A novel bilevel programming model for designing feeder bus routes to simultaneously consider bus stop selection, bus routing, and passenger route/trip choice behavior is developed in this work. The objective of the upper-level model is to minimize the total in-vehicle travel time of passengers. The objective of the lower-level model is to determine bus stop locations and assign passenger demand to stops by minimizing the total passenger walking time. An ant colony optimization-based two-stage heuristic algorithm is developed to solve the bilevel programming model within an acceptance computation time. Different from conventional methods, the proposed methodology benefits from using real temporal and spatial characteristic datasets of travel demand aggregated from cellular data and traveling time and distance matrices under actual traffic conditions resulted from an open geographic information system (GIS) tool. Results of a real-life case study in Chongqing municipality, China, show that the proposed methodology could achieve a significant reduction of 2.9% total passenger travel time compared with a traditional method. Finally, sensitivity analyses are conducted to further understand the performance of the model. It is anticipated that the proposed methodology could be very useful in designing attractive and cost-effective feeder transit service in practice.

Ride-sharing (RS) plays an important role in saving energy and alleviating traffic pressure. The vehicles in the demand-responsive feeder transit services (DRT) are generally not ride-sharing cars. Therefore, we proposed an optimal DRT model based on the ride-sharing car, which aimed at assigning a set of vehicles, starting at origin locations and ending at destination locations with their service time windows, to transport passengers of all demand points to the transportation hub (i.e., railway, metro, airport, etc.). The proposed model offered an integrated operation of pedestrian guidance (from unvisited demand points to visited ones) and transit routing (from visited ones to the transportation hub). The objective was to simultaneously minimize weighted passenger walking and riding time. A two-stage heuristic algorithm based on a genetic algorithm (GA) was adopted to solve the problem. The methodology was tested with a case study in Chongqing City, China. The results showed that the model could select optimal pick-up locations and also determine the best pedestrian and route plan. Validation and analysis were also carried out to assess the effect of maximum walking distance and the number of share cars on the model performance, and the difference in quality between the heuristic and optimal solution was also compared.