Research On Bus Arrival Time Prediction Based On Particle Filter

Bus arrival time is one of the most concerned traffic information for travellers. Providing accurate bus arrival time can help to improve the service quality of a transit system, enhance bus attractiveness, ease traffic jams and promote urban development.Firstly, based on the existing prediction methods and the factors influencing bus arrival time, this paper designs a complete set of processing method for GPS data and lines data. Then, the average speed at intervals and dwell time at stops are obtained through the C#, Maplnfo and SQL Server, providing a necessary data basis of the model variables.Secondly, this paper introduces the basic principle of particle filter (PF) systemically. Concerning that PF can be well applied to nonlinear and non-Gaussian systems, this paper exploratively establishes the bus arrival time prediction model based on PF, and introduces weight coefficient into the model in order to integrate the average spped at intervals and the instaneous speed with the expection to enhance prediction accuracy of the model and the algorithm.Finally, this paper chooses several typical bus lines in Beijing as analysis demonstration, with a mean absolute error (MAE) to evaluate the effectiveness of indicators. The selected lines are 1,2,300 Clockwise Loop,345 Express,438 and 464 and then bus arrival time are predicted at morning peak hour (8:00), off-peak hour (11:00) and evening peak hour (17:00). For further evaluation of the algorithm, taking morning peak hour (8:00) as an example, different methods including PF and Kalman filter (KF), are used to predict the bus arrival time. The results show that PF has better stability and higher accuracy, and the results based on PF are improved by 22.87% than that based on KF.