Research On Key Technologies Of Hybrid Traffic State Analysis And Prediction For Urban Road Network

In order to solve the increasingly intense contradiction of the imbalance between traffic supply and demand in the urban road network,all countries worldwide have vigorously promoted the research on a new generation of intelligent transportation systems about vehicleroad communication and collaborative control.Intelligent connected vehicles(ICVs)enhance the real-time information interaction with surroundings to achieve efficient collaboration between people,vehicles,roads and the environment.Internet companies and automobile enterprises,such as Google and Tesla,have started to conduct ICVs related research,which currently has been on the stage of road test.Therefore,it is the general trend of urban traffic development that ICVs and traditional vehicles gradually constitute ”hybrid traffic” in the urban road network.With a view to ICVs,taking the route planning,driving advice and traffic prediction as starting points,this dissertation have carried out the Research on Key Technologies of Hybrid Traffic State Analysis and Prediction for Urban Road Network,to provide theoretical foundation and technical support for efficient safe driving of ICVs in the urban hybrid traffic.The main content and innovations of the dissertation are as follows:1.Aiming at the demand of information sharing among vehicles,roads and the environment in the urban hybrid traffic,a real-time traffic spatial-temporal geographic information database is constructed.It is composed of static empirical field,dynamic real-time field,and vehicle ego field.Static empirical data such as road network topology,road attributes and historical traffic information,dynamic real-time data such as vehicle attitude and traffic states,and vehicle ego data such as vehicle kinematic performance and driving strategies are stored as layers.In order to obtain the accurate information about vehicle attitude,a down-sampling strategy for GNSS signal acquisition is proposed,which has greatly improved the ability of real-time positioning while ensuring the accuracy of signal acquisition.2.Aiming at the characteristics of strong nonlinearity and numerous factors of the travel time prediction problem,a spatial-temporal inspired particle filtering algorithm for multi-step travel time prediction is proposed.By analyzing the spatial-temporal characteristics of travel time series using the harmonic analysis method and Moran's I global index,this dissertation constructs a third-order-spatial-adjacency extended traffic spatial-temporal matrix to describe historical traffic patterns,further adaptively select effective particles based on the prediction error and its confidence.Then,the correlation analysis between state prediction results of particles and the travel time observation sequence is conducted to establish the particle weight update model.It is also proposed that a resampling strategy based on the spatial-temporal similarity,to resample the low-weight particles based on the cumulative distribution of effective particle weights.In this way,the travel time multi-step prediction of urban roads is realized.The proposed algorithm has effectively solved the problem of particle degradation,as well as improved the real-time performance of travel time prediction and the stability of multi-step prospective prediction.3.Aiming at the problem that the vehicle driving independence is gradually weakening due to real-time traffic influence,an optimization algorithm is proposed for vehicle driving speed constrained by road conditions and traffic states.Based on the features of urban road design and real-time traffic states,this dissertation quantifys the influence of urban road alignment parameters,traffic compositions and incidents on vehicle driving speed,and deduces the multi-linear fitting equation of driving speed constrained by road conditions.Relying on the evolution laws of traffic states in the urban roads,the variable-length model of traffic flows is derived as the driving speed model constrained traffic states.Considering the transformation rules of traffic state parameters in the measured traffic data,driving speed equations constrained traffic fundamental diagram are established.The instantaneous travel time,total travel time and total travel distance are selected to form the performance evaluation matrix,and finally the optimal driving speed with the lowest driving cost is obtained by combining and solving these constraint equations.Simulation experiments with measured data of urban road alignment and traffic states have proved that,the proposed algorithm can find out the optimal driving speed considering urban road conditions and real-time traffic states,so as to provide speed suggestions for ICVs driving in the urban hybrid traffic.4.In order to investigate the formation of traffic shockwave and its effects in the urban road network,a numerical detection algorithm of traffic shockwave propagation trajectories is proposed.According to the traffic dynamic model and vehicle conservation law,the spatial-temporal map of traffic state evolution is deduced.For the first time,gradient detection,non-maxima suppression,and dual-threshold methods for image processing are used to detect the shockwave propagation trajectories in the spatial-temporal map of traffic states.In order to verify the consistency between the proposed numerical detection algorithm and traffic analytical equations,a polar-parameter coordinate projection method is adopted,and an error correction model also is established to decouple the angle and distance errors of line segments of the extracted trajectories.Simulation experiments of no-controlled and variable speed limit controlled traffic bottleneck scenarios have proved that,the proposed algorithm can describe the spatial-temporal evolution process of traffic states,such as congestion and queuing in the bottleneck scenarios,and accurately extract traffic shockwave propagation trajectories with the relative distance error reduced to 0.1%-3%.In order to realize the information sharing and cooperative control in the hybrid traffic of urban road network,this dissertation has carried out the research on key technologies,such as the construction of real-time traffic database,the prediction of road travel time,the constraint optimization of vehicle driving speed,and the analysis of traffic state evolution.The measured data and simulation experiments have proved the effectiveness of the proposed algorithms.