A safety evaluation and optimization method for mountainous roads

Abstract

The present application relates to a kind of mountain road safety evaluation and optimization method, comprising: obtaining road alignment parameter and the motion state of vehicle;Road operation index prediction model is constructed, significant variable is selected from potential explanatory variable and optimal upstream and downstream distance is selected;Complex road operation index prediction model is used to verify significant variable;If verification is passed, vehicle operation index is predicted based on road operation index prediction model, if verification is not passed, road operation index prediction model is reconstructed and the significant variable of re-determination is verified, until verification is passed;Accident substitution index is calculated;Based on accident substitution index, using descriptive statistical method, the relationship between road alignment parameter and accident substitution index is determined, the safety of road is evaluated, and road alignment parameter is optimized.Compared with prior art, the present application can realize prospective road safety evaluation and optimization in road design stage.

Description

Technical Field

[0001] This invention relates to the field of road safety assessment and optimization, and in particular to a method for safety assessment and optimization of mountain roads. Background Technology

[0002] With the rapid development of my country's expressways, expressway traffic safety has received increasing attention from all sectors of society. The main construction areas for future highways in my country will shift to the central and western regions, where complex terrain conditions will lead to a wider application of combined alignment designs. However, existing highway design and safety evaluation standards do not adequately consider driver safety and impose only simplistic constraints on alignment design values.

[0003] There is a significant relationship between road alignment design and traffic accidents. Road curve radius, sight distance, maximum gradient and slope length, and unfavorable alignment combinations are key factors considered in road design safety evaluations. This is because drivers, while driving, take appropriate driving actions to control vehicle operation based on the road environment information they receive. Incompatible road alignments, such as abrupt changes in alignment design elements on adjacent road sections, can not only lead to unconscious dangerous driving behaviors but may also catch drivers off guard and cause improper driving operations.

[0004] In traditional road linear safety research, many existing speed models employ speed prediction strategies based on constructing speed feature points from fixed-point data. These existing prediction models have two main drawbacks: First, the segmentation of road sections disrupts the continuity of road alignment changes. Drivers do not repeatedly accelerate and decelerate within road segment units; instead, they are influenced by the road segment with the greatest road alignment constraints, maintaining inertia throughout the driving process. However, existing prediction models analyze roads by dividing them into different units, affecting their applicability. Second, the assumed speed feature points are not necessarily extreme points of speed change. For example, researchers have found that drivers do not always reach their minimum speed near the midpoint when traversing horizontal curves or sloping sections, thus introducing a bias in the baseline of existing prediction models.

[0005] Accidents are the most direct indicator for safety assessment, but research on the estimation of accidents based directly on linear parameters has some shortcomings. On the one hand, it requires a large number of accident samples and research sites as support. On the other hand, due to the randomness and low probability of accidents, the correlation between accidents and linear design parameters is often not significant. Summary of the Invention

[0006] The purpose of this invention is to provide a method for safety assessment and optimization of mountain roads.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] S1. Obtain the road alignment parameters and vehicle motion status of the target road segment;

[0009] S2. Construct dependent variables based on vehicle motion state and independent variables based on road alignment parameters, which are then used as potential explanatory variables.

[0010] S3. Construct a road operation index prediction model, solve the multicollinearity of variables based on the variance amplification factor, and screen significant variables from potential explanatory variables and the optimal upstream and downstream distances based on the model fit and the traffic meaning of the variables.

[0011] S4. Based on the differences among drivers, a complex road operation index prediction model is used to verify significant variables;

[0012] S5. If the verification passes, predict vehicle operation indicators based on the road operation indicator prediction model. If the verification fails, repeat steps S3-S5, rebuild the road operation indicator prediction model, and verify the newly determined significant variables until the verification passes.

[0013] S6. Calculate accident alternative indicators based on vehicle operation indicators;

[0014] S7. Based on the road alignment parameters of the target road segment, construct road segments with different alignment parameters, and based on the accident substitution index, use descriptive statistical methods to determine the relationship between the road alignment parameters and the accident substitution index, evaluate the road safety, and optimize the road alignment parameters.

[0015] The road alignment parameters include slope, curvature, whether it is a slope change point, and whether it is an S-curve.

[0016] The vehicle's motion state includes its operating speed, longitudinal acceleration, lateral acceleration, and lane departure value;

[0017] The dependent variables include operating speed, longitudinal acceleration, lateral acceleration, and lane departure level;

[0018] When constructing independent variables, the aggregated values ​​of curvature and slope of the upstream and downstream road segments at pre-configured distance intervals are extracted as potential explanatory variables.

[0019] The aggregate values ​​include extreme values, average values, proportions, and categorical variables.

[0020] The road operation index prediction model is constructed in two parts:

[0021] A multiple linear regression model was constructed for the vehicle's speed, longitudinal acceleration, and lateral acceleration:

[0022] y = β0 + β1x1 + β2x2 … + β p xp +ε

[0023] Where β0 is the regression constant, β1…β p These are regression coefficients, where y is the dependent variable, and x1, x2…x p There are p independent variables that can be precisely measured and controlled, and ε is the random error;

[0024] Construct an ordered Logit model for lane departure levels:

[0025]

[0026] Where, p j The dependent variable is the probability of the top j categories, α. j It is a constant of the j-th class, γ1…γ p These are model coefficients.

[0027] The complex road operation index prediction model consists of two parts:

[0028] A hybrid linear model is constructed for the vehicle's speed, longitudinal acceleration, and lateral acceleration:

[0029] y k =λ k0 +λ k1 x1+…λ kj x j …+λ kp x p +

[0030] λ[driver k ]0+λ[driver k ]1x1+…λ[driver k ] j x j …+λ[driver k ] p x p

[0031] Where, λ k0 Let λ be the intercept term for the k-th driver. kj To fix the coefficient of the independent variable, λ[driver] k ]0 is the random parameter intercept of the k-th driver, λ[driver k ] j Let y be the random independent variable parameter for the k-th driver, j = 1, 2, ..., p, and y be the dependent variable, x1, x2, ..., xn. p There are p independent variables that can be precisely measured and controlled;

[0032] Construct a hierarchical ordered Logit model for lane departure levels:

[0033]

[0034] η kj =η k +b j

[0035] Where, η k b represents the intercept variable for all groups. j Let μ be a random variable with a mean of zero, following a Gamma(0.01,0.01) distribution, and be used as a random effect to characterize within-group differences. j These are the model coefficients.

[0036] The accident substitution indicators include the difference between the operating speed and the design speed, the difference in speed across the section, the 85th quantile of the acceleration / deceleration rate, the speed variance, the minimum longitudinal acceleration, the proportion of no deviation, and the proportion of severe deviation.

[0037] The descriptive statistical method is a box plot.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] (1) The present invention establishes a continuous modeling strategy that takes into account the characteristics of upstream and downstream, which can more accurately describe the continuous change of speed on combined linear road sections.

[0040] (2) This invention selects potential explanatory variables for the distances between upstream and downstream aggregates, constructs a multiple linear regression model and an ordered Logit model, and screens out significant variables. In order to gain a deeper understanding of individual differences among drivers and explore the response of extreme driver samples to changes in road alignment, the significant variables are verified using a mixed linear model and a hierarchical ordered Logit model.

[0041] (3) This invention selects road operation indicators as observation indicators and selects typical accident substitute indicators for research, which can meet the sample size required for modeling and is easier to observe and collect than accidents with randomness and low probability. Attached Figure Description

[0042] Figure 1 This is a flowchart of the method of the present invention;

[0043] Figure 2 This is a box plot showing the difference between the radius of the uphill section of a horizontal curve and the vehicle speed in an embodiment of the present invention;

[0044] Figure 3 This is a box plot showing the radius of the uphill section of a horizontal curve and the minimum longitudinal acceleration in an embodiment of the present invention;

[0045] Figure 4This is a box plot showing the radius of the uphill section of the horizontal curve and the maximum lateral acceleration 85th in an embodiment of the present invention;

[0046] Figure 5 This is a box plot showing the radius of the uphill section of the horizontal curve and the ratio of the no-offset ratio in an embodiment of the present invention;

[0047] Figure 6 This is a box plot showing the average radius of the uphill section of the horizontal curve and the proportion of severe deviation in an embodiment of the present invention. Detailed Implementation

[0048] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0049] This embodiment provides a method for safety assessment and optimization of mountain roads, such as... Figure 1 As shown, it includes the following steps:

[0050] S1. Data Preparation: Obtain the road alignment parameters and vehicle motion status of the target road segment. The road alignment parameters include slope, curvature, whether it is a slope change point, and whether it is an S-curve. The vehicle motion status includes the vehicle's running speed, longitudinal acceleration, lateral acceleration, and lane departure value.

[0051] S2. Construct dependent variables based on vehicle motion state and independent variables based on road alignment parameters as potential explanatory variables.

[0052] In this embodiment, the dependent variables include running speed, longitudinal acceleration, lateral acceleration, and lane departure level.

[0053] When constructing independent variables, aggregated values ​​(such as extreme values, average values, proportions, and categorical variables) of the curvature and slope of the road segments at upstream and downstream points of 50m, 100m, 200m, 300m, 400m, and 500m were extracted as potential explanatory variables.

[0054] S3. Construct a road operation index prediction model, solve the multicollinearity of variables based on the variance amplification factor, and screen significant variables from potential explanatory variables and the optimal upstream and downstream distances based on the model fit and the traffic meaning of the variables.

[0055] A multiple linear regression model was constructed for the vehicle's speed, longitudinal acceleration, and lateral acceleration:

[0056] y = β0 + β1x1 + β2x2 … + β p x p +ε

[0057] Where β0 is the regression constant, β1…β p These are regression coefficients, where y is the dependent variable, and x1, x2…x p There are p independent variables that can be precisely measured and controlled, and ε is the random error.

[0058] That is, a multiple linear regression model is constructed for the vehicle's running speed, longitudinal acceleration, and lateral acceleration, and the parameters of each model are determined.

[0059] Construct an ordered Logit model for lane departure levels:

[0060]

[0061] Where, p j The dependent variable is the probability of the top j categories, α. j It is a constant of the j-th class, γ1…γ p These are model coefficients.

[0062] S4. Based on the differences among drivers, a complex road operation index prediction model is used to verify significant variables.

[0063] A mixed linear model is constructed for the vehicle's operating speed, longitudinal acceleration, and lateral acceleration. This mixed linear model, based on the linear model, assumes heterogeneity in the variance of the observed samples; in this invention, individual differences among drivers are classified as between-group differences. The specific model structure is as follows:

[0064] y k =λ k0 +λ k1 x1+…λ kj x j …+λ kp x p +

[0065] λ[driver k ]0+λ[driver k ]1x1+…λ[driver k ] j x j …+λ[driver k ] p x p

[0066] Where, λ k0 Let λ be the intercept term for the k-th driver. kj To fix the coefficient of the independent variable, λ[driver] k ]0 is the random parameter intercept of the k-th driver, λ[driver k ] jLet y be the random independent variable parameter for the k-th driver, j = 1, 2, ..., p, and y be the dependent variable, x1, x2, ..., xn. p There are p independent variables that can be precisely measured and controlled.

[0067] A hierarchical ordered Logit model is constructed for lane departure levels. This hierarchical ordered Logit model adds a layer to the ordered Logit model to reflect inter-group differences. This ensures that the model uses the same solution and the same set of model coefficients when considering the same group, while using different intercepts and different model coefficients between different groups. Specifically, this invention uses a fixed intercept η. k Introducing random variables, the structure of the hierarchical ordered logit model is as follows:

[0068]

[0069] η kj =η k +b j

[0070] Where, η k b represents the intercept variable for all groups. j Let μ be a random variable with a mean of zero, following a Gamma(0.01,0.01) distribution, and be used as a random effect to characterize within-group differences. j These are the model coefficients.

[0071] S5. If the verification passes, predict vehicle operation indicators based on the road operation indicator prediction model. If the verification fails, repeat steps S3-S5 to rebuild the road operation indicator prediction model and verify the newly determined significant variables until the verification passes.

[0072] S6. Calculate accident alternative indicators based on vehicle operation indicators.

[0073] The accident substitution indicators in this embodiment include the difference between the operating speed and the design speed, the difference in speed at the cross section, the 85th quantile of the acceleration / deceleration rate, the speed variance, the minimum longitudinal acceleration, the proportion of no deviation, and the proportion of severe deviation.

[0074] S7. Based on the road alignment parameters of the target road segment, construct road segments with different alignment parameters, and based on the accident substitution index, use descriptive statistical methods such as box plots to determine the relationship between the road alignment parameters and the accident substitution index, evaluate the road safety, and optimize the road alignment parameters.

[0075] The research section used in this embodiment is the Yongji Expressway in Hunan Province. Based on the design drawings of the Yongzhou-Jishou section of the expressway from the Hunan Provincial Road Survey and Design Institute and vehicle video recordings of similar local road sections, road alignment parameter information was collected. This mountainous expressway is 23.96 km long, with a speed limit of 100 km / h, four lanes in both directions, and adopts a separated roadbed design. Potential explanatory variables for upstream and downstream distances of 50m, 100m, 200m, 300m, 400m, and 500m are constructed, as shown in Table 1. Road operation index change curves were collected from 38 valid driver samples based on driving simulation experiments. Lane deviation values ​​were divided into three levels: less than 0.25m (no deviation), greater than 0.25m but less than 0.5m (moderate deviation), and greater than 0.5m (severe deviation).

[0076] Table 1 Explanatory Variables for Potential Road Alignment Parameters

[0077]

[0078] Multiple linear regression models were used to model vehicle speed, longitudinal acceleration, and lateral acceleration, while an ordered logit model was used to model lane departure level. After addressing spatial autocorrelation and multicollinearity, the adjusted R-squared of the model fit was calculated, and the upstream and downstream ensemble distances with the largest adjusted R-squared were selected. The optimal ensemble distance was 300m for vehicle speed, 400m for longitudinal acceleration, 50m for lateral acceleration, and 300m for lane departure level. Considering driver differences, a mixed linear model and a hierarchical ordered logit model were used to test the significance of the selected variables. The models used and the selected optimal significant variables and their coefficients are shown in Table 2.

[0079] Table 2. Significant independent variables and coefficients of the prediction model

[0080]

[0081] Five accident surrogate indicators were selected: mean speed difference of road segments, minimum longitudinal acceleration, 85th percentile of maximum lateral acceleration, proportion of no deviation, and proportion of severe deviation. These accident surrogate indicators passed correlation analysis and Tobit model testing, demonstrating their effectiveness on the relevant road segments.

[0082] The parameters that determine the combined alignment of the road segment include: the radius of the circular curve R, the spiral parameter A, the deflection angle of the circular curve angle, the front slope grade1, the back slope grade2, and the relative position P. Uphill horizontal curves, downhill horizontal curves, convex horizontal curves, and concave horizontal curves are constructed under different parameters, and corresponding box plots are drawn and analyzed. Figures 2-6 It shows the relationship between the upslope radius of the curve and the surrogate indicators for five types of accidents.

[0083] Comprehensive analysis reveals that when the radius of the circular curve is less than 1000m, as the radius decreases, the absolute value of the speed difference in the road segment increases significantly, the minimum longitudinal acceleration decreases significantly, the maximum lateral acceleration increases significantly, the proportion of no deviation decreases, and the proportion of severe deviation increases. Both the longitudinal smoothness and lateral stability of the road segment decline. However, when the radius of the circular curve is greater than 1000m, the radius has almost no impact on the safety of the combined alignment. Therefore, for uphill sections with horizontal curves, an effective way to improve road segment safety is to choose a larger circular curve radius. Considering factors such as cost, a circular curve radius of around 1000m would be an optimal range.

[0084] Similarly, for each combination type, a comprehensive analysis can be conducted on the relationship between various road alignment parameters and various accident surrogate indicators to evaluate road safety and optimize road alignment parameters.

[0085] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for safety assessment and optimization of mountain roads, characterized in that, Includes the following steps: S1. Obtain the road alignment parameters and vehicle motion status of the target road segment; S2. Construct dependent variables based on vehicle motion state and independent variables based on road alignment parameters, which are then used as potential explanatory variables. S3. Construct a road operation index prediction model, solve the multicollinearity of variables based on the variance amplification factor, and screen significant variables from potential explanatory variables and the optimal upstream and downstream distances based on the model fit and the traffic meaning of the variables. S4. Based on the differences among drivers, a complex road operation index prediction model is used to verify significant variables; S5. If the verification passes, predict vehicle operation indicators based on the road operation indicator prediction model. If the verification fails, repeat steps S3-S5, rebuild the road operation indicator prediction model, and verify the newly determined significant variables until the verification passes. S6. Calculate accident alternative indicators based on vehicle operation indicators; S7. Based on the road alignment parameters of the target road segment, construct road segments with different alignment parameters, and based on the accident substitution index, use descriptive statistical methods to determine the relationship between the road alignment parameters and the accident substitution index, evaluate the road safety, and optimize the road alignment parameters. The road operation index prediction model is constructed in two parts: A multiple linear regression model was constructed for the vehicle's speed, longitudinal acceleration, and lateral acceleration: in, It is the regression constant. It is the regression coefficient. It is the dependent variable. yes An independent variable that can be precisely measured and controlled. It is random error; Construct an ordered Logit model for lane departure levels: in, It is the dependent variable, and is the previous variable. The probability of different categories It is the first Category of constants These are model coefficients; The complex road operation index prediction model consists of two parts: A hybrid linear model is constructed for the vehicle's speed, longitudinal acceleration, and lateral acceleration: in, For the first The intercept term for each driver, The coefficients of the fixed independent variable, For the first The random parameter intercept of each driver, For the first Random independent variable parameters for each driver, , It is the dependent variable. yes A single independent variable that can be precisely measured and controlled; Construct a hierarchical ordered Logit model for lane departure levels: in, The intercept variable represents all groups. Let be a random variable with a mean of zero, following a Gamma(0.01, 0.01) distribution, and be used as a random effect to characterize within-group differences. These are the model coefficients.

2. The method for safety assessment and optimization of mountain roads according to claim 1, characterized in that, The road alignment parameters include slope, curvature, whether it is a slope change point, and whether it is an S-curve.

3. The method for safety assessment and optimization of mountain roads according to claim 1, characterized in that, The vehicle's motion state includes its speed, longitudinal acceleration, lateral acceleration, and lane departure value.

4. The method for safety assessment and optimization of mountain roads according to claim 3, characterized in that, The dependent variables include operating speed, longitudinal acceleration, lateral acceleration, and lane departure level.

5. The method for safety assessment and optimization of mountain roads according to claim 2, characterized in that, When constructing independent variables, the aggregated values ​​of curvature and slope of the upstream and downstream road segments at pre-configured distance intervals are extracted as potential explanatory variables.

6. The method for safety assessment and optimization of mountain roads according to claim 5, characterized in that, The aggregate values ​​include extreme values, average values, proportions, and categorical variables.

7. The method for safety assessment and optimization of mountain roads according to claim 1, characterized in that, The accident substitution indicators include the difference between the operating speed and the design speed, the difference in speed across the section, the 85th quantile of the acceleration / deceleration rate, the speed variance, the minimum longitudinal acceleration, the proportion of no deviation, and the proportion of severe deviation.

8. The method for safety assessment and optimization of mountain roads according to claim 1, characterized in that, The descriptive statistical method is a box plot.

Images