TL;DR
This paper proposes a novel methodology for designing customized bus (CB) networks by leveraging online car-hailing (OCH) data. It introduces a multi-step process that includes data processing, demand clustering, initial line generation, and a line selection model to balance operator, social, and passenger interests, offering a more efficient and data-driven approach to urban transit planning.
What Problem Does This Paper Address?
Traditional public transport often struggles with issues like overcrowding, low punctuality, and long travel times, leading passengers to opt for private cars, taxis, or OCH. While OCH offers convenience, it contributes to increased congestion and environmental pollution. Customized Bus (CB) systems aim to bridge this gap by offering a more personalized and efficient public transit option. However, designing effective CB networks, particularly the crucial "line planning" aspect, is complex. Existing methods often rely on less detailed questionnaire data or don't fully account for diverse vehicle capacities and operational costs. This paper addresses the need for a robust, data-driven CB network design methodology that utilizes readily available OCH data to optimize routes, balance stakeholder interests, and improve urban mobility.
Methods
The paper outlines a four-step CB network design methodology:
- Individual Reservation Travel Demand Data Processing: This involves cleaning and screening OCH trip data, focusing on origin-destination (OD) pairs, boarding/alighting times, and passenger mileage. It uses ArcGIS to convert geographic coordinates into planar coordinates for further analysis.
- CB Line OD Area Division: Agglomerative Hierarchical Clustering (AHC) is employed to group passenger origins and destinations into distinct "OD areas." This clustering considers constraints on demand quantity within areas and distance limitations to define logical service zones for bus stops.
- Initial Set of CB Lines Generating: Based on the clustered OD areas, potential CB lines are generated by pairing origin and destination areas. Lines are filtered based on minimum demand and length constraints to create an initial set of viable routes.
- CB Line Selection Model Building: A generalized cost objective function is developed to select the most suitable lines. This function balances the company's operating costs, environmental costs, traffic congestion costs, and passenger costs. The model considers multiple CB vehicle types (e.g., 49-seat, 18-seat buses) and uses a discrete optimization approach (branch-and-bound) to determine the optimal network.
Datasets
The study utilizes real-world OCH trip data from Beijing, China. Specifically, data from 27 July 2017 (Monday) was used, focusing on the morning rush hour (70,537 trips between 08:00 and 09:00). After initial filtering, 44,043 trips were retained for planning. The data includes fields such as order number, boarding/alighting locations (longitude and latitude), times, and passenger mileage.
Key Insights and Ideas
- Leveraging OCH Data: The paper demonstrates the feasibility and value of using OCH data, which is more granular and dynamic than traditional survey data, for CB network planning.
- Hierarchical Clustering for OD Zoning: AHC proves effective in grouping passenger origins and destinations into meaningful areas, which is crucial for defining efficient CB routes and stop locations.
- Multi-objective Optimization: The line selection model's strength lies in its ability to balance competing interests: operator profitability, societal benefits (reduced pollution/congestion), and passenger satisfaction.
- Vehicle Mix is Key: The study highlights that a combination of different-sized CB vehicles (e.g., 49-seat and 18-seat) is generally more cost-effective than using a single vehicle type.
- Capacity vs. Cost: Lower-capacity CBs tend to be more cost-effective, suggesting a trade-off between passenger volume and operational efficiency.
Key Results
- The combination of CB vehicles with 49 seats and 18 seats was found to be the most cost-effective across various scenarios.
- CBs with lower capacity were generally more cost-effective than those with larger capacities.
- Prioritizing environmental pollution and congestion issues in decision-making leads to better service acceptance rates and spatial coverage for the CB network.
- The service acceptance rate and spatial coverage of the CB network increase with the fixed operating cost per OCH vehicle, while CB vehicle usage decreases.
- The study generated several operating schemes for the Beijing CB network, identifying specific high-demand routes.
Limitations and Future Work
- Mode Choice Assumption: The current model assumes passengers will travel by either CB or OCH, not considering other modes or a more complex travel mode selection process.
- Fare Influence: The impact of different fare structures on CB line planning was not explicitly discussed and is suggested for future research.
- Further Refinement: The authors suggest incorporating other transport modes into the model and developing more precise travel mode selection models to capture all potential CB passengers.
Summary updated