关键词: 城市交通, 公交调度, 联合优化, 需求响应接驳公交, 全服务时间窗, 候选站点

Abstract: In the context of demand-responsive feeder transit, it is important to consider the passenger's requirements for the entire journey, including the start and end points, as well as the mobility of people and the characteristics of urban roads. Passengers have the earliest departure time and the latest arrival time to ensure that passengers are delivered to their destination before the service schedule of the following leg. Furthermore, when passengers specify their pickup stops, they often have multiple walkable stops to choose from. Considering these aspects of passenger requirements, this study focuses on designing a demand-responsive feeder transit service model that considers candidate stops and full-service time windows and optimizes the scheduling of the system. Firstly, this paper establishes a mapping between dummy stops and realistic demand and stop sets based on candidate stops for requests. An impedance matrix is constructed, which takes into account node relationships and spatial distances. With the goal of minimizing the start cost and route cost of vehicles, the problem is transformed into a joint optimization problem of fleet size configuration and vehicle routing with a full-service time window and capacity constraint. A mixed integer linear programming model that can be solved using a commercial solver is established. To improve the efficiency of the solution process, it introduces the vehicle number symmetry elimination constraint and the uniqueness constraint for demand sequences at the same stop. These constraints prevent the same solution from being obtained by only exchanging vehicle numbers or sequencing requirements differently at the same stop. To handle larger-scale instances, multiple neighborhood structures and an adaptive neighborhood simulation annealing algorithm based on problem characteristics are designed. Finally, taking the sub-network of transit in Guangzhou University Town as an example, numerical experiments under different influencing factors were designed. The experimental results show that compared with the requests specifying exactly one pick-up stop, the introduction of candidate stops reduces the fleet size and vehicle travel distance required for service, reducing fixed costs and variable costs by 25% and 22.19%, respectively, thereby reducing the total cost by 23.32%. In addition, the full-service time window constraints are stricter than single-point time window and tour time constraints. However, the introduction of candidate stops can compensate for the increased effect of the time windows' tight temporal requirements on route costs and passenger travel time. The results indicate that the full-service time window can ensure the quality of service of passengers, and the introduction of candidate stops can effectively reduce operating costs.

Key words: urban traffic, transit scheduling, joint optimization, demand responsive feeder-transit, full-service time window, candidate stops