合流区域多车交互风险实时评估方法

Real-time risk assessment method of multi-vehicle interaction at merging area

为探究合流区多车交互换道条件下的交通安全状况,开展了合流区多车交互风险实时评估研究。考虑车辆动力学以及多车交互作用特性,构建了基于车辆物理状态层、多车交互层和风险概率层的贝叶斯分层风险实时评估模型。利用MCMC吉布斯取样法标定模型参数,采用后验预测p值、参数分位数变化检验模型参数的有效性,通过仿真分析评价模型的实时评估性能,采用k-means聚类法对多车交互风险进行分级。结果表明:模型参数BGR统计值均小于0.1,Durbin-Watson后验p值、正态性假设p值、检验分布对称性p值和峰度p值都接近0.5,说明该风险评估模型收敛且具有良好的拟合性,各级风险的AUC结果表明模型对低风险、较高风险及高风险的识别较为准确,性能较好。本文研究可对一定时段内的车辆运行风险进行评估,为驾驶决策提供参考依据。

To explore the traffic safety characteristics of the multi-vehicle interactive lane-changing at the merging area,the study of real-time multi-vehicle interactive risk assessment was carried out.Considering the characteristics of vehicle dynamics and multi-vehicle interaction,a Bayesian hierarchical model of realtime risk assessment was constructed based on the physical state layer,multi-vehicle interaction layer and risk probability layer of vehicles.The model parameters were calibrated by MCMC Gibbs sampling method.The model validity was tested by posterior prediction of p value and variation of parameter quantile.And the predictive performance of the model was evaluated by simulation analysis.The method of k-means clustering was adopted to classify the risk level of multi-vehicle interaction.The results show that the statistical BGR of model parameters are all less than 0.1.And the p value of Durbin-Watson′s posterior,normal hypothesis,test distribution symmetry and kurtosis are all close to 0.5.So the risk assessment model is convergent and has good fitting.The AUC results show that the model is more accurate in identifying low risk,higher risk and highest risk,which has good performance.The research can evaluate the vehicle operation risk in a certain period of time and provide a reference for driving decision-making.