This work aims to investigate the dispatching of flex-route transit (FRT) based on the forecasted demand probabilities by integrating history data and real-time appointment information. An operating FRT line is used as the studied object, where massive historical data of dynamic origin-destination (OD) have been obtained. To address the uncertainty of demand, a prediction probability model of each OD pair using the deep autoregressive recurrent networks (DeepAR) method is constructed based on historical data. Then, a model to adjust probabilities is constructed using Bayesian inference while reservation information is added. An optimal dispatching strategy for FRT is proposed, where a departure timetable is obtained by Monte Carlo simulation and the predicted probabilities of demand. It is updated with real-time newly added reservation information. The model of planning departure timetables based on expected demand is established. The rules for adjusting the departure timetable are designed based on new advanced appointments. The optimization model for bus route operation is established. The method of combining genetic algorithm and Monte Carlo simulation is adopted in the concrete optimization solution. The proposed method is validated through a sample application of Bus Route 385 in Tianjin City, China. The results show that the profit of public transport enterprise is increased by 35.81%, 53.44%, and 27.14%, respectively, and the generalized travel cost of passengers is reduced by 9.29%, 12.05%, and 5.21%, respectively, compared to the method only using real-time reservation information, the method of demand probabilities using frequency estimation based on historical data, and the demand prediction method using long short-term memory networks. This approach enables a more accurate prediction of demand, resolving the contradiction between the profit increase of enterprise and the decrease of generalized travel cost of passengers.
Dispatching of Flex-Route Transit Based on Forecasted Demand Probabilities Using DeepAR and Bayesian Inference
J. Transp. Eng., Part A: Systems
2025-08-01
Article (Journal)
Electronic Resource
English
Evaluation of OmniLink Demand-Driven Transit Operations: Flex-Route Services
Online Contents | 1998
|Examining individuals’ adoption of flex route transit
Taylor & Francis Verlag | 2024
|Comparison and selection of feeder transit between flex-route transit and fixed-route transit
British Library Conference Proceedings | 2021
|