Identification And Modeling Of Vehicle Travel Characteristics In Affected Areas Of Urban Arterial Road Driven By Trajectory Data

With the continuous increase in the number of vehicles,the phenomenon of urban road traffic congestion has continued to worsen.New building road projects can no longer meet the growing demand for travel.Establishing the travel characteristics of vehicle groups in detail and providing travel demand files is significant.To provide a theoretical basis for customized intelligent traffic management and scientific traffic planning is an important issue for traffic managers.Accurately grasping the rules of vehicle travel requires detailed analysis of vehicle trajectories,sublimation of trajectory information,and refinement of effective travel indicators.The number of vehicles is large and the vehicle travel are complicated.The object of traffic management can not only target a single vehicle,or nor can all vehicles be treated the same.Therefore,according to the needs of urban traffic management,scientific,effective and customized classification of vehicles is the basic work of intelligent urban traffic management and planning.This article mainly establishes travel characteristic index system such as travel frequency,online time,trajectory repetition rate,travel time,activity preference area,main line influence area preference,etc.by mining historical RFID trajectory data.The travel time-space distribution of different types of vehicles is analyzed.Based on the Gaussian HMM model,the travel of private cars,taxis,and trucks is characterized.Finally,according to the analysis of the travel characteristics of private cars,taxis,and trucks,selection of travel characteristic indicators,and then based on the combination of the density peak clustering algorithm(CFSFDP)& the BP neural network model,identify different travel modes,that is,to achieve Identification of travel characteristic groups.The main research contents of this article are as follows:(1)Analysing the RFID data,explain the RFID detection principle and data advantages,and propose detailed steps and procedures for raw data pre-processing for the redundancy and error of RFID data.Finally,the technical process of RFID data association and matching is given to get effective Complete visualization data.(2)Study on the extraction of travel characteristic indicators based on RFID trajectory data.Based on the comparison of RFID data and Baidu map itinerary planning data,a study on determining the dwell time threshold is made to break the vehicle travel chain and lay the foundation for index extraction.Based on the analysis of the distribution characteristics of vehicle operation,combined with the characteristics of RFID data,a characteristic index system such as travel frequency,online time,trajectory repetition rate,travel time,activity preference area,and trunk influence area preferences has been established.(3)It’s a methods are propoers based on the vehicle travel characteristics of private cars,trucks,and taxis.By analyzing the statistical distribution characteristics of private cars,trucks,and taxis on trip frequency,online time,trajectory repetition rate,first trip period,and end-of-travel feature indicators,the differences in travel characteristics of different types of vehicles are studied.The Gaussian Hidden Markov Model is used to model the travel characteristics of private cars,trucks,and taxis.(4)Based on the travel characteristic index system,the secondary travel group division of private cars,trucks,and taxis is studied.A travel feature group identification model based on the improved density peak(CFSFDP)algorithm & BP neural network algorithm was established.Based on the density peaking algorithm,aiming at the shortcomings of the algorithm,the reasonable value of the parameters(cutoff distance)and the reasonable selection of the local density calculation method,the idea of Kneighbor value is introduced to improve the original algorithm,which makes the algorithm compatible and more Suitable for diverse big data environments.Based on the classification data after cluster analysis,BP neural network learning training is performed.(5)It be selected the RFID data of Chongqing for experimental analysis.First,based on the travel feature indicators extracted from private cars,taxis,and trucks,cluster analysis of the improved CFSFDP algorithm was performed to divide private cars into three characteristic groups,namely commercial private car groups,commuter private car trips,and confused private car trips.Trucks are divided into two characteristic groups,which are long-term truck trips and short-term truck types.taxis are also divided into two types: other regions prefer taxi groups,and trunk-affected zones prefer taxi groups.Then based on the classification data for BP neural network training and learning,the recognition accuracy rate of the private car trip feature recognition model is as high as97.2%.The recognition rate of the truck trip feature recognition model is as high as95.64%.The taxi trip feature recognition model is correctly identified.The rate is as high as 99.18%.