A driver stress training scenario generation system based on highway accident data

Abstract

The present invention provides a driver stress training scenario generation system based on massive highway accident data. The system pre-processes the collected traffic accident data, establishes a stress scenario library based on the massive highway accident data, constructs corresponding indicators of the scenario library, classifies the scenario library by clustering, establishes a scenario generation method model, and performs three-level screening based on driver customization. The model driver stress training scenario generation system disclosed by the present invention provides drivers with more targeted and rich stress scenarios, which can effectively reduce the accident rate.

Description

Technical Field

[0001] The present invention relates to the field of driving technology, and in particular to a driver stress training scenario generation system driven by highway accident data. Technical Background

[0002] With rapid economic development, the number of motor vehicles and drivers in my country has increased annually, and road traffic safety has attracted widespread attention from all sectors of society. Drivers' untimely response to dangerous stressful situations or incorrect operation during real-world driving have become a major cause of accidents. Therefore, studying the generation of driving stress scenarios and using them for driver stress response training has far-reaching significance for improving drivers' stress response capabilities during real-world driving.

[0003] In existing research on the generation of driving stress scenarios, the generated driving scenarios are mostly used for autonomous driving tests. The number of scenarios is small and most of them are generated based on subjective feelings. They lack the support of real road accident data and have a low degree of fit with the stress response scenarios of real drivers. Summary of the Invention

[0004] In response to the problems existing in existing scenario generation, the present invention provides a driver stress training scenario generation system based on massive highway accident data. The purpose is to establish a stress scenario library based on massive highway accident data, and provide drivers with more targeted and rich stress scenarios for training based on the driver's customized "three-layer screening" extraction mode, thereby reducing the accident rate.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A driver stress training scenario generation system based on highway accident data includes the following steps:

[0007] S1. Preprocessing the collected traffic accident data;

[0008] S2. Construct corresponding indicators of the scenario library;

[0009] S3, using clustering to classify the scene library;

[0010] S4. Establish a scene generation method model.

[0011] Furthermore, S1 specifically includes:

[0012] S11. Select the accident type, and select two types: vehicle-pedestrian accident and vehicle-to-vehicle accident;

[0013] S12: First, process the vehicle parameters, process the vehicle motion state in combination with the turn signal state, and then generate the relative motion direction in combination with the high-risk driving behavior;

[0014] S13. Select driving stress scenario factors.

[0015] Furthermore, the processing process of the vehicle motion state in S12 is specifically as shown in the following table:

[0016] Motion state processing

[0017]

[0018] The generation process of the relative motion direction is specifically shown in the following table:

[0019] Generation of relative motion direction

[0020]

[0021] Furthermore, the driving stress scenario factors described in S13 are selected from four aspects: vehicle factors, traffic conflict factors, road factors, and environmental factors, as shown in the following table:

[0022] Driving stress scenario factors

[0023]

[0024] Furthermore, S2 specifically includes:

[0025] S21. Construction of risk index:

[0026]

[0027]

[0028] Where, is the average number of deaths and S 2 is the variance of the number of deaths in the scene, X i is the number of deaths in the i-th accident, and N is the frequency of accidents involved in the scene.

[0029] S22. Organize and hierarchize environmental factors, road factors, vehicle factors, and traffic conflict factors to construct an orderly structural model and build a complexity index:

[0030]

[0031] In the formula, C represents the complexity, α i represents the complexity weight of the i-th category of the influencing factor classification, β ij represents the complexity weight of the jth parameter in the i-th category.

[0032] Furthermore, S3 specifically includes:

[0033] S31. Code each scenario factor;

[0034] S32. Determine cluster data:

[0035]

[0036] In the formula, SSE represents the sum of squared errors, K represents the number of clusters, and C i represents the i-th family, and p represents C i The sample points in m i C i The center of mass.

[0037] S33. Measure the similarity of each scene:

[0038]

[0039] Where d(A, B) represents the Euclidean distance between point A and point B.

[0040] S34. Statistically analyze the clustering results and classify the scenes into 11 major categories.

[0041] Furthermore, S4 specifically includes:

[0042] S41, screening the scene difference layer;

[0043] S42, screening the driving willingness layer;

[0044] S43. Screen the indicator setting layer.

[0045] Furthermore, the specific process of S41 is as follows: evenly extract scenes from the 11 major categories obtained in S34, and the number of scenes to be extracted M is calculated as follows:

[0046]

[0047] Where m i is the number of scenes extracted from the i-th category, and M is the set number of extracted scenes.

[0048] Furthermore, the specific process of S42 is as follows:

[0049] The scene extraction ratio is determined based on the different types of drivers and the roads they travel on. The specific extraction ratio W is set as follows:

[0050]

[0051] Where A ijis the jth scenario factor under the i-th category, and R is the scenario chain

[0052] Furthermore, the specific process of S43 is as follows:

[0053] S431, based on the average death toll X and the variance S of the death toll in each scenario calculated in S21. 2 The risk level is divided into three levels: high risk, medium risk, and low risk. The set proportions of high risk, medium risk, and low risk are expressed as P 高危险度设定值 、P 中危险度设定值 、P 低危险度设定值 According to the complexity calculated in S22, the complexity is divided into three levels: high complexity, medium complexity, and low complexity. The set proportion values ​​of high complexity, medium complexity, and low complexity are respectively expressed as P 高复杂度设定值 、P 中复杂度设定值 、P 低复杂度设定值 .

[0054] S432: Randomly extract a specified number of training scenarios from the scenario library, calculate the complexity and risk level of each extraction result, and express the calculation result as P 高危险度结果 、P 中危险度结果 、P 低危险度结果 、P 高复杂度结果 、P 中复杂度结果 、P 低复杂度结果 .

[0055] S433. Compare the thresholds based on the set value of S431 and the calculated result value of S431. If the difference between the set value and the result value of each level does not exceed 5%, it can be output. Otherwise, continue to randomly draw and perform threshold judgment again until the threshold requirement is met.

[0056]

[0057] Compared with the prior art, the beneficial technical effects of the present invention are:

[0058] The present invention proposes a driver stress training scenario generation system driven by real data from massive highway accidents. A stress scenario library is established based on massive highway accident data. The number of scenarios is large and more in line with various driving scenarios in reality, with high reference value. The driver's customized "three-layer screening" extraction mode provides drivers with more targeted and rich stress scenarios for training, thereby effectively reducing the accident rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 stress scenario factor diagram;

[0060] Figure 2 Schematic diagram of the scene generation rule module structure;

[0061] Figure 3 Driver stress training scenario generation system interface;

[0062] Figure 4 Result diagram of driver stress training scenario generation. Specific embodiments

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0064] The present invention proposes a driver stress training scenario generation system based on highway accident data, the process is as follows:

[0065] S1. Preprocess the collected traffic accident data:

[0066] S11. Select the accident type, and select two types: vehicle-pedestrian accident and vehicle-to-vehicle accident;

[0067] S12. First, process the vehicle parameters and the vehicle motion state in combination with the turn signal state, as shown in Table 1. Then, generate the relative motion direction in combination with the high-risk driving behavior, as shown in Table 2.

[0068] Table 1 Motion state processing process

[0069]

[0070] Table 2 Generation of relative motion direction

[0071]

[0072]

[0073] S13. Reference Figure 1 ,The factors of driving stress scenarios are selected from accident data, mainly from four aspects: vehicle factors, traffic conflict party factors, road factors, and environmental factors. As shown in Table 3, vehicle factors include vehicle traffic mode and vehicle motion state; traffic conflict party factors include conflict party traffic mode, conflict party driving state, and relative motion state; road factors include road type, intersection section type, road surface condition and road surface condition; environmental factors include traffic signal mode, protective facility type, lighting conditions, weather, and visibility. Each dangerous scenario is generated by 14 specific scenario parameters, which can intuitively show the various aspects of the dangerous scenario.

[0074] Table 3 Driving stress scenario factors

[0075]

[0076] S2. Build corresponding indicators of the scene library:

[0077] S21. Construction of risk index:

[0078]

[0079]

[0080] Where, is the average number of deaths and S 2 is the variance of the number of deaths in the scene, X i is the number of deaths in the i-th accident, and N is the frequency of accidents involved in the scene.

[0081] S22. Organize and hierarchize environmental factors, road factors, vehicle factors, and traffic conflict factors to construct an orderly structural model and build a complexity index:

[0082]

[0083] In the formula, C represents the complexity, α i represents the complexity weight of the i-th category of the influencing factor classification, β ij represents the complexity weight of the jth parameter in the i-th category.

[0084] S3. Use clustering to classify the scene library:

[0085] S31, Reference Figure 1 , coding each scenario factor;

[0086] S32. Determine cluster data:

[0087]

[0088] In the formula, SSE represents the sum of squared errors, K represents the number of clusters, and C i represents the i-th family, and p represents C i The sample points in m i C i The center of mass.

[0089] S33. Measure the similarity of each scene:

[0090]

[0091] Where d(A, B) represents the Euclidean distance between point A and point B.

[0092] S34. Statistically analyze the clustering results and classify the scenes into 11 major categories.

[0093] S4. References Figure 2 , to establish the scene generation method model:

[0094] S41, screening of scene difference layers; evenly extracting scenes from the 11 major categories obtained in S34 to avoid excessive duplication of the extracted scenes. The number of scenes to be extracted, M, is calculated as follows:

[0095]

[0096] S42. Screening the driving intention layer: selectively increase the extraction ratio of certain scenes based on different types of drivers and different roads they frequently travel on. The specific extraction ratio W is set as follows:

[0097]

[0098] S43. Screening the indicator setting layer:

[0099] According to the average number of deaths X and the variance of the number of deaths in the scene S calculated in S21 2 The risk level is divided into three levels: high risk, medium risk, and low risk. The set proportions of high risk, medium risk, and low risk are expressed as P 高危险度设定值 、P 中危险度设定值 、P 低危险度设定值 According to the complexity calculated in S22, the complexity is divided into three levels: high complexity, medium complexity, and low complexity. The set proportion values ​​of high complexity, medium complexity, and low complexity are respectively expressed as P 高复杂度设定值 、P 中复杂度设定值 、P 低复杂度设定值 .

[0100] A specified number of training scenarios are randomly selected from the scenario library, and the complexity and risk level of each extraction result are calculated. The calculation result is expressed as P 高危险度结果 、P 中危险度结果 、P 低危险度结果 、P 高复杂度结果 、P 中复杂度结果 、P 低复杂度结果 .

[0101] According to the set value of each level proportion and the calculated result value, the threshold is compared. If the difference between the set value and the result value of each level proportion does not exceed 5%, it can be output. Otherwise, continue to randomly draw and perform threshold judgment again until the threshold requirement is met.

[0102]

[0103] Figure 3 This is a specific interface diagram of the driver stress training scenario generation system. The driver needs to input relevant parameters and click "Generate" to generate recommendations for training scenarios based on various factors customized by the driver.

[0104] refer to Figure 4 ,The driver inputs various parameters, the number of scenarios to be trained is set to 100, high risk accounts for 0.2, medium risk accounts for 0.6, low risk accounts for 0.2, high complexity accounts for 0.15, medium complexity accounts for 0.55, low complexity accounts for 0.3, and the road type of the intended scenario factors is urban road.

[0105] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention, and such changes and modifications fall within the scope of the invention as claimed.

Claims

1. A driver stress training scenario generation system based on highway accident data, characterized in that: The steps include: S1. Preprocessing the collected traffic accident data; S2. Build the corresponding indicators of the scenario library, including: S21. Construction of risk index: Where, is the average number of deaths and S 2 is the variance of the number of deaths in the scene, X i is the number of deaths in the i-th accident, and N is the frequency of accidents involved in the scene; S22. Organize and hierarchize environmental factors, road factors, vehicle factors, and traffic conflict factors to construct an orderly structural model and build a complexity index: In the formula, C represents the complexity, α i represents the complexity weight of the i-th category of the influencing factor classification, β ij represents the complexity weight of the jth parameter in the i-th category; S3, using clustering to classify the scene library; S4. Establish a scenario generation method model, specifically including: S41, screening of scene difference layers: evenly extract scenes from the classification obtained in S3, and the number of scenes to be extracted M is calculated as follows: Where m i is the number of scenes extracted from the i-th category, and M is the set number of extracted scenes; S42. Screening the driving intention layer: Based on the training intentions of different types of drivers, a higher extraction ratio W is set for the scene chains containing the intention scene factors. The specific extraction ratio W is set as follows: Where A ij is the jth scenario factor under the i-th category, and R is the scenario chain; S43. Screen the indicator setting layer.

2. A driver stress training scenario generation system based on highway accident data according to claim 1, characterized in that: S1 specifically includes: S11. Select the accident type, and select two types: vehicle-pedestrian accident and vehicle-to-vehicle accident; S12: First, process the vehicle parameters, process the vehicle motion state in combination with the turn signal state, and then generate the relative motion direction in combination with the high-risk driving behavior; S13. Select driving stress scenario factors.

3. The driver stress training scenario generation system based on highway accident data according to claim 2 is characterized in that: The processing process of the vehicle motion state in S12 is specifically as follows: Motion state processing The generation process of the relative motion direction is specifically shown in the following table: Generation of relative motion direction 4. The driver stress training scenario generation system based on highway accident data according to claim 3 is characterized in that: The driving stress scenario factors described in S13 are selected from four aspects: vehicle factors, traffic conflict factors, road factors, and environmental factors, as shown in the following table: Driving stress scenario factors 5. The driver stress training scenario generation system based on highway accident data according to claim 1 is characterized in that: S3 specifically includes: S31. Code each scenario factor; S32. Determine the number of clusters: In the formula, SSE represents the sum of squared errors, K represents the number of clusters, and C i represents the i-th family, and p represents C i The sample points in m i C i The center of mass; S33. Measure the similarity of each scene: Where d(A, B) represents the Euclidean distance between point A and point B; S34. Statistically analyze the clustering results and classify the scenes into 11 major categories.

6. The driver stress training scenario generation system based on highway accident data according to claim 1 is characterized in that: The specific process of S43 is as follows: S431. Average number of deaths calculated in S21 and the variance of the number of deaths in the scene S 2 The risk level is divided into three levels: high risk, medium risk, and low risk. The set proportions of high risk, medium risk, and low risk are expressed as P 高危险度设定值 、P 中危险度设定值 、P 低危险度设定值 According to the complexity calculated in S22, the complexity is divided into three levels: high complexity, medium complexity, and low complexity. The set proportion values ​​of high complexity, medium complexity, and low complexity are respectively expressed as P 高复杂度设定值 、P 中复杂度设定值 、P 低复杂度设定值 ; S432: Randomly extract a specified number of training scenarios from the scenario library, calculate the complexity and risk level of each extraction result, and express the calculation result as P 高危险度结果 、P 中危险度结果 、P 低危险度结果 、P 高复杂度结果 、P 中复杂度结果 、P 低复杂度结果 ; S433: Compare the thresholds based on the set value of S431 and the calculated value of S431. If the difference between the set value and the calculated value of each level does not exceed 5%, the output is performed. Otherwise, continue to randomly select and perform threshold judgment again until the threshold requirement is met.

Images