Risk Analysis And Model Calibration Based On Microscopic Driving State

With the development of economy,vehicles increase largely.Then,traffic safety issues have become increasingly serious,and unsafety traffic environment will not only bring the loss of property,but also a threat to human's lives.And compared with the remedy after the accident,it is very important to take measures before accidents to avoid the dangers.Firstly,based on the NGSIM data,we analyzed the state and the risk from macroscopic and microscopic perspective.Then according to different traffic state and risk indicators,the data was divided into three clusters by FCM clustering method.It was found that there were differences of risks among different clusters.And the Helly model was used to validate the data of different clusters.Finally,the risk indicators were introduced into the Helly model,and simulation results showed that the drivers could adjust their behaviors based on the real situation after introducing the risk indicators,and the new Helly model could make the drivers to avoid traffic risk effectively.Secondly,vehicle platoons were extracted form the NGSIM data.And we analyzed how the head vehicle influences the following ones,and the influence of the two subsequent vehicles.And there was obvious correlation between leader and its followers in terms of traffic states and risks.The influences of the leader could reach about fifth vehicle.So the platoons were divided into three clusters based on the leader's speed and relative speed,and the followers' speed and spacing.The results showed that different clusters have different risks.With the gradually increasing of the vehicles' index,the risk in cluster 1 changed little,the risk in cluster 2 gradually decreased,and the risk in cluster 3 gradually increased.Overall,the risk in cluster 2 was the highest.Finally,the measured data was used to analyze the influence of different vehicles'position on the risks.It was found that vehicles at different positions have different risks,and drivers would give attentions to adjust relative speed to reduce or avoid the risk.Finally,we analyzed the traffic states and risks by using a car-following experimental data at the suburb of Hefei,China.We found that due to the smaller speed and larger spacing,cluster 3 has low risks.Cluster 1 has both small relative speed and spacing,and cluter 2 has lager relative speed and spacing.Then,both cluster 1 and cluster 2 have certain risks.Based on these analyses,we compared the NGSIM data of United States and the experimental data at Hefei.From the aspects of both macroscopic and microscopic states,the distributions of speed and density form the experimental data at Hefei were much smaller,and the risks were smaller.The correlations of leaders speed and its followers were much largerer for the experimental data.However,the risk correlation between vehicles was much smaller.For different clusters,the distribution of spacing of NGSIM data has wider ranges,and the distribution of relative speed of experimental data has wider ranges.Thus corresponding NGSIM data has lower risk.