| With the booming growth of vehicle car,the expressway traffic congestion is becoming increasingly common.Under this condition,the accurate detection and prediction of expressway traffic congestion which are important parts of Intelligence Transportation System(ITS)can provide valuable information for users to make appropriate travel plan.Although with the continuous improvement of the construction of the Electric Toll Collection(ETC)gantry,realtime travel time and vehicle arrival time of the expressway network could be obtained,it is still impossible to extract the parameters which could directly determine traffic conditions,such as density and instantaneous speed.Also,the travel time could be easily affected by various factors such as service areas,which could affect the real average travel time of sections on expressways when evaluating traffic state.Therefore,the accuracy and robustness of traffic congestion detection and prediction models based on ETC transaction data could be incorrect.In addition,the noise in the raw data caused by the process of data collection and data transformation is too serious to ignore.This study proposes a framework to preprocess the ETC transaction data to ensure the reliability of data.Then this study explores the characteristic of expressway traffic congestion under ETC transaction data to detect and predict expressway traffic congestion.The main research work and contributions are as follows:(1)This paper analyzed the characteristics of expressway traffic congestion flow from the perspective of ETC transaction data.By analyzing the ETC transaction data,this paper proposed a transaction data preprocessing framework to clean abnormal data.Then,this paper explored the flow-change and average speed-change during the process of traffic congestion.The results show that expressway traffic congestion is similar to urban traffic congestion,but the difference is the time the traffic congestion occurs.In addition,the evolution of traffic congestion obeys the process,which is free flow to synchronous flow,synchronous flow to congestion flow,congestion flow to synchronous flow and synchronous flow to free flow.(2)This paper proposed an expressway traffic congestion detection model under the impact of expressway service areas.This paper explored the impact of expressway service areas on expressway traffic congestion identification by comparing travel time differences between users who get into service areas and users who don’t.Based on this,this paper constructs a traffic congestion detection model based on Speed Performance Index(SPI).The model incorporated three modules: 1)pause rate prediction;2)linear regression of pause rate and the bias caused by the service area;3)correction.The pause rate is used to quantify service areas.The result shows that the detection model could detect the traffic congestion under the impacts caused by service areas accurately.(3)This paper proposed an expressway traffic congestion prediction model based on traffic congestion maps.To solve the accuracy degradation caused by sudden changes in traffic phenomena in data driven models,this paper proposed a traffic congestion prediction model based on traffic pattern recognition.In addition,to reduce the time consumption when predicting traffic congestion,this study defines the concept of consensus days.Firstly,the traffic congestion map was established based on historical data,then clustering algorithms were employed to divide traffic congestion maps into several traffic modes.Secondly,the consensus days were extracted for each traffic mode by similarity calculation method based on Rand Index and create a set of consensus days.Finally,to predict traffic congestion,the model compared the real data with consensus day set to find the most similar consensus day to predict traffic congestion in next time.The result shows that the proposed traffic congestion prediction model based on traffic congestion map could predict the traffic congestion well. |
