Active Safety Management Based On Driving Behavior And Risk Analysis

Traffic safety of trucks is an important safety problem faced by many countries.In 2019,the number of trucks in China accounted for about 10.96% of the total number of motor vehicles,but the number of deaths caused by truck crashes accounted for 29.75%of the total number of deaths.Among them,the consequences of crashes involving heavy trucks are more serious,and the average death rate is about 2.38 times that of small buses.Trucks have the characteristics of large volume,many blind areas and high crash fatality rate,which brings a severe test to road traffic safety.About 87% of the federal transportation administration’s decision-making is related to the operation of large vehicles and trucks.As one of the key factors affecting the occurrence of traffic crashes,the behavior characteristics of drivers have attracted extensive attention.Identifying the risk factors of daily driving behavior and travel characteristics of trucks and drivers and establishing the relationship between dangerous driving behavior and crash risk are the keys to improving the active safety management level of the freight industry.According to the provisions of the “Measures for the Dynamic Supervision and Administration of Road Transport Vehicles”(2022 Amendment)issued by the Ministry of Transport of the People’s Republic of China,it is required that all heavy trucks,dangerous goods transport vehicles and semi-trailer tractors entering the transport market should be installed with satellite positioning devices and connected to the " National Public Supervision and Service Platform for Road Freight Vehicles".At present,the number of vehicles connected to the network in the platform has exceeded 7 million.At the same time,the transportation management departments in Jiangsu,Zhejiang,Guangdong provinces and other places have also vigorously promoted freight enterprises to install on-board active safety early warning systems in vehicles to collect data on driver behavior(such as fatigue and mobile phone use)and vehicle operation status(such as following too close and lane departure),which makes it possible to improve the safety of trucks and drivers.Through the analysis of the active safety warning data of more than 30000 trucks in China,it is found that "following too close" and "yawning" warnings are the two types of dangerous driving behaviors with the highest frequency,accounting for 35.6%and 17.5% respectively,reflecting the high risk of rear-end collision and fatigue driving of trucks during travel.However,the existing early warning for "following too close" only judges whether the time gap between two vehicles meets the early warning standard,and does not systematically improve the hazardous car-following habits accumulated by drivers.The early warning logic of some active safety monitoring equipment for driver fatigue usually only depends on whether the driver yawns to judge whether the driver is tired or not.The degree of fatigue is not graded,and the cumulative effect of fatigue over time is not considered,resulting in inaccurate fatigue early warning.In addition,most freight enterprises usually only use on-board monitoring data to query the vehicle location,driving route,early warning frequency and other contents,lack of systematic research on the key factors affecting the occurrence of traffic crashes,and insufficient safety evaluation of truck travel.However,due to the limitations of low acquisition frequency and non-continuous storage of all driving videos during vehicle travel,it is difficult to extract various micro-driving behavior indicators during data analysis.It is necessary to consider expanding research means(such as natural driving research and driving simulator Research)for analysis.In view of this,the research goal of this paper is from the perspective of crash prevention,at the driver’s behavior level,aiming at the key problems(rear-end collision risk,fatigue driving)found in the data analysis of active safety monitoring of trucks,to study the relationship of daily car-following behavior with rear-end collision,and put forward the driver fatigue level detection method.At the freight vehicle level,based on the on-board monitoring data of trucks,quantify the key characteristics affecting the crash risk of trucks,and design a comprehensive evaluation method for the driving safety of trucks.Finally,based on the research results,suggestions for the implementation of active safety management countermeasures are put forward.The main research contents and key results of this dissertation are as follows:First,the dangerous car-following behaviors related to the risk of rear-end collision were revealed.Using the high-precision and continuously recorded natural driving study data in Shanghai,this study extracts and analyzes the car-following events and rear-end collision dangerous events of drivers on urban surface roads and urban expressways.By establishing the Poisson hurdle model and Poisson regression model,the relationship between car-following behavior and the number of rear-end crashes under the two road types was quantified respectively.The results show that vehicle longitudinal control characteristics(such as relative speed standard deviation,maximum acceleration and deceleration),time control characteristics(such as the proportion of time gap less than 1s,minimum time gap),and emergency behavior(large jerks)are the key following behaviors that affect the risk of rear-end collision.This study makes an empirical study on the relationship between daily driving behavior and rear-end collision risk,which provides a guiding basis for truck driver education and driving assistance system optimization.Second,a drowsiness level discrimination model considering driver’s individual differences and the cumulative effect of drowsiness is proposed.This study uses the driving simulator experiment to carry out continuous data acquisition and analysis of drivers’ drowsy driving.By establishing a mixed-effect ordered logit model considering time cumulative effects and driver individual differences,i.e.,the MOLTCE model,it is found that the percentage of eyelid closure(PERCLOS)and the standard deviation of lane position are two effective indicators to judge the drowsiness level.Compared to the model’s accuracy without considering the time cumulative effect(67.75%),the MOL-TCE model has significantly improved the accuracy of drowsiness detection(80.35%).This study identifies the behavior characteristics significantly related to drowsy driving,optimizes the drowsiness level discrimination model,and provides support for the improvement of the early warning scheme of active safety monitoring equipment in feature selection and early warning logic design.Third,using the on-board monitoring data of trucks,the key characteristics affecting crash risk of trucks are explored and quantified.By extracting the travel trajectory information and the early warning information of active safety monitoring equipment from the active safety cloud platform,the zero-inflated Poisson regression model was induced in this study to establish the relationship between the active safety early warning characteristics,travel behavior,driving behavior and crash risk of trucks.The results show that the introduction of active safety early warning features effectively improves the model fitting results on the data.Yawning,smoking and the percentage of trips driven at night are the three most significant features affecting the crash risk of trucks.This study quantifies the relationship between the behavior characteristics of trucks and crash risk,provides theoretical support for optimizing the transportation task arrangement of freight companies and improves the company safety management strategy.Fourth,a comprehensive evaluation method of truck driving safety is proposed to support the active safety management of freight companies.Based on the active safety early warning information,travel behavior and driving behavior information extracted by the active safety monitoring equipment,this study constructs a comprehensive evaluation index system of truck driving safety based on considering the drivers’ dangerous car-following habits and drowsy driving.Based on the entropy weight method and the hierarchical analysis(i.e.,EWM-AHP)method,a comprehensive evaluation method with a combination of subjective and objective is constructed.The EWM-AHP is used to calculate the weight of truck safety evaluation indexes.Combined with k-Means cluster analysis and decision tree analysis,the detailed options of each index are obtained.By assigning different indexes and options,a comprehensive evaluation model of truck driving safety is established.The results show that: 1)the EWM-AHP method better balances the defects of subjective and objective evaluation methods in the evaluation process,making the evaluation system easy to interpret and use.2)The score of safety evaluation is significantly related to the number of crashes.The score shows a significant difference between trucks with and without traffic crashes,indicating that this method can effectively divide the safety of trucks.The research results provide basic theoretical and technical support for the function design of active safety monitoring equipment,driver’s hazardous driving behavior intervention education,transportation task arrangement,and vehicle and driver safety management.The suggestions are as follows: 1)Add a dangerous behavior feedback module to the forward collision warning system to provide drivers feedback on dangerous car-following behavior.2)In the driver drowsiness monitor system,the driving duration variables should be collected to characterize the cumulative effect of drowsiness,and the physiological and behavioral characteristics of drivers should also be extracted to realize more accurate drowsy driving warnings based on considering drivers’ individual differences.3)For car-following behavior and fatigue driving,strengthen the driver’s horizontal and vertical stability practice during driving,and learn to identify the phenomenon of "microsleep".4)Freight companies shall regularly carry out the identification of risk factors of truck crashes and driving safety assessment,timely identify key risk problems and high-risk trucks,and optimize the arrangement of transportation tasks.