Private Car Travel Pattern Mining And Application Using Electronic Registration Identification Data Of Vehicles

With the development of modern society and economy,people’s living standards have improved and the ownership of private cars has increased.However,urban traffic congestion,environmental pollution,and other "urban diseases" have become increasingly serious,being the social focus.The "transportation power" strategy proposed by the state points out that intelligent transportation is the trend of modern transportation.Large-scale traffic data based on the ubiquitous perception of the Internet of things is the basis of intelligent transportation.The existing research on urban travel is mainly based on public bus data and taxi data,and rarely involves private car data.In fact,private car is the most important part of urban road traffic,and the research on private car travel has strong practical significance.Electronic Registration Identification of Vehicles(ERI)is an emerging traffic information collection technology,realizing the travel information collection of all urban motor vehicles,including private cars.However,different from the common traffic data,such as GPS data from taxi and online car-hailing and bus IC card data,ERI data has the characteristics of sparse spatial distribution and full vehicle coverage,not suitable for the traditional traffic data mining methods.This thesis focuses on " data quality → traffic regularity → traffic optimization ".To solve the time synchronization problem in ERI data,a relevant data repairing method is carried out.Then,the urban traffic regularity is studied from the perspective of vehicles and the road network.The private car travel regularity is reflected through the discovery of urban attractive areas,and the space-time regularity of road network traffic is reflected through the road network travel time estimation model.Finally,according to the discovered urban traffic regularity,the customized bus schemes for private car commuters are studied,providing a new idea for urban traffic optimization and helping solve urban diseases.The main innovations and contributions of this thesis are as follows:(1)Aiming at the time synchronization problems in ERI data,we propose a data repairing framework DR-TSP.DR-TSP mainly includes three parts: finding the problem ERI readers,inferring the clock drift rate and clock leap values,and repairing the problem ERI data.Firstly,statistical analysis for the travel time of all roads is employed to detect the problem ERI readers,and the method based on smoothing is utilized to find the clock leaps.Then,the neural network is used to construct the estimation model of the true travel time.The loss function considers the clock drift rate and clock leap values,which would be inferred in the process of model training.Finally,the inferred clock drift rate and clock leap values are employed to repair the problem data.Experiments show that the proposed method can effectively improve the quality of ERI data.(2)An urban private car travel pattern mining method based on ERI data is proposed.The method mainly includes two parts: trajectory segmentation and urban attractive area discovery.Firstly,we divide the private car trajectories into various trips by combining the link travel time distribution model with Bayesian classification,and the parameters of each link travel time distribution model are trained by the Expectation Maximum(EM)algorithm.Then,we propose a grid clustering algorithm based on the data field model,and utilize the algorithm to mine urban attractive areas according to private car trip data.Finally,the real ERI data of Chongqing are used for verification,and the private car travel pattern is visualized,revealing the travel regularity of urban private cars.(3)Construct an accurate network-level travel time estimation model and apply it to traffic simulation.Firstly,we construct the road network travel time estimation model MGCN,which integrates the traditional traffic models with the deep neural network models through the attention mechanism.MGCN fully considers the temporal and spatial characteristics of road network traffic,and uses MACD(Moving Average ConvergenceDivergence)and GCN(Graph Convolutional Network)to extract temporal and spatial characteristics respectively.Based on MGCN,a traffic simulation framework is proposed to realize the high-precision simulation of urban road network traffic under the given road network information and all vehicle trip information.Experiments show that the proposed method has superior performance in travel time estimation and traffic simulation,and the simulation accuracy for macro traffic is significantly better than the general traffic simulation tool SUMO.(4)Design customized bus schemes for urban private car commuters.We use private car trip data to study customized bus services,hoping to improve urban traffic by transforming private car commuters into customized bus commuters.The research is carried out from two aspects: customized bus demand and customized bus scheme.For customized bus demand,we mine private car commuters from private car trip data as potential service objects of customized buses,according to the temporal-spatial similarity and high-frequency characteristics of commuting trips.Then,we model the customized bus scheme with the number of service passengers as the goal.The model considers the spatial-temporal constraints of passengers,adopts buses with multiple capacities to make the customized bus scheme more flexible and efficient,and considers the minimum passenger load rate to ensure the benefit of each bus.The model is solved by the Differential Evolution algorithm.The real ERI data of Chongqing are used to evaluate the customized bus scheme,verifying the great potential of customized buses in alleviating traffic congestion and reducing urban traffic energy consumption.