Optimizing On-Demand Bus Services for Remote Areas

Xudong Li, Zhongzhen Yang, Lian FengSustainability 2023 · DOI

TL;DR

This paper introduces a bi-level optimization model to improve on-demand bus services in remote areas by jointly optimizing departure frequencies and ticket prices. The model successfully reduces total generalized costs for both operators and travelers, demonstrating significant improvements in efficiency and user experience for underserved regions.

What Problem Does This Paper Address?

Remote areas often suffer from low population density and limited public transit options, leading to infrequent services, longer travel times, and reduced accessibility. Traditional fixed-route bus services struggle to cater to the dispersed travel demand and time-sensitive needs of residents in these areas. While on-demand transit (ODT) offers a more flexible solution, optimizing its operation, especially considering factors like traveler willingness to pay and overlapping routes, remains a challenge. This paper aims to bridge this gap by developing a model that specifically addresses the unique characteristics of on-demand bus services in remote, low-demand environments.

Methods

The core of this research is a bi-level optimization model.

  • Upper-Level Model (UM): This model focuses on optimizing the on-demand bus service itself. It determines the optimal bus departure frequency across different time windows and sets the ticket prices for on-demand services. The primary objective is to minimize the total generalized cost, which encompasses both the operational costs for the bus companies and the time costs for the travelers. A key consideration here is incorporating travelers' willingness to pay for these services.
  • Lower-Level Model (LM): This model acts as a traveler behavior simulator. It uses a Logit model to predict the probability of travelers choosing between normal bus stops and on-demand stops, based on factors like travel time, walking distance, and ticket prices.

The two models work iteratively: the UM proposes a strategy (frequency and price), and the LM simulates traveler responses. This feedback loop continues until an optimal solution is found.

To solve this complex bi-level problem, the authors employ a genetic algorithm. This is a suitable choice because exact solution algorithms are not practical for such interdependent variables. The genetic algorithm encodes potential solutions (departure frequencies and ticket prices) as chromosomes and iteratively evolves them through selection, crossover, and mutation to find the best-performing solutions.

Datasets

The study utilizes data from a case study in Meishan District, Ningbo City. This data was collected through:

  • Geographic Information: MapInfo software was used to project and overlay geographic information of bus stops and road networks.
  • Questionnaire Survey: A survey was conducted in October 2022 to gather data on residents' demographics, travel choices, walking times, waiting times, and ticket price preferences. A total of 273 valid questionnaires were collected, exceeding the minimum sample size requirement.
  • Traveler Arrival Patterns: Data on the number of travelers boarding buses at normal stops across different time windows was collected to understand arrival patterns.
  • Parameter Calibration: The parameters for the Logit model (α1, α2, α3, α4) were calibrated using data from the questionnaire survey and the NLOGIT 4.0 software.

Key Insights and Ideas

  • Traveler Willingness to Pay is Crucial: The model's success hinges on incorporating travelers' willingness to pay for on-demand services. This acknowledges that users are often willing to pay more for increased flexibility and reduced travel time.
  • Joint Optimization is Key: Simply optimizing departure frequency or ticket price in isolation is insufficient. The paper demonstrates the benefit of jointly optimizing both, as they are interdependent and influence each other's effectiveness.
  • On-Demand Stops Offer Significant Advantages: Even in remote areas, travelers show a strong inclination to use on-demand stops, especially during rush hours. This is attributed to the reduced walking and waiting times compared to traditional stops.
  • Overlapping Routes Can Be Leveraged: The model implicitly handles overlapping routes by considering how travelers might choose between different services at on-demand stops, potentially leading to more efficient service provision.
  • Reduced Generalized Cost is Achievable: The most significant insight is that on-demand services, when optimized, can substantially reduce the total cost for both operators and users, making public transit more viable in low-demand areas.

Key Results

The study's numerical analysis on Meishan Island yielded several positive outcomes:

  • Total Generalized Cost Reduction:
    • Rush hour (7:00–8:00): Reduced by 30.36% (from 2672.6 RMB to 1861.3 RMB).
  • Off-rush hour (11:00–12:00): Reduced by 15.35% (from 1506.2 RMB to 1275.1 RMB).
  • Traveler Time Cost Reduction:
    • Rush hour: Reduced by 37.08%.
  • Off-rush hour: Reduced by 20.33%.
  • Reduced Waiting Times:
    • Average waiting time at on-demand stops: 4.3 minutes during rush hour and 6.8 minutes during off-rush hour.
  • Significant reductions in waiting times at normal stops as well, due to denser departure frequencies.
  • Increased Revenue: The on-demand bus mode also led to an increase in revenue for bus companies, by 15.89% during rush hour and 61.24% during off-rush hour.
  • Traveler Preference: Travelers showed a higher inclination to choose on-demand stops, with percentages reaching 73.75% during rush hour.

Limitations and Future Work

  • Constant OD Traffic Flow Assumption: The study assumes that the total number of bus trips remains constant before and after optimization. This doesn't account for potential shifts in transportation mode choice (e.g., from cars to buses) that might be induced by improved services.
  • Focus on Boarding Demands: The model primarily considers boarding demands at on-demand stops, assuming all travelers alight at normal stops.
  • Fixed Ticket Prices: The model assumes fixed ticket prices for different services, not dependent on travel distance or time.

Future research directions suggested by the authors include:

  • Analyzing the impact of on-demand bus services on residents' overall transportation mode choices.
  • Developing more sophisticated models that account for potential changes in travel demand due to service improvements.
  • Exploring bus operation modes that better meet the actual, dynamic needs of residents.

Summary updated

Abstract

This study proposes on-demand bus services for remote areas with low transit demand, incorporating travelers’ willingness to pay and values of time. To jointly optimize the on-demand service of overlapping bus routes, we construct a bi-level model. The upper-level model (UM) optimizes bus departure frequency in different time windows and ticket prices of on-demand services to minimize the total generalized cost, subject to travelers’ willingness to pay for on-demand services. The lower-level model (LM) calculates the probability of travelers choosing on-demand stops. A numerical analysis based on Meishan Island data in Ningbo indicates that with on-demand bus services, the total generalized cost incurred by buses and travelers can be reduced by 30.36% and 15.35% during rush and off-rush hours, respectively. Additionally, the waiting time at an on-demand bus stop is only 4.3 min during rush hours and 6.8 min during off-rush hours.
BibTeX
@inproceedings{li2023optimizing,
  author = {Xudong Li and Zhongzhen Yang and Lian Feng},
  title = {{Optimizing On-Demand Bus Services for Remote Areas}},
  booktitle = {Sustainability},
  year = {2023},
  doi = {10.3390/su15097264},
}
Figure