TL;DR
This paper tackles the challenge of optimizing customized bus routes in rural areas, where traditional public transport struggles with low demand and dispersed populations. They propose a two-stage model that balances operator costs and passenger travel time, incorporating both pre-booked and real-time requests, and introduce a novel hybrid algorithm (BA-APSO) to find efficient solutions.
What Problem Does This Paper Address?
Rural areas often suffer from inadequate public transportation. Low population density, dispersed travel origins and destinations, and fluctuating demand make fixed-route bus services economically unviable, leading to high operating costs, frequent empty buses, and long passenger wait times. This paper addresses this by focusing on "demand-responsive customized bus" (CB) services, which aim to provide flexible, on-demand transportation tailored to passenger needs. The core problem is how to efficiently plan the routes for these CB services, considering both the operator's costs (fuel, vehicle usage) and the passengers' travel time, while also accounting for real-time changes in demand and various operational constraints.
Methods
The authors propose a two-stage optimization model:
- Reservation Phase (Static): This stage deals with pre-booked rides. A vehicle route optimization model is formulated to minimize a weighted sum of operator costs and passenger travel time costs. Key constraints include passenger time windows, vehicle capacity, and route continuity.
- Real-time Phase (Dynamic): This stage incorporates passengers who request rides during the operation of the buses. The model is adapted to include penalty costs for refusing real-time requests and for exceeding tolerable time windows, aiming to improve responsiveness and operational efficiency.
For solving this complex, multi-objective optimization problem, the paper introduces a hybrid algorithm combining the Bat Algorithm (BA) and Adaptive Particle Swarm Optimization (APSO).
- The Bat Algorithm (BA) is known for its fast convergence and strong global search in the early stages but can get stuck in local optima later.
- Particle Swarm Optimization (PSO) offers high search accuracy and optimization capability but has weaker global search.
- The hybrid BA-APSO algorithm leverages the strengths of both. In the early stages, it uses BA with modifications inspired by APSO's inertia weight to maintain early search effectiveness. In later stages, it transitions to an APSO-based approach with adjusted learning factors to improve search accuracy and convergence speed, aiming to overcome the limitations of each individual algorithm.
Datasets
The study uses an actual road network from some townships in Jing'an County, Jiangxi Province, China, as its basis for empirical analysis.
- Road Network Data: Distances between stations were obtained using a map.
- Passenger Information: Pick-up/drop-off stations and expected boarding time windows for passengers were generated using software, with the total number of passengers based on actual survey data. This is noted as a limitation, with a suggestion for future work to use real passenger flow data.
Key Insights and Ideas
- Rural Public Transport Needs a Flexible Approach: The paper strongly advocates for demand-responsive customized bus services as a superior alternative to fixed-route buses in low-density rural areas.
- Two-Stage Optimization is Crucial: Effectively managing both pre-booked and last-minute travel requests requires a model that can adapt dynamically, balancing immediate needs with planned routes.
- Hybrid Algorithms Offer Synergistic Benefits: Combining algorithms like BA and APSO can lead to better overall performance than using either alone, addressing their individual weaknesses in convergence speed and global search capability.
- Balancing Operator and Passenger Interests: The objective function explicitly considers both minimizing operator costs and passenger travel time, highlighting the importance of a holistic approach to public transport planning.
Key Results
The proposed optimization model and the BA-APSO hybrid algorithm were tested on the Jing'an County road network.
- Static Phase Optimization: The hybrid algorithm achieved an optimal objective function value of 2117, using five cars and two models.
- Algorithm Comparison: Compared to standalone BP (a variant of PSO) and APSO algorithms, the BA-APSO hybrid algorithm demonstrated superior performance:
- It achieved the lowest target value (2117).
- It reduced the total driving distance by 8 km and total travel time by 9 min compared to APSO.
- It increased the average load factor by 18.8% compared to BP.
- Dynamic Phase Performance: In a dynamic phase simulation with 8 passengers, the algorithm successfully responded to 7 of them, with a maximum delay of 12.1 minutes, which was deemed acceptable.
- Overall Improvement: The paper states that the optimization model and hybrid algorithm reduced the objective function value by 5.5% compared to previous methods.
- Comparison with Regular Bus: A comparison with a hypothetical regular bus service showed that the demand-responsive customized bus service reduced travel distance by 248.4 km, travel time by 186.7 min, and operating costs by 2529.2 yuan for 81 passengers.
Limitations and Future Work
- Assumptions on Vehicle Speed: The study assumes a fixed vehicle speed, which is unrealistic in practice due to varying road conditions, environment, and weather. Future work should incorporate dynamic speed adjustments.
- Passenger Data Generation: Passenger information was randomly generated. The authors suggest using real passenger flow data for more accurate analysis.
- Focus on Passengers: The current model primarily focuses on passenger demand. Future research could integrate passenger, cargo, and mail transportation to develop a more comprehensive rural transport model.
- Limited Scope of Comparison: While comparisons were made with BP and APSO, a broader comparison with other state-of-the-art algorithms could further validate the proposed hybrid approach.
Summary updated