Traffic Conflict Prediction And Safety Evaluation Application In Freeway Work Zone

The freeway traffic environment in the work zone of the freeway is complex and changeable,and has considerable potential safety hazards.Therefore,it is urgent to solve the traffic safety problems in the work zone of the freeway.In view of the short construction period of the work zone of the freeway and the lack of accident data,the traffic conflict technology that can "analyze in advance" is used to conduct a safety study on the work zone of the freeway.First,classify the driving status of vehicles in the work zone of the freeway,and analyze the operating characteristics of different construction sections,and then give the principle of traffic conflicts in the work zone.Then the field survey was carried out by video recording of unmanned aerial vehicles,and three types of work zones of amplitude modulation,closed overtaking lanes and compressed overtaking lanes were collected,and the basic traffic parameter data was extracted using the development platform based on Visual Studio.And analyze the traffic characteristics of the work zone from the aspects of traffic flow and vehicle type distribution,running speed,headway distance,lane changing characteristics and so on.Secondly,the rear-end conflict index TTC and the lane change conflict index PET based on the closeness of time and space are selected.The measurement methods of two types of conflicts are given separately,and the threshold of conflict risk degree is determined by combining statistical analysis and empirical values.Based on the spatial correlation of the two types of conflicts,regional risk factors are established to characterize the overall degree of conflict risk in the region.On this basis,explore the spatial distribution of conflicts,analyze the relationship between conflicts and traffic characteristics,and fully explore the relevant traffic influencing factors for building conflict prediction models.Thirdly,through simulation and SSAM to supplement the conflict conditions under different traffic conditions,combined with the analysis results of the relationship between conflict and traffic characteristics,according to the results of the multi-collinearity test to select the appropriate variables,the three types of work zones were established based on single-dimensional and multi-dimensional Bayesian Poisson log-normal distribution prediction model for rear-end conflicts and lane-change conflicts with different severity dimensions.Based on Bayesian inference,the appropriate prior distribution is set for the parameters.The M-H algorithm and Gibbs algorithm are used to realize the Markov chain Monte Carlo sampling.The likelihood function is used to obtain the parameter posterior probability function and the conflict prediction model is calibrated.Using DIC and criteria to test the superiority of the model,the results show that the goodness of fit of the multivariate model considering the correlation between conflicts in various dimensions is significantly higher than that of the single-dimensional model.By verifying the effectiveness of the model,it is found that the average relative error value of each prediction model is less than 10%,indicating that the constructed conflict model has good prediction ability.Finally,the four equivalent conflict rates based on different conflict types(rearend conflict,lane change conflict)under different risk dimensions are selected as safety evaluation indicators,and the overall safety evaluation method of highway work zone based on gray clustering theory is constructed and defined The safety level of each evaluation object is shown.Select the regional risk factors to evaluate the safety of each section of different work zone types,and use single-factor analysis of variance to make judgments on the significance and safety level of traffic safety differences in different sections.The results show that the overall evaluation level of the work zone is basically consistent with the on-site investigation status,and the safety level of various construction sections has a high consistency with the distribution of the conflicting spatial characteristics of the previous section,indicating that the gray clustering evaluation method and the regional risk factor The one-way ANOVA method is suitable for the safety study of highway work zone.