Analysis And Evaluation Of Traffic Conflict At Signal Intersection Based On Connected And Automated Vehicle

The connected and automated vehicle(CAV)technology can effectively improve road traffic safety through vehicle-to-infrastructure communication and autonomous driving,but at present,CAV is still in the development stage of low to high penetration,and their impact on traffic safety still needs further quantitative assessment.Therefore,this paper takes a signalized intersection in Yibin,Sichuan Province as the main research scenario,takes a mixed traffic simulation environment composed of human-driven vehicles and CAVs as the research object,deeply analyzes the correlation between dynamic traffic parameters and traffic conflicts at signal intersections.Based on Bayesian theory and Markov Chain Monte Carlo method,a comprehensive CAV traffic conflict model and a multi-dimensional CAV traffic left-turn conflict model are constructed for different traffic safety evaluation needs.The experiment is cross-validated through the SUMO traffic simulation platform.The specific research work of this paper is as follows:(1)To address the issue of factors affecting mixed traffic conflicts at signalized intersections,this study proposes a dynamic traffic parameter correlation analysis based on Spearman’s correlation coefficient.Based on SUMO traffic microsimulation platform,a typical cross-shaped signalized intersection scenario is built to simulate the random traffic flow state under CAV penetration rates and nine dynamic traffic parameters such as straight traffic volume and left-turn traffic volume are counted and calculated in terms of signal cycles,combined with TTC and PET conflict indicators to estimate the cumulative number of traffic conflicts and analyze the change trend of traffic conflicts under the influence of different parameters,while based on The correlation between different dynamic traffic parameters and traffic conflicts as well as traffic parameters is studied based on Spearman coefficient,and the dynamic traffic parameters with strong correlation are screened to provide theoretical support for the establishment of traffic conflict model.(2)To address the demand for a comprehensive evaluation of traffic safety at signalized intersections in CAV environments,this study proposes a comprehensive conflict model of signalized intersections in CAV environments based on multivariate combination.Based on the Poisson log-normal distribution model,six dynamic traffic parameters such as straight traffic volume,left-turn traffic volume,CAV penetration rate and conflict number are defined as model inputs to establish the CAV traffic conflict model with different combinations of explanatory variables,and the model parameter coefficients are calibrated by Markov chain Monte Carlo method.By comparing the traffic conflict models under the six parameter combinations,the experiments verify that the multivariate combined CAV traffic model has better estimation effect on the comprehensive quantitative traffic conflict analysis compared with the univariate model,using the model goodness-of-fit and Bayesian p-value as evaluation indexes.Meanwhile,the traffic conflicts under the dynamic threshold range of the model combination variables are quantified and analyzed.(3)To address the impact of through-left turning ratio on traffic conflicts at signalized intersections in CAV environments,a straight-left conflict model for multiple scenarios of signalized intersections in CAV environments is proposed.multi-dimensional through-left turning conflict model for smart connected traffic is proposed.Based on the multivariate Poisson log-normal distribution model,the multivariate straight-left conflict model of signalized intersections in CAV environments is constructed by considering the parameters of CAV penetration rate and straight-left ratio.Based on the road network structure and road section observation data of the signal intersection of Rongzhou Road,Yibin City,12 traffic scenarios with low,medium and high CAV penetration rates and different combinations of straight-left ratios are designed,and the model parameter coefficients are calibrated by Markov chain Monte Carlo method and compared with the traffic conflict model under a single traffic scenario.The results show that the multivariate direct left conflict model fits better,and the traffic conflicts under different traffic scenarios are quantified and analyzed by this model.Finally,a cross-validation was conducted at the signal intersection of on Shangmao Road to analyze the prediction accuracy and adaptability of the conflict model.In summary: 1)The calibration results of the multivariate combined conflict model show that if other influencing factors remain unchanged,a 1% increase in straight traffic flow is expected to increase traffic conflict by 0.30%;a 1% increase in left-turn traffic flow is expected to increase traffic conflict by 0.37%;and a 10 m increase in maximum queue length increases traffic conflict by 26.2%.2)The impact of straight traffic flow and left-turn traffic flow on straight-left conflict in different traffic scenarios has different results,and at the left-turn ratio of 0.2,the impact of smart network penetration on traffic conflict is the largest,indicating that increasing network penetration at this time can minimize traffic conflict,and traffic conflict is expected to decrease by 0.37% for every 1% increase in CAV penetration rates.