Feeder Bus Route Design and Vehicle Allocation Under Influence of Shared Bikes

For the "first-and last-mile" of rail transit, feeder buses and shared bikes are two most prevalent modes to provide connection with rail transit for commuters. To understand the impact of bike-sharing on the planning and operation of feeder bus travel demand and route design, this study examines the feeder bus route design and vehicle allocation challenges based on the interaction of demand and supply. From the demand side, the actual travel demand of feeder buses is dynamically estimated depending on the user's mode choice between shared bikes and feeder buses, considering the travel time and travel cost. Comparatively, a mixed-integer non-linear programming model with the objective of minimizing the sum of bus operating cost and user travel cost is developed from the supply perspective to optimize the bus route design and vehicle allocation, including vehicle capacity, vehicle quantity, and flow balance constraints. The Lagrangian relaxation algorithm is used to solve the model. This strategy is applied to the planning of feeder bus routes in the Beijing suburbs surrounding the Huilongguan Metro Station. The actual smart card data and Mobike cycling data are used to obtain the total travel demand. The travel time by various modes between stops is derived from the AutoNavi route planning API (Application Programming Interface). In the case where the total number of vehicles is 10 and the number of lines is 2, the experimental results show that the difference between the assumed bus travel demand and the computed bus ridership can be effectively avoided if the influence of bike-sharing on bus travel demand is considered. The average running time between each bus stop and the station is 15.58 minutes, while the average passenger waiting time is 3.35 minutes. In the case where there are four lines, the average running time from each bus stop to the station is 8.53 minutes, which is almost half of the case with only two lines; the average waiting time for passengers is 3.44 minutes. Nonetheless, the computing time for the model grows exponentially with the increasing of the number of lines. Consequently, from the perspective of model calculation efficiency, both scenarios in which the number of lines is set to 2 or 3 can satisfy the application requirement of updating lines every half hour. © 2023 Science Press. All rights reserved.