Study On Bus Arrival Time Prediction Based On GPS Data

The development of public transportation system (PTS) is the most efficient approach to resolve the traffic problems in a metropolis. PTS information can improve the work efficiency and service level and accelerate the management transfer from simple static pattern to intelligent dynamic one.As one of the most concerned public transportation information, bus arrival time plays a significant role in reducing consumers'anxiety of waiting, saving travel time and dispatching buses effectively. Therefore, it is necessary to develop the bus arrival time prediction based on advanced information and communication technology in order to improve the manangement and service level and thus implement intelligent PTS.According to the analysis of three typical prediction models: models based on historical data, multi-linear regression models and artificial neural network models, this paper presents an algorithm to predict the expected bus arrival times at individual bus stops along a service route based on the technology of map matching and traffic detection. This proposed prediction algorithm combines GPS data with linearize route and real-time average travel speeds of individual route segments. To reduce the measurement errors of GPS data and improve the prediction accuracy, it also presents pretreatment and parameters selection methods and prediction regulations.Sample prediction algorithm can't detect the running situation and take action to unexpected incident. Based on the proposed prediction algorithm, this paper therefore designs an intelligent prediction system, which divides the whole buses'running into Beginning Status, Prediction Status, Terminus Status and Unknown Status. And all service bus can be managed due to the theory of Object-Oriented. Six key modules make the system implement determining of bus service route and running direction, predicting arrival time, detecting abnormal turning and switching among states.Finally, the proposed algorithm and system are tested by the data from Route 871 of Chongqing. The experiments demonstrate the prediction algorithm outperforms tapical methods in terms of accuracy and this accuracy improves and becomes stable gradually as the distance increases. For the prediction system, the performance of detecting service route and running direction is up to 98% and 100%. It's proved that this system cannot only manage buses, but also make the prediction algorithm efficient.