Complex interleaving area safety risk assessment method in intelligent network connection mixed driving environment

By building a traffic simulation model and building a traffic conflict analysis model, combined with gray cluster evaluation, the problem of evaluating the impact of intelligent connected vehicles on traffic safety in complex intertwined areas is solved, and quantitative analysis of traffic flow and safety risks is realized, and traffic is improved. Accuracy of safety evaluation.

CN118968747APending Publication Date: 2024-11-15CHONGQING JIAOTONG UNIV
View PDF 0 Cites 4 Cited by

Patent Information

Application Number
CN202410935959.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

After the existing technology has been integrated into the transportation system, it has failed to effectively evaluate its impact on traffic safety in complex intertwined areas. Especially in the mixed environment of urban complex intertwined areas, traditional methods are difficult to quantify traffic safety risks, and it has not Fully consider the impact of traffic composition and the penetration rate of autonomous vehicles on safety.

Method used

The safety risk assessment method of complex intertwined areas in the intelligent network hybrid driving environment is adopted, including building a traffic simulation model, building a traffic conflict analysis model and verifying it, establishing traffic safety evaluation indicators and methods, through field research and graphical road network editing tools, Simulation experiment design and data analysis, determine the parameter calibration of artificial driving and autonomous driving vehicles, build a negative binomial distribution model, and use gray clustering for safety evaluation.

Benefits of technology

It has achieved stability assessment of traffic flow in complex intertwined areas under intelligent networked environments and reduced traffic safety risks, provided quantitative analysis tools, able to more accurately evaluate traffic safety levels under different autonomous driving penetration rates, and improved traffic safety Accuracy and credibility of evaluation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118968747A_ABST
    Figure CN118968747A_ABST
Patent Text Reader

Abstract

The invention discloses a complex interleaving area safety risk assessment method in an intelligent network connection mixed driving environment, and the method comprises the following steps: S1, building a traffic simulation model, including scene building and vehicle parameter calibration; s2, constructing and verifying a traffic conflict analysis model; and S3, establishing a traffic safety evaluation index and method, and constructing a safety grade division method. The method has the advantages that the prominent problem of restricting road traffic safety can be solved, the traffic flow of the complex interlaced area in the intelligent network connection environment is stabilized, and the traffic safety risk is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of traffic safety risk assessment, and in particular to a method for assessing safety risks in complex interweaving areas under an intelligent networked mixed driving environment. Background Art

[0002] In the study of mixed traffic flow, the current research on mixed traffic flow at home and abroad mainly focuses on the following model of mixed traffic flow vehicle driving model and its safety and stability. Most of the research on safety is aimed at urban intersections and basic sections of highways. In recent years, it has begun to involve traffic safety in merging and diverging areas of expressways. Research on traffic safety in interweaving areas is relatively scarce and urgently needs to be supplemented and improved.

[0003] In the research of traffic conflict technology, the indicators for measuring traffic conflicts can be summarized into three categories: time-based indicators, kinematics-based indicators and collision energy indicators. Among them, time-based indicators are widely used in the construction of conflict models due to their intuitiveness and ease of use. Commonly used model building methods mainly include time series analysis, regression analysis, generalized linear models (such as negative binomial distribution, Poisson distribution and zero stacking distribution) and neural networks.

[0004] In terms of weaving zone safety assessment methods, existing methods mostly focus on analyzing the impact of key variables such as weaving length, weaving ratio, traffic volume, weaving flow ratio and speed, and generally use traffic conflict technology as the main means of safety assessment. The impact of traffic composition is rarely considered, especially at a time when intelligent network technology and autonomous driving vehicles are gradually integrated into the transportation system. The research on weaving zone safety issues considering different autonomous driving penetration rates is relatively weak, and the research on traffic safety in mixed traffic environments in typical scenarios such as complex urban weaving zones has not yet been carried out.

[0005] At present, intelligent networked vehicles are still in the development stage of low penetration to high penetration, and their impact on road traffic safety needs further quantitative evaluation. There are significant differences between traditional manually driven vehicles and intelligent networked autonomous driving vehicles in terms of driving behavior, reaction speed and decision-making logic, and mixed traffic flow may cause new traffic safety problems. At present, some scholars have conducted traffic safety research after the mixing of autonomous driving in typical scenarios such as intersections, merging areas, and toll station diversion areas. However, since the traffic flow in the weaving area is more chaotic and complex, and the traffic characteristics analysis is more difficult, there is still a gap in the quantitative analysis of traffic safety of intelligent networked vehicles in the weaving area. Summary of the invention

[0006] In view of the above-mentioned deficiencies in the prior art, the technical problem to be solved by this patent application is how to provide a safety risk assessment method for complex interweaving areas in an intelligent connected mixed driving environment that can solve the prominent problems that restrict road traffic safety, stabilize the traffic flow in complex interweaving areas in an intelligent connected environment, and reduce traffic safety risks.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] A safety risk assessment method for complex interweaving areas in an intelligent connected mixed driving environment includes the following steps:

[0009] S1: Build a traffic simulation model, including scene construction and vehicle parameter calibration;

[0010] S2: Build and verify the traffic conflict analysis model;

[0011] S3: Establish traffic safety evaluation indicators and methods, and construct a safety level classification method.

[0012] As an optimization, in step S1, the road geometry parameters obtained from the field survey and the processed traffic flow data are used to accurately control the flow, speed, path, and vehicle category parameters of the weaving area by lane; the road network modeling is performed using the graphical road network editing tool netedit based on the geometric parameters of the weaving area; and the vehicle type, traffic path, vehicle input, expected speed distribution and other parameter settings and additional files of the routing file are created in combination with the actual survey traffic flow data.

[0013] As an optimization, according to the above steps, the calibration of parameters of the manually driven vehicle and the calibration of simulation parameters of the autonomous driving vehicle are determined;

[0014] In the manual driving vehicle parameter calibration, the IDM following model is selected as the following model, and the lane changing model adopts LC2013; the expression of the IDM following model is as follows

[0015]

[0016] In the formula, a s is the acceleration of the following vehicle; a m is the maximum acceleration of the vehicle; v 0 is the expected speed; v S is the speed of the following vehicle; δ is the vehicle acceleration index; s 0is the headway time under congestion; T is the safe headway time; Δv is the speed difference between adjacent following vehicles; Δx is the headway distance; b is the expected deceleration; the IDM model setting parameters are deceleration (m / s2) is 4.4; the maximum expected speed (m / s2) is 5.56; the minimum headway distance (m) is 2.5; the expected headway time (s) is 1; the speed factor is 1; the driver's driving proficiency is 0.5; the speed deviation is 0.1;

[0017] The CACC following model is used as the following model in the simulation parameter calibration of the autonomous driving vehicle, and the SL2013 lane changing model is used as the lane changing model. The expression of the CACC following model is as follows:

[0018] e s (t-1) = Δx(t-1) - t c v s (t-1)-ls 0

[0019]

[0020] Where: e s (t-1) is the headway error of the vehicle at time t-1; Δx(t-1) is the headway between the following vehicle and the preceding vehicle at time t-1; t c is the expected headway between vehicles; l is the vehicle length; s 0 is the headway distance in congestion; v s (t) is the speed of the vehicle at time t; v s (t-1) is the speed of the vehicle at time t-1; is the inter-vehicle distance error increment at time t-1; k p and k d are adjustment coefficients, which are 0.45 and 0.25 respectively.

[0021] As an optimization, sensitivity analysis is performed on the minimum vehicle spacing (minGap), expected minimum headway (tau), speed factor (Speed ​​Factor) in the IDM following model in the above steps, and strategic lane change (lcStrategic), cooperative lane change (lcCooperative), speed-up lane change (lcSpeedGain), minimum acceptance spacing for lane change (lcAssertive), and keeping right (lcKeepRight) in the SL2013 lane change model; the ratio of the difference between the maximum and minimum values ​​to the mean is used as the mathematical indicator for sensitivity analysis of the correction parameters; the following formula is used

[0022]

[0023] According to the simulation sensitivity analysis results, the correction index can be determined to calibrate the core parameters.

[0024] As an optimization, step S2 is to establish a traffic conflict simulation in a mixed traffic environment and analyze the influencing factors before constructing the traffic conflict analysis model; first, three assumptions are made: the mixing of the autonomous driving vehicle does not affect the driving operation of the manual driver; no traffic accidents occur during the simulation; the driving behavior parameters of the autonomous driving vehicle are set based on the existing research and the traffic flow characteristics of the complex interweaving area; then, the orthogonal experimental design method is used to construct a simulation experiment orthogonal table for the mixed traffic environment in the complex interweaving area, and uniform sampling is performed according to the pre-set influencing factors and their level values, so as to obtain comprehensive and representative sample data within a limited number of simulation experiments; secondly, the SPSS software is used to perform an orthogonal experimental design on the selected factors and level variables, and the combination of each factor in each experiment is input into SUMO as a simulation experiment input value to run the simulation model; finally, the SSAM software is used to perform conflict analysis, and the conflict data is exported. According to the conflict threshold, the rear-end collision conflict data of different severities and the lane change conflict data of different severities are respectively counted.

[0025] As an optimization, a conflict model was constructed based on the above steps. Stata software was used to analyze the data, and the AIC criterion (Akaike Information Criterion) and BIC criterion (Bayesian Information Criterion) were used as the judgment criteria. The Poisson or negative binomial distribution was compared by overdispersion and goodness of fit to determine the final model. The negative binomial distribution explained each conflict. The negative binomial distribution model was established on the basis of the Poisson distribution model. When the data was relatively discrete, the disturbance term ε was added to take into account the existence of errors. i Balance the relationship between expected value and variance. At this time, the expected value λ i ' can be expressed as:

[0026] λ i '=exp(β i x i +ε i )=λ i exp(ε i )

[0027] At this time, the probability of the number of random events occurring can be expressed as:

[0028]

[0029] The formula conditional distribution is converted into unconditional distribution to obtain the probability density of the number of traffic conflicts, that is, the negative binomial regression distribution model:

[0030]

[0031] Where: P(xi ) is the occurrence of x at position i i The probability of traffic conflict; α is the parameter to be calibrated.

[0032] Severe Lane Change Conflict Calculation Model

[0033] The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of serious lane-changing conflicts, and the alpha value of the model is 0.063, indicating that it has passed the overdispersion test; therefore, the model is constructed as follows:

[0034] N 1 =exp(1.9225128+0.0020886V-0.5406693MR+0.0131275L

[0035] -0.7760446lR+2.6839670VR)

[0036] Where: N 1 is the number of serious lane change conflicts, veh / h; V is the average hourly traffic volume of the lane, veh / h; MR is the penetration rate of autonomous driving, %; L is the weaving length, m; IR is the weaving ratio; VR is the weaving flow ratio.

[0037] General lane-changing conflict calculation model

[0038] The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of general lane change conflicts, and the alpha value of the model is 0.046, indicating that it has passed the overdispersion test. Therefore, the model is constructed as follows:

[0039] N 2 =exp(2.5629645+0.0009639V-0.5953913MR+0.0171947L

[0040] +2.0972626VR)

[0041] Where: N 2 is the number of general lane change conflicts, pcs / h; V is the average hourly traffic volume of the lane, veh / h; MR is the penetration rate of autonomous driving, %; L is the weaving length, m; VR is the weaving flow ratio;

[0042] Severe rear-end collision calculation model

[0043] The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of serious rear-end collisions, and the alpha value of the model is 0.132, indicating that it has passed the overdispersion test. Therefore, the model is constructed as follows:

[0044] N 3=exp(0.1211006+0.0035253V+1.6124795VR+0.0070063L)

[0045] Where: N 3 is the number of serious rear-end collisions, veh / h; V is the average hourly traffic volume of the lane, veh / h; VR is the weaving flow ratio; L is the weaving length, m;

[0046] General rear-end collision calculation model

[0047] The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of general rear-end collisions, and the alpha value of the model is 0.034, indicating that it has passed the overdispersion test. Therefore, the model is constructed as follows:

[0048] N 4 =exp(3.5436384+0.0015160V-0.4050758MR+0.0120833L

[0049] -0.6183874IR)

[0050] Where: N 4 is the number of general rear-end collisions, veh / h; V is the average hourly traffic volume of the lane, veh / h; MR is the penetration rate of autonomous driving, %; L is the weaving length, m; IR is the weaving ratio.

[0051] As an optimization, in step S2, the traffic conflict analysis model is verified by using SSAM software to extract conflict data from the SUMO simulation experiment results in a mixed traffic environment, and using it as a test reference group. Then, the constructed model is used to calculate the conflict data under different traffic conditions. By calculating the absolute error, relative error and mean absolute error between the two groups of data, the effectiveness of the four traffic conflict calculation models is verified; the mean absolute error MAE is standardized into a percentage form, namely MAE%; the calculation formula is as follows:

[0052]

[0053] Where: n is the number of samples; Calculate the value for the model; y i Extract values ​​for simulation results.

[0054] As an optimization, in step S3, the safety evaluation index is established, and the "conflict occurrence rate" is selected as the evaluation index of the safety status of the complex weaving area. The traffic conflict occurrence rate k i Refers to the number of traffic conflicts N per hour i With traffic volume Q i The specific calculation formula is as follows:

[0055]

[0056] Where: k i is the conflict occurrence rate; i represents the conflict type, i=1, 2, 3, 4, representing severe lane-changing conflict, general lane-changing conflict, severe rear-end collision conflict and general rear-end collision conflict, respectively.

[0057] As an optimization, the safety evaluation method in step S3 uses grey clustering to evaluate the traffic safety of complex interweaving areas under mixed traffic conditions, and constructs a sinusoidal grey clustering safety evaluation model based on combined weighting.

[0058] As an optimization, step S3 constructs the safety level division and compares the clustering coefficients corresponding to different gray classes. The gray class to which the maximum clustering coefficient belongs is the safety risk level of the experimental scheme. The sample values ​​of different evaluation indicators under different safety levels in complex interweaving areas are statistically analyzed. The maximum evaluation index value is used as the threshold for levels one, two, and three to determine the evaluation index thresholds for each safety level in the intelligent connected mixed traffic environment.

[0059] In summary, this safety risk assessment method for complex weaving areas in an intelligent connected mixed driving environment has the advantages of being able to solve prominent problems that restrict road traffic safety, stabilize traffic flow in complex weaving areas in an intelligent connected environment, and reduce traffic safety risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a flow chart of a method for safety risk assessment of complex interweaving areas in an intelligent connected hybrid driving environment described in the present invention.

[0061] Figure 2 This is the SUMO control flow chart.

[0062] Figure 3 Flowchart for calibrating driving behavior parameters.

[0063] Figure 4 This is a diagram showing the impact of various factors on traffic conflicts.

[0064] Figure 5 is the possibility function of each evaluation level. DETAILED DESCRIPTION

[0065] The present invention is further described in detail below in conjunction with the accompanying drawings. In the description of the present invention, it should be understood that the directions or positional relationships indicated by directional words such as "upper, lower" and "top, bottom" are usually based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description. Unless otherwise stated, these directional words do not indicate or imply that the devices or components referred to must have a specific direction or be constructed and operated in a specific direction, and therefore cannot be understood as limiting the scope of protection of the present invention; the directional words "inside and outside" refer to the inside and outside relative to the contours of each component itself.

[0066] like Figure 1-5 As shown, a safety risk assessment method for complex interweaving areas in an intelligent connected mixed driving environment includes the following steps: S1: building a traffic simulation model, including scene construction and vehicle parameter calibration; S2: building a traffic conflict analysis model and verifying it; S3: establishing traffic safety evaluation indicators and methods, and building a safety level classification method.

[0067] In this embodiment, in step S1, the road geometry parameters obtained from field surveys and the processed traffic flow data are used to accurately control the flow, speed, path, and vehicle category parameters of the weaving area by lane; the road network is modeled using the graphical road network editing tool netedit based on the geometry parameters of the weaving area; and the vehicle type, traffic path, vehicle input, expected speed distribution, and other parameter settings and additional files of the routing file are created in combination with the actual survey traffic flow data.

[0068] In this embodiment, according to the above steps, the manual driving vehicle parameter calibration and the automatic driving vehicle simulation parameter calibration are determined;

[0069] In the manual driving vehicle parameter calibration, the IDM following model is selected as the following model, and the lane changing model adopts LC2013; the expression of the IDM following model is as follows

[0070]

[0071] In the formula, a s is the acceleration of the following vehicle; a m is the maximum acceleration of the vehicle; v 0 is the expected speed; v S is the speed of the following vehicle; δ is the vehicle acceleration index; s 0 is the headway time under congestion; T is the safe headway time; Δv is the speed difference between adjacent following vehicles; Δx is the headway distance; b is the expected deceleration; the IDM model setting parameters are deceleration (m / s2) is 4.4; the maximum expected speed (m / s2) is 5.56; the minimum headway distance (m) is 2.5; the expected headway time (s) is 1; the speed factor is 1; the driver's driving proficiency is 0.5; the speed deviation is 0.1;

[0072] Specifically, the lane-changing model uses LC2013. Under the control of this model, the speed of the vehicle is mainly affected by the vehicle in the same lane or the target lane ahead. Referring to the selection of driving behavior parameters of vehicle models in mixed traffic flow research by domestic and foreign scholars, the following parameters are determined for sensitivity analysis.

[0073] Table 1 Key parameter selection

[0074]

[0075]

[0076] In the routing file, three types of manually driven vehicles are created for cars according to different vehicle routes and lane change requirements. After adjusting the selected driving behavior parameters, simulation experiments are conducted through SUMO-GUI. The calibration results of driving behavior parameters in different types of manually driven vehicles are shown in Table 2. Finally, the driving behavior parameters in different types of manually driven vehicles are calibrated and the measured values ​​of the evaluation indicators from the field traffic survey statistics are compared with the simulation values ​​of the final calibration results.

[0077] Table 2 Calibration results of driving behavior parameters of manually driven vehicles

[0078]

[0079] The CACC following model is used as the following model in the simulation parameter calibration of the autonomous driving vehicle, and the SL2013 lane changing model is used as the lane changing model. The expression of the CACC following model is as follows:

[0080] e s (t-1) = Δx(t-1) - t c v s (t-1)-ls 0

[0081]

[0082] Where: e s (t-1) is the headway error of the vehicle at time t-1; Δx(t-1) is the headway between the following vehicle and the preceding vehicle at time t-1; t c is the expected headway between vehicles; l is the vehicle length; s 0 is the headway distance in congestion; v s (t) is the speed of the vehicle at time t; v s (t-1) is the speed of the vehicle at time t-1; is the inter-vehicle distance error increment at time t-1; k p and k dare adjustment coefficients, which are 0.45 and 0.25 respectively.

[0083] In this embodiment, sensitivity analysis is performed on the minimum vehicle spacing (minGap), the expected minimum headway (tau), the speed factor (Speed ​​Factor) in the IDM following model in the above steps, and the strategic lane change (lcStrategic), cooperative lane change (lcCooperative), speed-up lane change (lcSpeedGain), the minimum acceptance spacing for lane change (lcAssertive), and keep right (lcKeepRight) in the SL2013 lane change model; and the ratio of the difference between the maximum and minimum values ​​to the mean is used as the mathematical indicator for the sensitivity analysis of the correction parameters; the following formula is used:

[0084]

[0085] According to the simulation sensitivity analysis results, the correction index can be determined to calibrate the core parameters.

[0086] Specifically, by calculating the sensitivity of the average speed and the number of lane changes, the results shown in Table 2 are obtained, among which the Speed ​​Factor, lcStrategic, lcCooperative, lcSpeedGain, lcAssertive, and lcKeepRight parameters are relatively sensitive to the average speed, and the minGap, tau, Speed ​​Factor, lcStrategic, lcAssertive, and lcSpeedGain parameters are relatively sensitive to the number of lane changes, indicating that at different value levels, the above parameters have a significant impact on the average speed and the number of lane changes in complex weaving areas.

[0087] Table 3 Parameter values ​​of connected autonomous driving vehicles

[0088]

[0089] In this embodiment, before constructing the traffic conflict analysis model, step S2 is to establish a traffic conflict simulation in a mixed traffic environment and analyze the influencing factors; first, three assumptions are made: the mixing of the autonomous driving vehicle does not affect the driving operation of the manual driver; no traffic accidents occur during the simulation; the driving behavior parameters of the autonomous driving vehicle are set based on the existing research and the traffic flow characteristics of the complex interweaving area; then, the orthogonal experimental design method is used to construct a simulation experiment orthogonal table for the mixed traffic environment in the complex interweaving area, and uniform sampling is performed according to the pre-set influencing factors and their level values, so as to obtain comprehensive and representative sample data within a limited number of simulation experiments; secondly, the SPSS software is used to perform an orthogonal experimental design on the selected factors and level variables, and the combination of each factor in each experiment is input into SUMO as a simulation experiment input value to run the simulation model; finally, the SSAM software is used to perform conflict analysis, and the conflict data is exported. According to the conflict threshold, the rear-end collision conflict data of different severities and the lane change conflict data of different severities are respectively counted.

[0090] Specifically, in the simulation experiment design, the expected speed is set to 60m / h, the expected speed of the autonomous driving vehicle is a fixed value set by the system, and the speeds of other vehicles are taken as the calibration parameters of the manually driven vehicle model. The hourly traffic volume in the weaving area, the penetration rate of autonomous driving, the ratio of large vehicles, the weaving ratio, the weaving flow ratio, the longitudinal slope of the road, and the weaving length are selected as the influencing factors of the orthogonal experiment to analyze traffic conflicts of different degrees and categories. The experimental level values ​​of each influencing factor are as follows

[0091] Table 4 Table of simulation experiment level values ​​of various influencing factors

[0092]

[0093] Specifically, SPSS software was used to conduct an orthogonal experimental design on the selected 7 factors and 5 levels of variables. The variables with less than 5 levels were processed using the quasi-horizontal method. The blank levels of the road longitudinal slope were filled with 1%. The combination of each factor in each experiment was input into SUMO as a simulation experiment input value to run the simulation model. The simulation duration was 4200s, and the first 600s was used as a simulation warm-up. SSAM software was used to perform conflict analysis, export conflict data, and statistically analyze rear-end collision data of different severity and lane change conflict data of different severity according to the conflict threshold.

[0094] Table 5 Orthogonal experimental design table

[0095]

[0096] In this embodiment, a conflict model is constructed based on the above steps; Stata software is used to analyze the data, and the AIC criterion (Akaike Information Criterion) and BIC criterion (Bayesian Information Criterion) are used as the judgment criteria. The Poisson or negative binomial distribution is selected by overdispersion and goodness of fit to determine the final selected model; the negative binomial distribution explains each conflict; the negative binomial distribution model is established on the basis of the Poisson distribution model. When the data is relatively discrete, considering the existence of errors, the disturbance term εi is added to balance the relationship between the expected value and the variance. At this time, the expected value λ i ' can be expressed as:

[0097] λ i '=exp(β i x i +ε i )=λ i exp(ε i )

[0098] At this time, the probability of the number of random events occurring can be expressed as:

[0099]

[0100] The formula conditional distribution is converted into unconditional distribution to obtain the probability density of the number of traffic conflicts, that is, the negative binomial regression distribution model:

[0101]

[0102] Where: P(x i ) is the occurrence of x at position i i The probability of traffic conflict; α is the parameter to be calibrated.

[0103] Severe Lane Change Conflict Calculation Model

[0104] The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of serious lane-changing conflicts, and the alpha value of the model is 0.063, indicating that it has passed the overdispersion test; therefore, the model is constructed as follows:

[0105] N 1 =exp(1.9225128+0.0020886V-0.5406693MR+0.0131275L

[0106] -0.7760446lR+2.6839670VR)

[0107] Where: N 1 is the number of serious lane change conflicts, veh / h; V is the average hourly traffic volume of the lane, veh / h; MR is the penetration rate of autonomous driving, %; L is the weaving length, m; IR is the weaving ratio; VR is the weaving flow ratio.

[0108] General lane-changing conflict calculation model

[0109] The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of general lane change conflicts, and the alpha value of the model is 0.046, indicating that it has passed the overdispersion test. Therefore, the model is constructed as follows:

[0110] N 2 =exp(2.5629645+0.0009639V-0.5953913MR+0.0171947L

[0111] +2.0972626VR)

[0112] Where: N 2 is the number of general lane change conflicts, pcs / h; V is the average hourly traffic volume of the lane, veh / h; MR is the penetration rate of autonomous driving, %; L is the weaving length, m; VR is the weaving flow ratio;

[0113] Severe rear-end collision calculation model

[0114] The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of serious rear-end collisions, and the alpha value of the model is 0.132, indicating that it has passed the overdispersion test. Therefore, the model is constructed as follows:

[0115] N 3 =exp(0.1211006+0.0035253V+1.6124795VR+0.0070063L)

[0116] Where: N 3 is the number of serious rear-end collisions, veh / h; V is the average hourly traffic volume of the lane, veh / h; VR is the weaving flow ratio; L is the weaving length, m;

[0117] General rear-end collision calculation model

[0118] The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of general rear-end collisions, and the alpha value of the model is 0.034, indicating that it has passed the overdispersion test. Therefore, the model is constructed as follows:

[0119] N 4 =exp(3.5436384+0.0015160V-0.4050758MR+0.0120833L

[0120] -0.6183874IR)

[0121] Where: N 4is the number of general rear-end collisions, veh / h; V is the average hourly traffic volume of the lane, veh / h; MR is the penetration rate of autonomous driving, %; L is the weaving length, m; IR is the weaving ratio.

[0122] Specifically, for typical scenarios such as complex weaving areas with multiple lanes and short spacing in cities, six factors, including average hourly traffic volume of lanes, autonomous driving penetration rate, large vehicle ratio, weaving flow ratio, weaving ratio and weaving length, were selected as alternative variables of the model; the Pearson correlation coefficient was used to test the correlation of the alternative variables.

[0123] Table 6 Description of selected variables

[0124]

[0125] The Pearson correlation coefficient was used to test the correlation of the selected variables. It is generally believed that when the correlation coefficient is greater than 0.8, there is a strong correlation between the two variables. In order to ensure the fitting effect of the model, one of the variables should be eliminated, as shown in Table 8.

[0126] Table 7 Correlation analysis of selected variables

[0127]

[0128] Specifically, referring to the existing research experience in the field of traffic conflicts, according to the traffic conflict investigation, based on the 85% cumulative frequency statistical method, the value corresponding to the 20% quantile cumulative frequency is set as the severe conflict threshold, and the value corresponding to the 85% quantile cumulative frequency is set as the general conflict threshold. See Table 5 for details. The traffic conflicts in complex interweaving areas are divided into four categories: severe lane change conflicts, general lane change conflicts, severe rear-end collision conflicts, and general rear-end collision conflicts, and conflict calculation models are established accordingly;

[0129] Table 8 Conflict indicator thresholds in complex interweaving areas

[0130]

[0131] In this embodiment, in step S2, the traffic conflict analysis model is verified by using SSAM software to extract conflict data from the SUMO simulation experiment results in a mixed traffic environment, and using it as a test reference group. Then, the constructed model is used to calculate the conflict data under different traffic conditions. By calculating the absolute error, relative error and mean absolute error between the two groups of data, the validity of the four traffic conflict calculation models is verified; the mean absolute error MAE is standardized into a percentage form, i.e., MAE%; the calculation formula is as follows:

[0132]

[0133] Where: n is the number of samples; Calculate the value for the model; y i Extract values ​​for simulation results.

[0134] In this embodiment, in step S3, the safety evaluation index is established, and the "conflict occurrence rate" is selected as the evaluation index of the safety status of the complex weaving area. The traffic conflict occurrence rate k i Refers to the number of traffic conflicts N per hour i With traffic volume Q i The specific calculation formula is as follows:

[0135]

[0136] Where: k i is the conflict occurrence rate; i represents the conflict type, i=1, 2, 3, 4, representing severe lane-changing conflict, general lane-changing conflict, severe rear-end collision conflict and general rear-end collision conflict, respectively.

[0137] In this embodiment, the safety evaluation method in step S3 uses grey clustering to evaluate the traffic safety of the complex weaving area under the mixed traffic environment, and constructs a sinusoidal grey clustering safety evaluation model based on combined weighting.

[0138] Specifically, a combined weighted sinusoidal grey clustering safety evaluation model is constructed.

[0139] First, construct the evaluation index matrix; assuming there are n samples, m evaluation indexes, s gray classes, and the value of the evaluation index is x ij , the evaluation matrix is:

[0140]

[0141] Where: x ij is the value of the i-th evaluation object at the j-th indicator; n is the number of evaluation objects, n=49; m is the number of evaluation indicators, m=4.

[0142] According to the traffic conflict incidence formula, the incidence rates of four types of traffic conflicts in 49 groups of experiments were calculated respectively, and the 49*4 evaluation matrix A was formed as shown in the formula:

[0143]

[0144] (1) Construction of Sine Curve Possibility Function

[0145] 1) Gray classification and eigenvalue determination

[0146] The safety status of complex weaving areas is divided into levels I, II, III and IV, which represent the increasing safety level. The eigenvalues ​​corresponding to different gray levels are determined by the cumulative frequency distribution method. The distribution of the incidence rate indicators of severe lane change, general lane change, severe rear-end collision and general rear-end collision is statistically analyzed respectively, and the 15%, 40%, 60% and 85% values ​​of the cumulative frequency curve are selected as the turning points of the possible degree function images of each indicator, that is, the eigenvalues ​​k of different gray classes.

[0147] Table 9 Different gray class eigenvalues ​​of evaluation indicators in complex interweaving areas

[0148]

[0149]

[0150] 2) Construction based on the possibility function of sine curve

[0151] According to the characteristic values ​​of the four security levels, the improved possibility function of different gray categories in the complex interweaving area is constructed. The possibility function corresponding to each security level is determined;

[0152]

[0153] Substitute the different gray class eigenvalues ​​of the four evaluation indicators into the formula to establish the improved possibility function.

[0154] (2) Combination weighting based on fuzzy consistency matrix-entropy method

[0155] 1) Fuzzy consistency matrix weight calculation

[0156] First, construct a fuzzy judgment matrix A. The fuzzy judgment matrix is ​​a matrix whose elements are in the interval [0,1] and is used to represent the fuzzy relationship between elements. ij Indicates index a i Same as index a j Compared with the importance of i is more important than a j When high, a ij The value is 1; when the index a i is more important than a j When low, a ij The value is 0; when the index a i and indicator a j When the importance of ij The value is 0.5.

[0157] A=[a ij ] n×n

[0158] Then, the judgment matrix is ​​converted into the fuzzy consistency matrix C by row and normalization sorting, and its calculation formula is as follows:

[0159] C=[c ij ] n×n

[0160] in,

[0161]

[0162] Finally, the elements in the fuzzy consistency matrix are added row by row and then normalized. The normalized value is the weight corresponding to the evaluation index. The relative weight of each index is:

[0163] ω j =[0.326 0.174 0.326 0.174]

[0164] 2) Entropy method weight calculation

[0165] When the probability of each state is p i (i=1,2,…,n), the system entropy is as follows:

[0166]

[0167] First, determine the original data matrix according to the indicator system Its standardized values ​​constitute the decision matrix C = (C ij ) m×n .

[0168] Then, calculate the proportion of the jth solution under the i-th indicator:

[0169]

[0170] Calculate the entropy value of the i-th indicator, where the constant k = 1 / ln(n):

[0171]

[0172] Calculate the weights of different indicators:

[0173]

[0174] The selected indicators are all negative indicators, but the sample data of some indicators are highly discrete, so the extreme value method is used to standardize the indicators to make the data more suitable for comprehensive comparison and analysis. The calculation formula of the negative indicator extreme value method is as follows:

[0175]

[0176] After standardization, the result is shifted to the right by 0.0001 units, and the relative weight of each indicator is obtained based on the entropy weight calculation method.

[0177] 3) Combination weight calculation

[0178] The combination of subjective and objective weighting methods, on the one hand, makes up for the strong subjectivity and randomness of the fuzzy consistency matrix, and on the other hand, avoids the defect of the entropy method that pays too much attention to data and ignores the indicator itself. In order to comprehensively consider the influence of both, the comprehensive weight calculation of the four types of conflict incidence indicators is carried out, ρ = 0.6.

[0179] w i =ρw i1 +(1-ρ)w i2

[0180] Where: w i is the comprehensive weight of each indicator; ρ is the correction coefficient; w i1 、w i2 are the weights of the fuzzy consistency matrix method and entropy method for each indicator.

[0181] Finally, the comprehensive weight results of each indicator are obtained through the combined weighting method.

[0182] (3) Construction of comprehensive clustering coefficient matrix

[0183] is the possible degree function of the jth indicator of observation object i in gray class k, η j is the weight of the j index, then the clustering coefficient of the observed object i with respect to the kth gray class is The calculation formula is shown in the formula:

[0184]

[0185] According to the above formula, the grey fixed weight clustering coefficient matrix is ​​constructed. Based on the evaluation matrix A, after determining the eigenvalue and possibility function of the gray class, the clustering coefficient of the evaluation object i (i = 1, 2, ... 49) in the gray class k (k = 1, 2, 3, 4) is calculated according to the clustering weight. The clustering coefficients of the 49 evaluation objects are calculated as follows:

[0186]

[0187] In this embodiment, step S3 constructs the security level division by comparing the clustering coefficients corresponding to different gray classes. The gray class to which the maximum clustering coefficient belongs is the security risk level of the experimental scheme. The sample values ​​of different evaluation indicators under different security levels in the complex interweaving area are counted. The maximum evaluation index value is used as the threshold for levels one, two, and three to determine the evaluation index thresholds for each security level in the intelligent network mixed traffic environment, as shown in the following table

[0188] Table 10 Evaluation index thresholds

[0189]

[0190] Beneficial Effects

[0191] 1) This safety evaluation method has high accuracy. The previous safety evaluation method was used to evaluate the sample data of this patent, and the accuracy of the evaluation method in this paper was compared. The safety evaluation method constructed by this patent, through subjective and objective combined weighting, considers the differences in the importance of evaluation indicators while considering the sample data information, making the evaluation results more credible; and by constructing a sinusoidal possibility function, the gray class division intervals are clearer and the safety level judgment is more accurate.

[0192] The index thresholds divided by the evaluation method are verified based on simulation experimental data, as shown in Table 7, Table 8, and Table 9. The evaluation accuracy is 98%, which is 20% and 4% higher than that of the conventional linear grey variable weight clustering and fixed weight clustering evaluation methods, respectively.

[0193] Table 11 Test results of safety evaluation level division using conventional linear grey variable weight clustering evaluation method

[0194]

[0195]

[0196] Table 12 Comparison of safety evaluation level division test of conventional linear grey fixed weight clustering evaluation method

[0197]

[0198] Table 13 Comparison of safety evaluation level division test based on combined weighted curve grey clustering evaluation method

[0199]

[0200]

[0201] (2) A calculation model for four types of traffic conflicts, namely severe lane changing, general lane changing, severe rear-end collision, and general rear-end collision, was constructed in complex weaving areas under intelligent connected mixed traffic environments. By quantifying the relationship between the geometric conditions of complex traffic weaving areas, traffic flow parameters, and traffic conflicts, the number of traffic conflicts under different traffic conditions can be obtained more conveniently.

[0202] (3) An evaluation index system and method for traffic safety in complex weaving areas under intelligent connected mixed traffic environments were established, which can quantify the traffic safety level in complex weaving areas under different autonomous driving penetration rates.

[0203] (4) Currently, there is no research on the traffic safety risks in complex weaving areas after the introduction of autonomous vehicles, which cannot provide a scientific basis for the technical improvement of intelligent connected vehicles and the formulation of intelligent traffic management strategies. This patent focuses on the evaluation of the impact of intelligent connected mixed traffic flow on the safety risks of complex weaving areas, which has good theoretical significance and practical application value.

[0204] Finally, it should be noted that those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A safety risk assessment method for complex interweaving areas in an intelligent connected mixed driving environment. It is characterized in that The following steps are involved: S1: Build a traffic simulation model, including scene construction and vehicle parameter calibration; S2: Build and verify the traffic conflict analysis model; S3: Establish traffic safety evaluation indicators and methods, and construct a safety level classification method.

2. According to the method for safety risk assessment of complex interweaving areas in an intelligent connected hybrid driving environment as described in claim 1, It is characterized in that In step S1, the road geometry parameters obtained from field research and the processed traffic flow data are used to accurately control the flow, speed, path, and vehicle type parameters of the weaving area by lane; the road network modeling is performed using the graphical road network editing tool netedit according to the geometric parameters of the weaving area; Combined with the actual survey traffic flow data, the routing file is created with parameter settings such as vehicle type, traffic path, vehicle input, expected speed distribution, and additional files.

3. According to claim 2, a safety risk assessment method for complex interweaving areas in an intelligent connected hybrid driving environment, It is characterized in that According to the above steps, the calibration of parameters of the manually driven vehicle and the calibration of simulation parameters of the autonomous driving vehicle are determined; In the manual driving vehicle parameter calibration, the IDM following model is selected as the following model, and the lane changing model adopts LC2013; the expression of the IDM following model is as follows In the formula, a s is the acceleration of the following vehicle; a m is the maximum acceleration of the vehicle; v 0 is the expected speed; v S is the speed of the following vehicle; δ is the vehicle acceleration index; s 0 is the headway time under congestion; T is the safe headway; Δv is the speed difference between adjacent following vehicles; Δx is the headway distance; b is the expected deceleration; the IDM model setting parameters are deceleration (m / s2) is 4.4; the maximum expected speed (m / s2) is 5.56; the minimum headway distance (m) is 2.5; the expected headway distance (s) is 1; the speed factor is 1; the driver's driving proficiency is 0.5; the speed deviation is 0.1; The CACC following model is used as the following model in the simulation parameter calibration of the autonomous driving vehicle, and the SL2013 lane changing model is used as the lane changing model. The expression of the CACC following model is as follows: e s (t-1)=Δx(t-1)-t c v s (t-1)-l-s 0 Where: e s (t-1) is the headway error of the vehicle at time t-1; Δx(t-1) is the headway between the following vehicle and the preceding vehicle at time t-1; t c is the expected headway between vehicles; l is the vehicle length; s 0 is the headway distance in congestion; v s (t) is the speed of the vehicle at time t; v s (t-1) is the speed of the vehicle at time t-1; is the inter-vehicle distance error increment at time t-1; k p and k d are adjustment coefficients, which are 0.45 and 0.25 respectively.

4. According to claim 3, a safety risk assessment method for complex interweaving areas in an intelligent connected hybrid driving environment, It is characterized in that Sensitivity analysis is performed on the minimum vehicle spacing (minGap), expected minimum headway (tau), speed factor (Speed ​​Factor) in the IDM following model in the above steps, and strategic lane change (lcStrategic), cooperative lane change (lcCooperative), speed-up lane change (lcSpeedGain), minimum acceptance spacing for lane change (lcAssertive), and keep right (lcKeepRight) in the SL2013 lane change model; the ratio of the difference between the maximum and minimum values ​​to the mean is used as the mathematical indicator for sensitivity analysis of the correction parameters; the following formula is used According to the simulation sensitivity analysis results, the correction index can be determined to calibrate the core parameters.

5. According to claim 4, a safety risk assessment method for complex interweaving areas in an intelligent networked mixed driving environment, It is characterized in that Step S2: Before constructing the traffic conflict analysis model, a traffic conflict simulation in a mixed traffic environment should be established and the influencing factors should be analyzed. First, three assumptions should be made: the mixing of the autonomous driving vehicle does not affect the driving operation of the manual driver; no traffic accidents occur during the simulation; the driving behavior parameters of the autonomous driving vehicle are set based on the existing research and the traffic flow characteristics of the complex interweaving area; then, the orthogonal experimental design method is used to construct a simulation experiment orthogonal table for the mixed traffic environment in the complex interweaving area, and uniform sampling is performed according to the pre-set influencing factors and their level values, so as to obtain comprehensive and representative sample data within a limited number of simulation experiments; secondly, the SPSS software is used to perform an orthogonal experimental design on the selected factors and level variables, and the combination of each factor in each experiment is input into SUMO as a simulation experiment input value to run the simulation model; finally, the SSAM software is used to perform conflict analysis, and the conflict data is exported. According to the conflict threshold, the rear-end collision conflict data of different severities and the lane change conflict data of different severities are respectively counted.

6. According to claim 5, a safety risk assessment method for complex interweaving areas in an intelligent networked hybrid driving environment, It is characterized in that Based on the above steps, the conflict model was constructed; Stata software was used to analyze the data, and the AIC criterion (Akaike Information Criterion) and BIC criterion (Bayesian Information Criterion) were used as the judgment criteria. The Poisson or negative binomial distribution was compared through overdispersion and goodness of fit to determine the final model; the negative binomial distribution explained each conflict; the negative binomial distribution model was established on the basis of the Poisson distribution model. When the data was relatively discrete, the disturbance term ε was added to take into account the existence of errors. i Balance the relationship between expected value and variance. At this time, the expected value λ i ' can be expressed as: l i '=exp(β i x i +e i )=λ i exp(e i ) At this time, the probability of the number of random events occurring can be expressed as: The formula conditional distribution is converted into unconditional distribution to obtain the probability density of the number of traffic conflicts, that is, the negative binomial regression distribution model: Where: P(x i ) is the occurrence of x at position i i The probability of traffic conflict; α is the parameter to be calibrated. Severe Lane Change Conflict Calculation Model The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of serious lane-changing conflicts, and the alpha value of the model is 0.063, indicating that it has passed the overdispersion test; therefore, the model is constructed as follows: N 1 =exp(1.9225128+0.0020886V-0.5406693MR+0.0131275L -0.7760446lR+2.6839670VR) Where: N 1 is the number of serious lane change conflicts, veh / h; V is the average hourly traffic volume of the lane, veh / h; MR is the penetration rate of autonomous driving, %; L is the weaving length, m; IR is the weaving ratio; VR is the weaving flow ratio. General lane-changing conflict calculation model The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of general lane change conflicts, and the alpha value of the model is 0.046, indicating that it has passed the overdispersion test. Therefore, the model is constructed as follows: <h2 style=";text-align:left;direction:ltr">N<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> =exp(2.5629645+0.0009639V-0.5953913MR+0.0171947L+2.0972626VR) Where: N 2 is the number of general lane change conflicts, pcs / h; V is the average hourly traffic volume of the lane, veh / h; MR is the penetration rate of autonomous driving, %; L is the weaving length, m; VR is the weaving flow ratio; Severe rear-end collision calculation model The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of serious rear-end collisions, and the alpha value of the model is 0.132, indicating that it has passed the overdispersion test. Therefore, the model is constructed as follows: N 3 =exp(0.1211006+0.0035253V+1.6124795VR+0.0070063L) Where: N 3 is the number of serious rear-end collisions, veh / h; V is the average hourly traffic volume of the lane, veh / h; VR is the weaving flow ratio; L is the weaving length, m; General rear-end collision calculation model The P value of each variable is less than 0.05, indicating that there is a significant correlation between the variables in the table and the number of general rear-end collisions, and the alpha value of the model is 0.034, indicating that it has passed the overdispersion test. Therefore, the model is constructed as follows: N 4 =exp(3.5436384+0.0015160V-0.4050758MR+0.0120833L -0.6183874IR) Where: N 4 is the number of general rear-end collisions, veh / h; V is the average hourly traffic volume of the lane, veh / h; MR is the penetration rate of autonomous driving, %; L is the weaving length, m; IR is the weaving ratio.

7. According to claim 6, a safety risk assessment method for complex interweaving areas in an intelligent networked hybrid driving environment, It is characterized in that In step S2, the traffic conflict analysis model is verified by using SSAM software to extract conflict data from the SUMO simulation experiment results in a mixed traffic environment, and using it as a test reference group. Then, the constructed model is used to calculate the conflict data under different traffic conditions. By calculating the absolute error, relative error and mean absolute error between the two groups of data, the validity of the four traffic conflict calculation models is verified. The mean absolute error MAE is standardized into a percentage form, namely MAE%. The calculation formula is as follows: Where: n is the number of samples; Calculate the value for the model; y i Extract values ​​for simulation results.

8. According to claim 7, a method for safety risk assessment of complex interweaving areas in an intelligent networked mixed driving environment, It is characterized in that In step S3, the safety evaluation index is established, and the "conflict occurrence rate" is selected as the evaluation index of the safety status of the complex weaving area. The traffic conflict occurrence rate k i Refers to the number of traffic conflicts N per hour i With traffic volume Q i The specific calculation formula is as follows: Where: k i for the incidence of conflict; i represents the conflict type, i=1, 2, 3, 4, representing severe lane-changing conflict, general lane-changing conflict, severe rear-end collision conflict, and general rear-end collision conflict, respectively.

9. According to claim 8, a safety risk assessment method for complex interweaving areas in an intelligent networked hybrid driving environment, It is characterized in that In step S3, the safety evaluation method uses grey clustering to evaluate the traffic safety of complex interweaving areas under mixed traffic conditions, and constructs a sinusoidal grey clustering safety evaluation model based on combined weighting.

10. According to claim 9, a method for safety risk assessment of complex interweaving areas in an intelligent networked hybrid driving environment, It is characterized in that Step S3 constructs the safety level division by comparing the clustering coefficients corresponding to different gray classes. The gray class to which the maximum clustering coefficient belongs is the safety risk level of the experimental scheme. The sample values ​​of different evaluation indicators under different safety levels in complex interweaving areas are statistically analyzed. The maximum evaluation index value is used as the threshold for levels one, two, and three to determine the evaluation index thresholds for each safety level in the intelligent connected mixed traffic environment.

Citation Information

Cited By

  • Interlacing area lane change safety evaluation method, terminal equipment and storage medium

    CN119580531A

  • A lane-changing safety evaluation method, terminal device and storage medium in an interweaving area

    CN119580531B

  • Risk real-time assessment method and system for confluence area and diversion area in man-machine mixed driving environment

    CN121811657A