Research On The Method Of Bus Arrival Time Prediction

Bus arrival time is one of the most important manifestation of public transportation intelligence.Improving the accuracy of bus arrival time prediction plays an important role on improving public transportation service level,easing traffic congestion,reducing passenger travel cost,realizing the information of public transport system.Firstly,this paper analyzed the bus operating data acquisition principle,method and characteristics,and designed the GPS data interpolation algorithm and bus lines discretization algorithm to deal with bus operating data.Through analyzed the bus operation process and the influence factors of the arrival time,this paper chose the dwell time and interval average speed as the input variables,and designed the algorithm which can obtain input variable.Secondly,this paper used dwell time and interval average speed as input variables,the Statistical Method Bus Arrival Time Prediction Model Based on Interval Length(SMBATP-IL),Kalman Filter Bus Arrival Time Prediction Model Based on Interval Length(KFBATP-IL)and Particle Filter Bus Arrival Time Prediction Model Based on Interval Length(PFBATP-IL)were established,and the concrete process and steps of implementing the algorithm were designed.Taking the PFBATP-IL model as a case to prove the feasibility of the proposed method and model in this paper.Finally,this paper analyzed 2 empirical bus lines,and used an average absolute error MAE as a measure of forecast results,which chose morning peak(8:00),off-peak(11:00)and evening peak(17:00).In different interval length(10 m,20 m,30 m)conditions,this paper used three models in this article to predict arrival time.Results showed that the different interval length have minimum influence on PFBATP-IL model,and KFBATP-IL model is middle,and have maximum influence on SMBATP-IL model.Under the optimal interval length,the predicted results of BATP-PF model is optimal,compared with predicted results of BATP-KF model and BATP-SM model improved by 16.86% and 28.46% respectively.