The expanding network of public bike-sharing system and increasing number of users leads to imbalances in the distribution of bikes causing full or empty station, especially during peak hours. Hence, bike sharing systems need to be properly rebalanced to meet the demand of users and to operate successfully.
Literatures showed several hybrid meta-heutistics to solve Vehicle Routing Problem with Simultaneously Pickup and Delivery. In this study, we present a concept of bike distribution area and combine Location Routing Problem, for determining the least number of distribution centers and vehicles, as well as minimum routing cost simultaneously. The proposed model have an optimization problem in each constraint, each results are interdependence, hence is decomposed into three sub-problems and solved by Simulated Annealing Algorithms.
The test example in model level three shows that if the total demand for each stations is positive, the minimum demand as the vehicle initial carry bikes cause minimum routing distance. If the total demand for each station is negative, the loaded vehicles result in minimum routing distance.
In this study, we use the reality network and coordinate of Taipei YouBike as a numerical example, discussing the trade-off between location of distribution centers, number of vehicles and routing cost. Crucial factors are longest service time of distribution centers and vehicles, vehicles capacity, time window and penalty cost. It is therefore we design five scenarios respectively, discussing the effect on distribution centers, number of vehicles and routing cost. The result shows that it can decrease vehicles by breaking time window, but resulting in lower service level and huge penalty cost. Enhance vehicle capacity leads to a higher vehicle purchase cost, but the increasing numbers of service stations can decrease the number of vehicles. Enlarge distribution centers service scale can increase the number of vehicles, though reducing building cost of distribution centers. The test example shows the optimal vehicle initial carry bikes on different demand, and the proposed model is proved to be operable, therefore can be used by relevant operators for planning dispatch strategy systematically to improve operation efficiency.