Research On The Prediction Of Bus Arrival Time Under Bus Priority Environment

With the acceleration of urbanization in Inner Mongolia Autonomous Region,people’s travel demand and the increase of motor vehicle ownership have brought great challenges to the urban road transportation capacity.In response to the national call,the autonomous region government advocates the strategy of giving priority to public transportation.On the one hand,it realizes "space priority" by setting up bus lanes;on the other hand,it researches the technology of cooperating public transportation and signal lights to realize "signal priority".Among them,the prediction of bus arrival time at the intersection is an important reference basis for realizing "signal priority".In this paper,buses with bus lanes are selected,and the bus GPS track data is taken as the research object to predict the time of bus arrival at the intersection,and then the time prediction model is proposed.Firstly,the processing method of bus GPS track data was introduced,including outlier filtering,range limiting,location matching,target vector acquisition,etc.The bus running speed,section travel time,intersection delay and other running characteristics were analyzed statistically,the bus running characteristics were studied,and the spatio-temporal relationship between the vehicle travel track and the line intersection was excavated.Secondly,according to the research idea of dividing the vehicle running time into road section driving time and waiting delay time at the intersection,relevant research was carried out.A travel time prediction model based on time series fusion converter(Temporal fusion transformer)was proposed.The method integrated the encoder framework and the attention mechanism,and realized feature screening and extraction by processing the multi-variable information in time series data,and realized the high performance prediction of the model.A Light GBM bus intersection delay time prediction model was proposed based on particle swarm optimization algorithm.Particle swarm optimization was used to search the optimal solution in the multi-dimensional hyper-volume to complete the parameter configuration of the Light GBM model,which improved the accuracy of the model prediction.Finally,the two models constructed in this paper were verified with real data.The GPS track data of real bus operation is selected for screening processing,and the data was formatted into the input standard of the model and the timing prediction was carried out.The regression prediction evaluation index was used to compare and analyze the prediction results of the model and the real value,which proved that the travel time prediction model of bus section and the delay time prediction model of bus intersection constructed in this paper have a high degree of fitting,and the prediction error was within a reasonable range.Compared with the existing standard model,the accuracy of the model built in this paper was significantly improved under the evaluation indexes of mean square error,mean absolute error and mean absolute percentage error,which verified the validity of the model.