capacity.
  • Finite capacity is crucial: Ignoring station saturation leads to an overly optimistic view of system performance.
  • "Power of two choices" works wonders: Incentivizing users to choose between two stations based on their load (even if only a fraction of users comply) dramatically improves performance, reducing problematic stations by an exponential factor.
  • Redistribution is effective but costly: Truck-based redistribution can eliminate problematic stations, but the optimal rate is inversely proportional to station capacity, suggesting significant operational costs for high-capacity systems.
  • Geometry matters for detailed metrics, less for overall availability: While local search behavior due to city geometry doesn't drastically alter the proportion of problematic stations, it significantly impacts other metrics like the time to return a bike.

Key Results

  • Optimal Fleet Size: For a homogeneous system, the optimal fleet size (total number of bikes) is approximately $K/2 + \lambda/\mu$ bikes per station, where $K$ is station capacity and $\lambda/\mu$ is the average number of bikes in circulation.
  • Baseline Performance: Without incentives, the minimal proportion of problematic stations is $2/(K+1)$, which decreases slowly with capacity $K$. For $K=30$, this is about 6.5%.
  • Two-Choice Incentive: With the "two-choice" incentive, the proportion of problematic stations can be as low as $\sqrt{K^2 - K/2}$, a significant improvement. Even a small percentage of users adopting this strategy yields substantial gains.
  • Redistribution Rate: The optimal redistribution rate $\gamma^*$ to suppress problematic stations is inversely proportional to station capacity, specifically $\gamma^* = \lambda/(K-1)$ when bikes are moved individually.
  • Trip Time Distribution: The specific distribution of trip times (e.g., exponential, deterministic, log-normal) has a negligible effect on the proportion of problematic stations, as long as the average trip time is the same.

Limitations and Future Work

  • Homogeneity Assumption: The current model assumes homogeneous stations and constant demand. Future work should extend this to time- and space-inhomogeneous systems to capture real-world urban dynamics (e.g., rush hours, residential vs. business districts).
  • Geometric Effects: While simulations show geometry has limited impact on the proportion of problematic stations, it significantly affects other metrics like the time to return a bike. Further investigation into these detailed metrics is needed.
  • Practical Implementation: The paper focuses on theoretical performance and does not delve into the practical costs or implementation challenges of incentives or redistribution mechanisms.
  • Transient Behavior: The analysis primarily focuses on steady-state performance. Studying the transient behavior of these mechanisms, especially in dynamic urban environments, is a promising direction.
  • User Behavior Modeling: The paper assumes users either follow rules or act randomly. More nuanced models of user behavior, including how they might adapt to incentives, could be explored.

Summary updated

Abstract

Bike-sharing systems are becoming important for urban transportation. In these systems, users arrive at a station, pick up a bike, use it for a while, and then return it to another station of their choice. Each station has a finite capacity: it cannot host more bikes than its capacity. We propose a stochastic model of an homogeneous bike-sharing system and study the effect of the randomness of user choices on the number of problematic stations, i.e., stations that, at a given time, have no bikes available or no available spots for bikes to be returned to. We quantify the influence of the station capacities, and we compute the fleet size that is optimal in terms of minimizing the proportion of problematic stations. Even in a homogeneous city, the system exhibits a poor performance: the minimal proportion of problematic stations is of the order of the inverse of the capacity. We show that simple incentives, such as suggesting users to return to the least loaded station among two stations, improve the situation by an exponential factor. We also compute the rate at which bikes have to be redistributed by trucks for a given quality of service. This rate is of the order of the inverse of the station capacity. For all cases considered, the fleet size that corresponds to the best performance is half of the total number of spots plus a few more, the value of the few more can be computed in closed-form as a function of the system parameters. It corresponds to the average number of bikes in circulation.
BibTeX
@article{fricker2016incentives,
  author = {Christine Fricker and Nicolas Gast},
  title = {{Incentives and redistribution in homogeneous bike-sharing systems with stations of finite capacity}},
  journal = {EURO J. Transp. Logist.},
  year = {2016},
  doi = {10.1007/s13676-014-0053-5},
}
Figure