| Pedestrians are a representative vulnerable group in the traffic system.Once an accident occurs,the severity of injury to pedestrians is significantly higher than that of other traffic participants.In recent years,data on road traffic accidents in China show that each pedestrian-related traffic accident will cause an average of 1.2 pedestrians to be injured or died,and the probability of death among these 1.2 pedestrians is as high as 29.2%.In the statistical data of the past five years,the proportion of pedestrian fatalities in the total traffic accident fatalities has been maintained at more than 25%.Zebra crossing is set up to solve the characteristics of the vulnerable group of pedestrians.However,the actual situation shows that the proportion of Chinese drivers who obey the zebra crossing"pedestrian first"rule is extremely low,and a large number of drivers have never even taken the initiative to perform a"yielding"action.In order to improve the safety of pedestrian street-crossing,the feasible methods include two aspects:The first is to promote the driver to take the initiative to yield through various means,implement the behavior of"yield to pedestrian",and reduce the frequency of pedestrian-vehicle conflicts.The second is to use the on-board early warning system to recognize the pedestrian-vehicle interaction,alert the driver when there is a risk,or directly perform autonomous braking.In response to the above two requirements,the lidar and video recorders were used to conduct continuous natural observations on a zebra crossing without signal lights for more than 3 years,and obtains pedestrian-vehicle interaction data in different years.The data collected for more than 200 hours,the total number of vehicles passing through the zebra crossing exceeded 100,000,and the number of pedestrians exceeded 20,000.During the continuous experimentation of this article,it covers three different stages of the enforcement of the"yield to pedestrian"policy by the traffic management department of Xi'an:no policy enforcement,strong policy enforcement,and weak policy enforcement.The study of the"yield to pedestrian"policy is equivalent to a huge social experiment in Xi'an with the help of the policy enforcement ways of the traffic management department.Meanwhile,during the data collection period,there was a missing zebra crossing due to the re-paving of the test site.Therefore,this paper also obtained the pedestrian-vehicle interaction data of the same site with or without zebra crossings.Relying on the above-mentioned naturally collected data,the main research contents and conclusions of this paper are as follows:(1)The characteristics of pedestrian street crossing with different policy enforcement are analyzed and studied.When the enforcement of the“yield to pedestrian”policy changed from none to strong,the yield rate increased rapidly from 3.6%to 68.6%.Once the policy enforcement was weakened,the yield rate quickly dropped to 34.1%.There is a strong correlation between the yield rate and the policy enforcement.With the strengthening of policy enforcement,pedestrian waiting time has been reduced by about 41.58%;Compared with the absence of policy enforcement,when the policy enforcement is strong,the critical acceptance for decision-making for street-crossing is the smallest,and pedestrians'risk-taking capacity and confidence in street-crossing increase by 16%from the data.The above results show that the“yield to pedestrian”policy enhances pedestrians'courage to cross the street significantly.(2)Under different levels of enforcement efforts of the policy,the characteristics of vehicles when passing by the zebra crossing were studied and analyzed.As the level of enforcement efforts increases,the mean speed of the vehicles on the zebra crossing zone declined by 17.6%compared with that before the enforcement of the policy.While when the enforcement efforts of the policy degrade,vehicles'mean speed when passing by the zebra crossing part starts to get larger again correspondingly,with rising amplitude of 10.2%.Of all the data collected,in the mode when the vehicles actively giving way to pedestrians,they traveled with a deceleration of 1.94m/s~2,while in the mode of vehicles passively yielding to pedestrians,the vehicles passed by with a deceleration of 2.20 m/s~2.So,even though the interaction between vehicles and pedestrians is complicated under different modes,we can also found that the mean value of deceleration in active yielding mode is smaller by 11.8%to that of passive yielding mode.From this point of view,it is verified that it's safer for the pedestrians to crossing the road in the active yielding mode,on the other hand,it proves that“yield to pedestrians”policy has positive effect on the safety of pedestrians crossing the road.(3)Pedestrians'crossing intention recognition model with/without zebra crossing line before the crossing behavior has begun was built considering the requirements of Pedestrian Warning Systems.Compare the results of the analysis of the data,we can find that,when there is no zebra crossing on the road and the pedestrian intend to cross the road immediately and directly,the mean distance between the vehicle and the zebra crossing zone is 42.1m,which is larger than that of the occasion 38.46m when there is zebra crossing on the road and the pedestrian intend to cross the road without waiting.When there is no zebra crossing on the road and the pedestrian choose to wait,the mean distance between the vehicle and the zebra crossing zone is 22.0m,which is larger than the mean distance 19.59m when there is zebra crossing on the road and the pedestrian is inclined to wait.When there is no zebra crossing on the road and the pedestrian is going to cross the road directly,the mean TTC is3.83s,while the mean TTC is 3.46s when there is zebra crossing and the pedestrian has the same intention.When there is no zebra crossing on the road and the pedestrian is going to wait,the mean TTC is 2.21s,which is larger than that of the situation 1.96s when there is zebra crossing on the road and the pedestrian intend to wait.When there is no zebra crossing and the pedestrian intend to cross the road directly,the mean speed of the pedestrian is4.61km/h,while the mean speed of the pedestrian is 4.15km/h when there is zebra crossing and the pedestrian has the same intention.When there is no zebra line and the pedestrian is going to wait,the mean speed of the pedestrian is 2.16km/h,which is smaller than that of the occasion 2.48km/h when there is zebra crossing and the pedestrian has the same intention.Based on attention mechanism&bidirectional long short-term memory(Atten-Bi LSTM),pedestrians'crossing intention model was built in this paper,the recognition accuracy when there is no zebra crossing is 93.89%,and 93.05%when there is zebra crossing on the road.(4)In order to improve the performance of Pedestrian Warning System,a model of the trajectory of the pedestrian when passing the road was developed when there is/isn't zebra crossing line on the road.In the situation with/without zebra crossing line,the mean length of waiting time for pedestrians tending to cross the road are 7.23s,8.89s separately,the time used for crossing the road are 3.71s,4.22s separately,the mean speed pedestrians crossing the road are 4.47km/h and 5.47km/h,the mean speed of vehicles traveling through the zone are35.42km/h and 29.17km/h.Affected by the policy,pedestrians are inclined to cross the road more boldly when there is zebra crossing line on the road,while when there is no line,they are more willing to take a crossing action in a situation with lower risks.Tested by real data,the model combined Hidden Markov Model(HMM)and Bi-LSTM can predict the trajectory of pedestrians when there is zebra crossing or no zebra crossing on the road. |
PDF Full Text Request