Research On The Skid Resistance Demand Of Ramp Based On Driving Information Captured By UAV

Ramps,generally composed of circular curve and transition curve,are indispensable parts of highway which connect main and auxiliary roads or guide driving routes,etc.Due to terrain constraints,radii of many ramps are designed small.When driving on small-radius ramps at a certain speed,drivers and passengers may have a certain sense of discomfort and risk;At the same time,the insufficient skid resistance of the road will lead to the increase of wheel slip rate,which is easy to make the vehicle drift or side-slip,and cause the failure to brake effectively.These conditions will not only aggravate the psychological burden of drivers,but also the main causes of accidents.Therefore,it is necessary to study the skid resistance demand of ramp.Different from the straight-line lane,the research on the skid resistance demand of ramp should consider both the braking factor and the lateral offset.Therefore,research should be carried out in combination of both the driving habits in curves,and skid resistance itself.How to accurately obtain drivers’ driving habits and what methods to evaluate the skid resistance demand of ramp are the main focus of this study.Starting with the evaluation of the skid resistance,this study firstly selects 7 indexes as the evaluation indexes from both vehicle dynamics and highway engineering(including 4 indexes of safety and 3 indexes of comfort),and determined the threshold of these 7 indexes;Then 6 typical vehicle models are selected and established in CarSim / TruckSim,and 32 groups of ramp models with different radius,design speed and slope are established according to the horizontal and vertical elements of ramp;After that,an ideal driver model is established according to braking perception response time and steering preview time,and three common driving conditions are set up.Based on the above design,the simulation test of ramp anti-skid demand is carried out,and the driving comfort and skid resistance demand are analyzed based on nearly 800 groups of data.Results show that: The skid resistance demand of small vehicles is much less than that of large vehicles;Due to the restriction of braking distance and lateral offset distance,the skid resistance demand of large vehicles does not strictly increase with the increase of vehicle weight;With the increase of basic speed,the skid resistance demand is largely increased with the same increment of speed,with the form similar to concave parabola;The skid resistance demand decreases with the increase of ramp gradient;When the design speed is 40km/h,the design slope of-4% and 5% is not suitable for driving,and the slope of-3%～4% is more appropriate.On the above basis,the data are fitted by various methods of regression,and finally the prediction formula of ramp is proposed.For the study of driving habits,this paper proposes to use UAV to collect and extract vehicle driving information.The collected track and speed information can effectively reflect the judgment of most drivers and extreme drivers in the face of specific road design,and can be collected in large quantities.To process the collected UAV video,Google Collaboration platform is used to modify and compile the YOLOv5 algorithm with Python,and YOLOv5 is retrain with the captured video.Results show that the precision rate P and recall rate R have satisfactory results,and the maximum F1 value is 0.86,reflecting a good P-R relationship;The loss function also stabilized at a very low level after 70 training epochs.The actual detection results show that the YOLOv5 recognition algorithm recombined in this study is very accurate and can be well applied to vehicle detection.Then,the trained YOLOv5 is used to replace the Faster R-CNN detector in the DeepSORT algorithm to improve the detection accuracy and speed,extract the vehicle driving information,and explain and calculate the coordinate transformation from the perspective of UAV.By coding,the coordinate information of vehicle trajectory is extracted,the trajectory is smoothed,and the frame difference method is used to calculate the real-time speed information,which is convenient for the establishment of real driver model;At the same time,the vehicle trajectory is drawn in the video,which is convenient for intuitive observation.Finally,taking the ramp of Cuipingshan Interchange as an example,the driving information of more than 150 vehicles in the UAV video is extracted and the trajectory is drawn,and three real driver trajectory models and four real driver speed models are established;The ramp elements are extracted by OpenStreetMap,and the ramp modeling is carried out in the form of GPS coordinate import in CarSim/TruckSim;Using the real driver trajectory model instead of the ideal driver model and the real driver speed model instead of the fixed speed model,the skid resistance demand and comfort of Cuipingshan Interchange ramp are studied;Using the skid resistance demand prediction formula proposed in this study,the theoretical anti slide demand of the ramp is calculated and compared with the actual simulation results.It shows that the simulation results are slightly larger than the calculated value of the anti sliding demand prediction formula.The reason is that the real driver model deviates from the center line of the road seriously,which is not conducive to the control of the lateral offset distance,so the skid resistance demand will naturally increase.From the overall results,the change trend of the prediction formula is the same as that of the simulation test established by the real driver model,and the values obtained are similar,which proves the accuracy of the prediction formula;From the perspective of identification accuracy,coordinate conversion accuracy and overall simulation accuracy,it also proves the feasibility of ramp anti sliding demand research based on UAV vehicle driving information acquisition.