Shared Subway Shuttle Bus Route Planning Based on Transport Data Analytics

• Published in: IEEE transactions on automation science and engineering 2018-10, Vol.15 (4), p.1507-1520
• Main Authors: Kong, Xiangjie, Li, Menglin, Tang, Tao, Tian, Kaiqi, Moreira-Matias, Luis, Xia, Feng
• Format: Article
• Language: English
• Subjects: Analytics; Buses; Crowdsourced data; Crowdsourcing; Data analysis; Dynamic programming; Machine learning; Optimization; passenger flow prediction; Planning; Predictions; Predictive models; Public transportation; Route planning; shared buses; Taxicabs; Traffic congestion; Urban transportation; User experience; Vehicle routing; Volatility
• ISSN: 1545-5955; 1558-3783
• DOI: 10.1109/TASE.2018.2865494

The development requirements of shared buses are extremely urgent to alleviate urban traffic congestions by improving road resource utilization and to provide a neotype transportation mode with good user experiences. The key to shared bus implementation lies in accurately predicting travel requirements and planning dynamic routes. However, the sparseness and the high volatility of shared bus data bring a great resistance to accurate prediction of travel requirements. Based on the consideration of user experiences, optimization objectives of shared bus route planning are significantly different from traditional public transportation and shared bus route planning is far more challenging than online car-hailing services due to the relatively high number of passengers. In this paper, we put forward a two-stage approach (SubBus), which is composed of travel requirement prediction and dynamic routes planning, based on various crowdsourced shared bus data to generate dynamic routes for shared buses in the "last mile" scene. First, we analyze the resident travel behaviors to obtain five predictive features, such as flow, time, week, location, and bus, and utilize them to predict travel requirements accurately based on a machine learning model. Second, we design a dynamic programming algorithm to generate dynamic, optimal routes with fixed destinations for multiple operating buses utilizing prediction results based on operating characteristics of shared buses. Extensive experiments are performed on real crowdsourced shared subway shuttle bus data and demonstrate that SubBus outperforms other methods on dynamic route planning for the "last mile" scene. Note to Practitioners -This paper is inspired by the problem of shared subway shuttle bus dynamic route planning for the "last mile" scene, and it is also applicable to other scenes, including commuting scenes, urban transportation hub scenes, and destination scenes of the tourist market. Shared bus operation routes at such scenes are usually aimed at trips with fixed destinations. Existing approaches to planning routes are generally designed for traditional transportation, such as traditional buses and taxis. In this paper, we propose a novel two-stage dynamic route planning approach (SubBus) based on the operation characteristics of shared subway shuttle buses. We perform a resident travel behavior analysis to improve the accuracy of travel requirement prediction. After that, we combine the prediction results and st