Multi-urban Night Bus Routes Planning

With the development of urbanization, the urban population increase. More and more people travel at late night and the time of the night that people travel is continue to lengthen. The existed public transportation system was designed with human surveyed passenger flows and user demands data. Bus transportation system is out of service at late night, which brings a lot inconvenience to the people traveling at late night. This paper is aim to solve the problem of people travel difficult at late night. We propose a novel method to plan the night bus system to fit the needs of people traveling at night. Nowadays on-board equipment is extremely rich, such as vehicle mounted GPS positioning equipment and vehicle monitoring equipment. We can obtain some information of the car, such as speed, taxi passenger status information.And we can understand the flow of the citizen. To plan Urban night bus routes system,we should know the flow of people travel at night. We can use the taxi GPS data to plan the night bus routes system because taxi is the main transportation at late night.The night bus routes planning method that we described in this paper considers many aspects, such as construction and operating cost, the capacity of the night bus transportation system. Though we want the capacity of the bus system to be more, but it will bring more cost. If we just consider the capacity of the bus system, it may cause some waste. To trade off all aspects, we propose a multi-aspects night bus routes planning method to solve the problem. In order to make full use of existing bus system resources, we use the existing bus stops as our night bus routes stops. The redundancy of the origin bus stops will be removed and we cluster the left stops to choose the night candidate bus stops. We score the bus stops which are the endpoint stops of existed bus routes in many aspects. The selection probability of each stop is calculated by the score of each stop. Then we select the first stop of one candidate bus route with the selection probability. The selected first stop is used to build the night bus routing graph.After the graph is built, we execute the pruning operation to reduce the space cost. then we generate one candidate night bus route in the graph. When one night candidate bus route is generated, we add it to the candidate night bus routes set and we update the night passenger flow network until the routes set generation is complete. We score the generated candidate night bus routes set to select the best set and then we map the set to the real geographical region. It is verified that our proposed method outperforms the Top-K approach in the planning of best multi-aspects trade-off bus route set.