Bus Arrival Time Prediction Study Based On Data From The Automatic Vehicle Location System Of Multiple Routes

With the improvement of the living standards of urban residents,travelers require a higher level of transit service.As one of the most important traffic information traveler concerns,the arrival time directly affects the attractiveness of public transportation.As for bus operators,real-time and accurate bus arrival time information can reduce delays at bus stops,optimize fleet size and timetables,etc.However,due to the complicated interference factors encountered by the bus during the operation,the existing bus arrival time prediction method based on single route is difficult to gain the desired result.In addition,on the basis of having abundant offline AVL data,the collective information of different bus lines running on the same road link was under used.In view of the above shortcomings,this paper proposes a method for predicting the bus arrival time based on the operation data of multiple routes.And the case study combined with data analysis method are conducted to evaluate the effectiveness of the prediction method using AVL data gathered at Chengdu.The specific work of this paper is as follows:The significant issues of present research in bus arrival time prediction were firstly pointed out by reviewing the literature from the perspective of prediction models and prediction object.And the pre-processing method of AVL data for bus arrival time prediction was established according to the features of historical data from intelligent transit systems of Chengdu.Then,a modeling framework for predicting bus arrival time with fused data from multiple bus routes is built.Specifically,time headways and travel time of all preceding vehicles affiliated with different bus routes running on the same road link were assembled to support arrival time forecasting of a subject vehicle.And the support vector machine(SVM)and feedforward neural network(FNN)algorithm was then employed to fulfill the mission of prediction.A case study was conducted on the road links of Renmin South Road and Shudu Avenue in Chengdu.The test results show that the predictions based on fused operational information of multiple routes outperformed the single-route based model in terms of several performance measures.Besides,the performances of SVM and FNN are assessed and compared for forecasting bus arrival time with multiple routes.Based on the prediction accuracy and practicability of the model,the prediction performance of the SVM is superior to the FNN.The effects of the travel time variability,the number of road segments and the number of bus routes sharing the same road links on the prediction performance of the model are discussed respectively.The results illustrate that there is a strong correlation between the fluctuation of the travel time and the prediction results.And with the increase of number of segments,the prediction accuracy shows an upward trend.As for the influence of the number of bus routes sharing the same road links on the prediction results,current research has not obtained a consistent conclusion,which needs to be further explored.