关键词: 城市交通, 需求响应公交, 接送策略, 线路优化, 精英遗传算法

Abstract: With the continuous advancement of information and internet technology, reservation-based travel has emerged as a prominent trend in developing public transport services. The demand responsive bus has gained significant attraction in various regions of China. These buses generate customized routes with flexible stops and schedules, catering to the personalized needs of passengers based on their travel reservation information. The demand responsive bus enhances the quality of public transportation services and mitigates the resource wastage associated with fixed bus schedule, making it particularly suitable for serving short-distance regional trips and accommodating weak spatiotemporal regularity travel patterns. The demand responsive bus service is divided into three components: demand travel information collection, route generation, and service provision. In particular, route generation is the most critical component in demand responsive bus service, which directly affects system ability and service quality. Due to its importance, researchers have dedicated their efforts to studying demand responsive bus routing problem. However, current research remains largely focused on routing optimization models using either a unitary pickup and delivery strategy or a single vehicle type. In fact, these simple operational methods are not enough to accommodate the complexity and variability of actual demand patterns. To overcome this gap, we attempt to incorporate diverse pickup and delivery operation strategies, which is beneficial for not only increasing the flexibility advantages of demand responsive bus, but also enriching operational methods to meet diversified service needs. Furthermore, we introduce multiple capacity vehicle types into demand responsive bus routing optimization, which facilitates a better allocation of passengers with various vehicle resources and provides more space for route optimization.
Hence, we propose a demand responsive bus routing optimization problem with multiple capacity vehicle types, incorporating two operational strategies: the pickup and delivery separation strategy and the pickup and delivery mixture strategy. In order to take into account both the interests of passengers and the bus company, we minimize the system cost as the objective, which comprises the cost of measuring service quality and operational costs. In the model, we also take buffer time, minimum load rate, and detour coefficients constraints into account. As a typical NP-hard problem, the difficulty of solving the model increases exponentially as the scale of the problem grows. To accelerate the solution, we propose a decomposition method that combines an elite genetic algorithm and solver. Subsequently, we take the demand responsive bus stops in the Wangjing area of Beijing as a case study, and generate randomly travel demands. We evaluate the service quality by applying the pickup and delivery separation strategy and the pickup and delivery mixture strategy across various demand scenarios, buffer times, detour coefficients, and fleet sizes.
The experiments results demonstrate what follows: (1)The pickup and delivery separation strategy is effective in reducing system operating costs. Compared to the pickup and delivery mixture strategy, the average operating cost is reduced by up to 27.5%. (2)The pickup and delivery mixture strategy is conducive to improving both the demand service rate and the average vehicle occupancy rate, which can significantly increase the system service rate by 24.5%, and the average vehicle occupancy by a maximum of 10.7% compared to the pickup and delivery separation strategy. (3)The buffer time is an essential feature affecting the system service rate and the average vehicle occupancy. In particular, when the system service rate has reached 1.0, continuing to prolong the buffer time can also contribute to the re-assignment of passengers to improve the average vehicle occupancy. (4)The vehicle detour coefficient, as one of the factors on the system service ability and total cost, has a crucial threshold of 2.0 in the experiment. When the detour coefficient exceeds the threshold, the system service ability and total cost remain constant. (5)The fleet size has a significant impact on the overall cost of the system. A small vehicle fleet leads to poor service ability, while a large fleet may result in a reduction of vehicle utilization. Furthermore, different pickup and delivery strategies exhibit their respective preferences for high-capacity vehicles when vehicle resources are sufficient in the case.
In conclusion, this paper offers valuable methodologies and a theoretical foundation for the practical design of demand responsive bus routes. It extensively examines the practical operation effectiveness of both the pickup and delivery separation strategy and the pickup and delivery mixture strategy in the responsive bus routing optimization model. At the same time, it also investigates the effects of incorporating multiple capacity vehicle types in the routing optimization problem. Moreover, the paper conducts several sensitive experiments to explore the impact of different detour coefficients and vehicle fleet sizes. In the future, we will be able to quantitatively evaluate the effectiveness of different pickup and delivery strategies, based on real reservation information and multi-source traffic data, and provide more customized demand responsive bus service with detailed demand scenarios.

Key words: urban transportation, demand responsive bus, pickup and delivery strategies, routing optimization, elite genetic algorithm