Method for determining the weights of autonomous driving hazard avoidance decisions that incorporate driving styles

Abstract

This invention discloses a method for determining the weights of autonomous driving risk avoidance decisions by integrating driving styles, belonging to the field of autonomous driving technology. The method includes the following steps: collecting driver subjective rating data, defining conflict scenarios and quantifying risk levels, clustering the data after normalization to obtain three driving styles, and constructing scenario-style units; through fuzzy matrix construction and defuzzification, AHP subjective weight calculation and consistency verification, factor correlation coefficient and conflict index calculation, and CRITIC objective weight calculation, integrating subjective and objective weights, and outputting a weight mapping table matching scenarios and driving styles to support personalized decision-making. This invention, employing the above-mentioned method for determining the weights of autonomous driving risk avoidance decisions by integrating driving styles, achieves personalized autonomous driving risk avoidance decisions, takes into account the risk tolerance capabilities of different traffic participants, and can flexibly adjust weights according to conflict scenarios and driving styles, improving the rationality and adaptability of decisions.

Description

Technical Field

[0001] This invention belongs to the field of autonomous driving technology, specifically relating to a method for determining the weights of autonomous driving hazard avoidance decisions that integrate driving styles. Background Technology

[0002] The automotive industry and intelligent driving technology are deeply integrated, and autonomous vehicles have become the mainstream development direction in the transportation field. The decision-making and planning module is the central hub of vehicle control, and the rationality of its trajectory directly determines driving safety and comfort.

[0003] However, existing technologies have the following shortcomings: insufficient integration of personalized driving styles, ignoring the differences in risk tolerance and decision-making habits among different drivers; insufficient consideration of the differences in vulnerability among traffic participants, especially neglecting the special needs of vulnerable groups such as pedestrians and non-motorized vehicles; and lack of dynamic adaptability, unable to flexibly adjust according to specific conflict scenarios.

[0004] Therefore, a new method is urgently needed. Summary of the Invention

[0005] The purpose of this invention is to provide a method for determining the risk avoidance decision weights of autonomous driving that integrates driving styles. This method enables personalized risk avoidance decisions for autonomous driving, takes into account the risk resistance capabilities of different traffic participants, and can flexibly adjust the weights according to conflict scenarios and driving styles, thereby improving the rationality and adaptability of the decisions.

[0006] To achieve the above objectives, this invention provides a method for determining the weights of autonomous driving hazard avoidance decisions that integrate driving styles, comprising the following steps:

[0007] S1. Collect drivers' subjective judgment scores on fairness factors, passenger safety factors, and vulnerable group safety factors through a combination of questionnaires and driving simulator tests. And the original scores of pairwise factor comparisons ;

[0008] Representing the The first valid sample The original scores of each influencing factor; Representing the The first valid sample The first influencing factor and the second The original scores of pairwise comparisons of each influencing factor; Represents factors of safety and balance. Represents factors related to the safety of drivers and passengers. Safety factors representing vulnerable groups;

[0009] S2. Define the conflict scenario as follows:

[0010] ;

[0011] In the formula, As the object of conflict, For conflict locations, To quantify the risk level based on conflict time, the formula for calculating conflict time is:

[0012] ;

[0013] In the formula, The relative distance between conflicting objects. The relative speed between the conflicting objects; For conflict time;

[0014] And based on Values ​​are used to classify risk levels;

[0015] S3. Based on the original subjective judgment scores collected in S1, the original scores are normalized and mapped to the [0,1] interval to obtain the normalized scores. The calculation formula is as follows:

[0016] ;

[0017] In the formula, For the first The first valid sample Normalized scores of each influencing factor; This is the original score; For the first The lowest original score of each factor in all valid samples; For the first The factor received the highest original score across all valid samples;

[0018] S4. Based on the normalized score obtained in S3, construct a three-dimensional feature vector and use a clustering algorithm to classify drivers into three driving styles: self-interested, neutral, and altruistic.

[0019] S5, S2-based conflict scenarios Combined with the driving style obtained from S4, an independent scene-style calculation unit is constructed;

[0020] S6. For each scene-style calculation unit built in S5, establish a weight allocation model that includes subjective weights, objective weights, and weight fusion.

[0021] S7. Based on the original comparison scores of the relative importance of each pair of factors collected in S1, calculate the triangular fuzzy number and construct the fuzzy judgment matrix;

[0022] S8. The centroid method is used to defuzzify the fuzzy judgment matrix constructed in S7 to obtain a clear positive reciprocal matrix. The calculation formula is as follows:

[0023] ;

[0024] In the formula, This represents the lower limit of the fuzzy comparison number. This is the median value of the fuzzy comparison number; This represents the upper limit of the fuzzy comparison number; For clear positive and negative reciprocal matrices;

[0025] S9. Perform column normalization, row sum calculation and row sum vector normalization operations on the clear positive and negative reciprocal matrix obtained in S8 to obtain the subjective weight vector of each influencing factor in the analytic hierarchy process.

[0026] S10. Calculate the largest eigenvalue of the clear positive reciprocal matrix in S8, and calculate the consistency index using the following formula:

[0027]

[0028] In the formula, To determine the largest eigenvalue of a matrix; Let be the order of the matrix;

[0029] The consistency ratio is obtained by combining the average random consistency index, and the formula is as follows:

[0030] ;

[0031] In the formula, As a consistency indicator; The average random consistency index; The consistency ratio;

[0032] when The decision matrix must satisfy the consistency requirement.

[0033] S11. Based on the normalized score obtained in S3, the correlation coefficient between any two influencing factors is calculated using the Pearson correlation coefficient formula.

[0034] S12. Based on the correlation coefficients obtained in S11, calculate the conflict quantification index for each influencing factor. The calculation formula is as follows:

[0035] ;

[0036] In the formula, The correlation coefficient between factors; For the first Quantitative indicators of the conflict of influencing factors;

[0037] S13. Based on the subjective weight vector obtained in S9 and the conflict quantification index obtained in S12, calculate the comprehensive information content of each influencing factor using the following formula:

[0038] ;

[0039] In the formula, For the first The comprehensive information content of each influencing factor; This is the subjective weight vector;

[0040] S14. Based on the comprehensive information obtained in S13, calculate the objective weight vector of each influencing factor using the following formula:

[0041] ;

[0042] In the formula, For the first The objective weight of each influencing factor; The total number of influencing factors ;

[0043] S15. Based on the subjective weight vector of S9 and the objective weight vector of S14, a weighted average method is used to fuse them to obtain the comprehensive weight vector.

[0044] ;

[0045] In the formula, For the first The combined weight of each influencing factor; The fusion coefficient;

[0046] S16. Match the comprehensive weight vectors corresponding to the three driving styles in S4 to the conflict scenario in S2, and generate a weight mapping table for real-time decision-making calls of autonomous vehicles.

[0047] Preferably, in S2, the conflicting object ,in, Conflicts between autonomous vehicles and non-motorized vehicles. Conflicts between autonomous vehicles and passenger cars Conflicts between autonomous vehicles and commercial vehicles;

[0048] Conflict location ,in, For intersection conflict, Due to road section conflicts;

[0049] The risk level classification criteria are as follows:

[0050] Low risk: TTC > 5.0s;

[0051] Medium risk: 2.5s <TTC≤5.0s;

[0052] High risk: 1.5s <TTC≤2.5s;

[0053] Extremely high risk: TTC≤1.5s.

[0054] Preferably, in S4, the clustering algorithm used is K-means, the distance metric is Euclidean distance, and the cluster centers are iteratively updated until the position change is less than a preset threshold or the number of iterations reaches the maximum value.

[0055] Preferably, in S6, the weight allocation model formula is:

[0056] ;

[0057] In the formula, To represent autonomous vehicles in In the scene The comprehensive decision-making evaluation index for driving style-based risk avoidance paths has a value range of [0,1]. To represent factors of fairness; To represent factors related to the safety of drivers and passengers; As a representative of safety factors for vulnerable groups; , , For respectively , , The weighting coefficients.

[0058] Preferably, in S7, the triangular fuzzy number is calculated using the following formula:

[0059] ;

[0060] In the formula, ,2,3, Representing fairness factors, Represents factors related to the safety of drivers and passengers. Safety factors representing vulnerable groups; For the other two categories of factors;

[0061] Representative factors Relative factors The importance of, and satisfying:

[0062] ;

[0063] ;

[0064] ;

[0065] In the formula, The fuzzy interval adjustment coefficient is calculated using the following formula:

[0066] ;

[0067] In the formula, The maximum risk coefficient among all known conflict scenarios; 1.0 is the basic fuzzy width; 0.5 is the risk impact gain coefficient. The risk coefficient is calculated using the following formula:

[0068] ;

[0069] Fuzzy judgment matrix , represented as:

[0070] ;

[0071] In the formula, all diagonal elements are (1,1,1), indicating that each factor is of equal importance to itself. For the first The first influencing factor and the second Fuzzy comparison number of influencing factors.

[0072] Preferably, S9 is as follows:

[0073] Traversing the clear matrix For each column, calculate the sum of all elements in that column using the following formula:

[0074] ;

[0075] In the formula, For the clear positive and negative reciprocal matrix The Middle The sum of all elements in the column;

[0076] based on For the matrix Perform column normalization to obtain the normalized matrix. The calculation formula is:

[0077] ;

[0078] In the formula, The new matrix obtained after column normalization The Middle line, number Column elements;

[0079] For normalized matrix Summing each row yields a row sum vector, using the formula:

[0080] ;

[0081] ;

[0082] In the formula, For column normalized matrix The Middle The sum of all elements in the row; For rows and vectors; Let be the order of the matrix;

[0083] For rows and vectors After normalization, the subjective weight vector is finally obtained. The formula is:

[0084] ;

[0085] In the formula, For the corresponding number The row elements of each influencing factor; The number of influencing factors; For the first The final subjective weight of each influencing factor.

[0086] Preferably, in S11, the correlation coefficient between any two influencing factors is calculated. for:

[0087] ;

[0088] In the formula, For the first The number of valid samples in the th... Normalized scores for each influencing factor; For the first The sample mean of the normalized scores of all valid samples under each influencing factor; For the first The number of valid samples in the th... Normalized scores for each influencing factor; For the first The sample mean of the normalized scores of all valid samples under each influencing factor; The total number of valid samples; The correlation coefficient between factors; For the first Sample standard deviation of normalized scores for each influencing factor; For the first The sample standard deviation of the normalized scores of each influencing factor.

[0089] Preferably, in S15, the fusion coefficient is calculated based on the driving style coefficient and the scenario risk coefficient, using the following formula:

[0090] ;

[0091] In the formula, The original fusion coefficient; Style coefficient; This represents the scenario risk coefficient.

[0092] Therefore, the present invention adopts the above-mentioned method for determining the weights of autonomous driving avoidance decisions that integrate driving styles. Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0093] (1) The technical means of using K-means clustering algorithm to divide driving styles into three categories: self-interested, neutral and altruistic, overcomes the technical problem of insufficient integration of driving styles, and thus achieves the technical effect of personalized decision-making.

[0094] (2) By adopting technical means that the safety factors of vulnerable groups are listed as core decision-making factors, the technical problem of ignoring the needs of vulnerable groups is overcome, thereby achieving the technical effect of taking into account the risk resistance capabilities of different traffic participants.

[0095] (3) By using the construction of scene-style coupled calculation unit and SSAHW hybrid weight method to dynamically integrate subjective and objective weights, the technical problem of fixed decision weights is overcome, thereby achieving the technical effect of flexibly adjusting weights according to conflict scenario (object, location, risk level) and driving style.

[0096] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0097] Figure 1 This is a flowchart of an embodiment of the autonomous driving hazard avoidance decision weight determination method integrating driving style of the present invention;

[0098] Figure 2 This is a three-dimensional scatter plot of the KMeans clustering results of an embodiment of the autonomous driving hazard avoidance decision weight determination method integrating driving style of the present invention.

[0099] Figure 3 This is a cluster center radar diagram of an embodiment of the autonomous driving hazard avoidance decision weight determination method integrating driving style of the present invention. Detailed Implementation

[0100] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used in the present invention should have the ordinary meaning understood by those skilled in the art.

[0101] Example 1

[0102] like Figures 1-3 As shown, this embodiment provides a method for determining the weights of autonomous driving avoidance decisions that integrate driving styles. It should be understood that the specific parameters, models, and protocols mentioned in this embodiment are merely examples to help those skilled in the art understand the present invention, and are not intended to limit the present invention.

[0103] The method for determining the risk avoidance decision weights of autonomous driving systems that integrates driving styles, according to the present invention, includes the following steps:

[0104] S1. Through questionnaires and driving simulator tests, quantitatively evaluate three factors: fairness, driver and passenger safety, and safety of vulnerable groups, and obtain the original single-factor scores of valid samples. Original scores compared with pairwise factors ;

[0105] Representing the The first valid sample The original scores of each influencing factor, for example: The original scores representing the safety factors of the third valid sample of drivers and passengers; Representing the The first valid sample The first influencing factor and the second The original scores of pairwise comparisons of each influencing factor ( Represents factors of safety and balance. Represents factors related to the safety of drivers and passengers. (Representing safety factors for vulnerable groups);

[0106] S2. Generate a list of scenario variables to address the pre-set conflict scenarios between the questionnaire design in S1 and the driving simulator experiment. Where A represents the conflicting object, B represents the conflicting location, and C represents the scenario risk level;

[0107] Conflict Object ,in, Conflicts between autonomous vehicles and non-motorized vehicles. Conflicts between autonomous vehicles and passenger cars Conflicts between autonomous vehicles and commercial vehicles;

[0108] Conflict location ,in, For intersection conflicts (including crossroads, T-junctions, and roundabouts). For road segment conflicts (including urban main roads, secondary roads, expressways, and rural roads);

[0109] Scenario Risk Level Assessing based on Time of Conflict (TTC), the calculation formula is as follows:

[0110] ;

[0111] In the formula, The relative distance between conflicting objects is measured using lidar with an accuracy of ±0.1m; The relative velocity between conflicting objects is measured using millimeter-wave radar with an accuracy of ±0.5 km / h.

[0112] Risk levels are categorized according to TTC values, as shown in the table below:

[0113] Table 1 Risk Level Classification Table

[0114]

[0115] The risk coefficient is expressed as:

[0116] ;

[0117] In the formula, Risk coefficient;

[0118] S3. Raw data for quantitative scoring of the three types of influencing factors obtained from S1. Outlier handling is performed using 3 Principle of exclusion The ratings, among which, The sample mean. Standard deviation;

[0119] Based on the quantitative scoring data of three influencing factors—fairness, passenger safety, and safety of vulnerable groups—obtained from S1, the original scores are normalized and mapped to the [0,1] interval to eliminate the impact of differences in the units of measurement between different indicators on subsequent weight calculations. The normalization calculation formula is as follows:

[0120] ;

[0121] In the formula, For the first The first valid sample Normalized scores of each influencing factor; This is the original score; For the first The lowest original score of each factor in all valid samples; For the first The factor received the highest original score across all valid samples;

[0122] S4. Driving style clustering analysis is based on the normalized scores of each sample obtained in S3 on three influencing factors: fairness, driver and passenger safety, and safety of vulnerable groups. The scores of each valid sample on these three influencing factors are constructed into a three-dimensional feature vector, represented as follows:

[0123] ;

[0124] in, To normalize the score for fairness; A normalized score for driver and passenger safety; Normalized safety scores for vulnerable groups;

[0125] The feature vectors of all samples together constitute a three-dimensional feature dataset. ,in, M The effective sample size;

[0126] K-means clustering algorithm is used to analyze the 3D feature dataset. Cluster analysis was performed, and the number of clusters was preset based on the three common driving behavior styles: selfish, altruistic, and neutral. The elbow rule was used to verify the data against actual data; the Euclidean distance formula was used to measure the similarity between samples.

[0127] Three significantly different samples are randomly selected as initial centers. The iteration stops when the number of iterations is ≥100 or the change in the cluster center position is ≤0.001 and the cluster assignment result is stable. Finally, three clusters are obtained, and the final cluster center of each cluster is output.

[0128] After clustering, based on the score distribution characteristics of each cluster center in three dimensions, the three clusters are defined and labeled as three types of driving styles: the highest normalized average score of safety factors for drivers and passengers is the self-interested type, the highest normalized average score of fairness factors is the neutral type, and the highest normalized average score of safety factors for vulnerable groups is the altruistic type.

[0129] S5, the core of constructing the scene-style calculation unit, is to process each conflict scene defined in S2. The three driving styles—selfish, neutral, and altruistic—defined by S4 are combined to form multiple independent calculation units.

[0130] S6. To quantify the quality of autonomous vehicles' hazard avoidance paths under specific scenarios and driving styles, a weight allocation model is established, with the following formula:

[0131] ;

[0132] In the formula, To represent autonomous vehicles in In the scene The comprehensive decision-making evaluation index for risk avoidance paths based on driving style ranges from [0,1]. The closer the value is to 1, the more the risk avoidance path meets the decision-making needs of the scenario and driving style. To represent fairness, the core is to balance the safety rights of drivers and passengers inside the vehicle with those of vulnerable road users outside the vehicle, and to avoid the concentration of risk in one group. To represent the safety factors of drivers and passengers, with the core objective of ensuring the personal safety and riding experience of the people in the vehicle, the risks brought about by aggressive control strategies such as sudden braking and sharp steering are avoided; To represent the safety factors of vulnerable groups, the focus is on traffic participants with weaker risk resistance capabilities, such as pedestrians and non-motorized vehicle users, and their safety is given priority. , , For respectively , , The weighting coefficients are used to reflect the relative importance of the three types of factors in decision-making;

[0133] S7. Based on the "scene-style" coupled calculation unit divided by S5 and the corresponding unitized questionnaire data of the unit; construct a fuzzy judgment matrix for each unit to ensure the unitized adaptability of the weight calculation;

[0134] Introducing triangular fuzzy numbers to represent the elements of the judgment matrix, the elements are represented as follows:

[0135] ;

[0136] In the formula, ,2,3, Representing fairness factors, Represents factors related to the safety of drivers and passengers. Safety factors representing vulnerable groups; This represents the lower limit of the fuzzy comparison number. This is the median value of the fuzzy comparison number; This represents the upper limit of the fuzzy comparison number; For the other two categories of factors, such as hour, ;

[0137] Representative factors Relative factors The importance of, and satisfying:

[0138] ;

[0139] ;

[0140] ;

[0141] In the formula, This is the fuzzy interval adjustment coefficient, used to control the width of the triangular fuzzy number. The calculation formula is:

[0142] ;

[0143] In the formula, The maximum risk coefficient among all known conflict scenarios; 1.0 is the basic fuzzy width; 0.5 is the risk impact gain coefficient.

[0144] Fuzzy judgment matrix It is 3×3 in dimension, represented as:

[0145] ;

[0146] In the formula, all diagonal elements are (1,1,1), indicating that each factor is of equal importance to itself. For the first The first influencing factor and the second Fuzzy comparison number of influencing factors;

[0147] S8. Judgment fuzzy matrix based on the output of S7 The triangular fuzzy number centroid method is used to deblur all elements of the matrix, converting the fuzzy elements into clear values. The calculation formula is as follows:

[0148] ;

[0149] In the formula, For deblurring values;

[0150] S9. Clear positive-negative judgment matrix based on the output of S8 (3×3 dimensions, corresponding to three categories of factors: fairness, safety of drivers and passengers, and safety of vulnerable groups), the subjective weight vector of each influencing factor is calculated using the analytic hierarchy process (AHP), which reflects the driver's subjective value orientation.

[0151] Traversing the clear matrix For each column, calculate the sum of all elements in that column using the following formula:

[0152] ;

[0153] In the formula, For the clear positive and negative reciprocal matrix The Middle The sum of all elements in the column;

[0154] based on For the matrix Perform column normalization to obtain the normalized matrix. The calculation formula is:

[0155] ;

[0156] In the formula, The new matrix obtained after column normalization The Middle line, number Column elements;

[0157] For normalized matrix Summing each row yields a row sum vector, using the formula:

[0158] ;

[0159] ;

[0160] In the formula, For column normalized matrix The Middle The sum of all elements in the row; For rows and vectors; Let be the order of the matrix, in this embodiment... There are three influencing factors.

[0161] For rows and vectors After normalization, the subjective weight vector based on AHP is finally obtained. The formula is:

[0162] ;

[0163] In the formula, For the corresponding number The row elements of each influencing factor; In this embodiment, the number of influencing factors is [not specified]. ; For the first The final AHP subjective weights of each influencing factor;

[0164] S10. Perform a standard AHP consistency check, including the following sub-steps:

[0165] S1001, By solving the characteristic equation:

[0166] ;

[0167] In the formula, To determine the matrix, It is the identity matrix; The determinant operator is used to calculate the determinant value of the matrix within the parentheses. Let be the eigenvalues ​​to be solved;

[0168] In practical applications, numerical algorithms such as the power method can be used, or mathematical computing libraries such as MATLAB's eig function can be directly called to solve the problem;

[0169] S1002, Calculate the consistency index The formula is:

[0170] ;

[0171] In the formula, To determine the largest eigenvalue of a matrix;

[0172] S1003, Query the average random consistency index Refer to the table to obtain the matrix order. corresponding value;

[0173] Calculate the consistency ratio The formula is:

[0174] ;

[0175] S1004. If CR < 0.10, the judgment matrix passes the consistency test; if CR ≥ 0.10, the elements in the judgment matrix need to be corrected, and S1001-S1003 need to be repeated until CR < 0.10, which satisfies the consistency requirement.

[0176] S11. Calculate the correlation coefficient between any two influencing factors using normalized scores, as shown in the following formula:

[0177] ;

[0178] In the formula, For the first The number of valid samples in the th... Normalized scores for each influencing factor; For the first The sample mean of the normalized scores of all valid samples under each influencing factor; For the first The number of valid samples in the th... Normalized scores for each influencing factor; For the first The sample mean of the normalized scores of all valid samples under each influencing factor; The total number of valid samples; The correlation coefficient between factors; For the first Sample standard deviation of normalized scores for each influencing factor; For the first Sample standard deviation of normalized scores for each influencing factor;

[0179] S12. Based on the correlation coefficients among factors obtained in S11, calculate the conflict quantification index for each influencing factor using the following formula:

[0180] ;

[0181] In the formula, For the first The conflict quantification index of each influencing factor; the larger the value, the higher the degree of conflict between that factor and other factors.

[0182] S13. Based on the subjective weights obtained in S9 and the conflict quantification index obtained in S12, calculate the comprehensive information content of each influencing factor using the following formula:

[0183] ;

[0184] In the formula, For the first The larger the value of the comprehensive information content of an influencing factor, the more comprehensive information the factor contains, the higher its relative importance, and the greater the weight it should be assigned.

[0185] S14. Based on the comprehensive information of each influencing factor obtained in S13, the objective weight vector is calculated through normalization, using the following formula:

[0186] ;

[0187] In the formula, For the first CRITIC objective weights of each influencing factor; The total number of influencing factors. ;

[0188] S15. By fusing the AHP subjective weight vector from S9 with the CRITIC objective weight vector from S14 using a weighted average method, the comprehensive weight vector corresponding to the "Scene-Style" unit is obtained. Specifically:

[0189] The formula for calculating the overall weight is:

[0190] ;

[0191] In the formula, For the first The combined weight of each influencing factor; The fusion coefficient is... ;

[0192] The method for determining the fusion coefficient is as follows:

[0193] A more self-centered driving style leads to decisions that rely more on subjective preferences; a more altruistic style leads to decisions that closely resemble objective statistical patterns. Style coefficients are set based on driving style clustering results. , self-interested type Neutral type altruism 2;

[0194] The original fusion coefficient is calculated based on the combined scenario risk level and driving style, using the following formula:

[0195] ;

[0196] In the formula, The original fusion coefficient;

[0197] To ensure that both subjective and objective weights play their roles, restrictions are imposed. The formula is given within the range [0.2, 0.8]:

[0198] ;

[0199] S16. The final output weight mapping table is shown in the table below:

[0200] Table 2 Weight Mapping Table

[0201]

[0202] Generate a complete weight mapping table for each conflict scenario defined in S2. The comprehensive weight vectors corresponding to the three driving styles—selfish, neutral, and altruistic—are matched respectively, and are expressed as follows:

[0203] ;

[0204] In the formula, This is the overall weight vector;

[0205] The generated weight mapping table is pre-installed in the control system or cloud decision-making platform of the autonomous vehicle; during vehicle operation, the environmental perception module continuously collects driving data and accurately identifies the current conflict scenario by analyzing the data. The system clarifies the conflicting parties (non-motorized vehicles / passenger vehicles / commercial vehicles), the conflict location (intersections / road sections), and the risk level (low / medium / high / extremely high) based on TTC quantification. The driver profiling module calls upon the driver's historical driving data or combines it with real-time driving behavior analysis to determine the current driver's driving style (selfish / neutral / altruistic).

[0206] Based on the identified conflict scenarios and determined driving styles, a matching query is performed in a pre-deployed weight mapping table to quickly obtain the corresponding comprehensive weight vector. The planning and control module uses this comprehensive weight vector as a basis to integrate three decision objectives: fairness, safety of drivers and passengers, and safety of vulnerable groups, and finally generates safe, balanced vehicle control commands that are tailored to the driver's style.

[0207] Therefore, the present invention adopts the above-mentioned method for determining the weight of autonomous driving risk avoidance decision based on integrated driving style. This method realizes personalized autonomous driving risk avoidance decision, takes into account the risk resistance capabilities of different traffic participants, and can flexibly adjust the weight according to conflict scenarios and driving styles, thereby improving the rationality and adaptability of decision-making.

[0208] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0209] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for determining the weights of autonomous driving hazard avoidance decisions that integrate driving styles, characterized in that, Includes the following steps: S1. Collect drivers' subjective judgment scores on fairness factors, passenger safety factors, and vulnerable group safety factors through a combination of questionnaires and driving simulator tests. And the original scores of pairwise factor comparisons ; Representing the The first valid sample The original scores of each influencing factor; Representing the The first valid sample The first influencing factor and the second The original scores of pairwise comparisons of each influencing factor; Represents factors of safety and balance. Represents factors related to the safety of drivers and passengers. Safety factors representing vulnerable groups; S2. Define the conflict scenario as follows: ； In the formula, As the object of conflict, For conflict locations, To quantify the risk level based on conflict time, the formula for calculating conflict time is: ； In the formula, The relative distance between conflicting objects. The relative speed between the conflicting objects; For conflict time; And based on Values ​​are used to classify risk levels; S3. Based on the original subjective judgment scores collected in S1, the original scores are normalized and mapped to the [0,1] interval to obtain the normalized scores. The calculation formula is as follows: ； In the formula, For the first The first valid sample Normalized scores of each influencing factor; This is the original score; For the first The lowest original score of each factor in all valid samples; For the first The factor received the highest original score across all valid samples; S4. Based on the normalized score obtained in S3, construct a three-dimensional feature vector and use a clustering algorithm to classify drivers into three driving styles: self-interested, neutral, and altruistic. S5, S2-based conflict scenarios Combined with the driving style obtained from S4, an independent scene-style calculation unit is constructed; S6. For each scene-style calculation unit built in S5, establish a weight allocation model that includes subjective weights, objective weights, and weight fusion. S7. Based on the original comparison scores of the relative importance of each pair of factors collected in S1, calculate the triangular fuzzy number and construct the fuzzy judgment matrix; S8. The centroid method is used to defuzzify the fuzzy judgment matrix constructed in S7 to obtain a clear positive reciprocal matrix. The calculation formula is as follows: ； In the formula, This represents the lower limit of the fuzzy comparison number. This is the median value of the fuzzy comparison number; This represents the upper limit of the fuzzy comparison number; For the elements of a clear positive and negative matrix; S9. Perform column normalization, row sum calculation and row sum vector normalization operations on the clear positive and negative reciprocal matrix obtained in S8 to obtain the subjective weight vector of each influencing factor in the analytic hierarchy process. S10. Calculate the largest eigenvalue of the clear positive reciprocal matrix in S8, and calculate the consistency index using the following formula: In the formula, To determine the largest eigenvalue of a matrix; Let be the order of the matrix; The consistency ratio is obtained by combining the average random consistency index, and the formula is as follows: ； In the formula, As a consistency indicator; The average random consistency index; Consistency ratio; when The decision matrix must satisfy the consistency requirement. S11. Based on the normalized score obtained in S3, the correlation coefficient between any two influencing factors is calculated using the Pearson correlation coefficient formula. S12. Based on the correlation coefficients obtained in S11, calculate the conflict quantification index for each influencing factor. The calculation formula is as follows: ； In the formula, The correlation coefficient between factors; For the first Quantitative indicators of the conflict of influencing factors; S13. Based on the subjective weight vector obtained in S9 and the conflict quantification index obtained in S12, calculate the comprehensive information content of each influencing factor using the following formula: ； In the formula, For the first The comprehensive information content of each influencing factor; This is the subjective weight vector; S14. Based on the comprehensive information obtained in S13, calculate the objective weight vector of each influencing factor using the following formula: ； In the formula, For the first The objective weight of each influencing factor; The total number of influencing factors ; S15. Based on the subjective weight vector of S9 and the objective weight vector of S14, a weighted average method is used to fuse them to obtain the comprehensive weight vector. ； In the formula, For the first The combined weight of each influencing factor; The fusion coefficient; S16. Match the comprehensive weight vectors corresponding to the three driving styles in S4 to the conflict scenario in S2, and generate a weight mapping table for real-time decision-making calls of autonomous vehicles.

2. The method for determining the weights of autonomous driving risk avoidance decisions that integrate driving styles according to claim 1, characterized in that, In S2, conflicting objects ,in, Conflicts between autonomous vehicles and non-motorized vehicles. Conflicts between autonomous vehicles and passenger cars Conflicts between autonomous vehicles and commercial vehicles; Conflict location ,in, For intersection conflict, Due to road section conflicts; The risk level classification criteria are as follows: Low risk: TTC > 5.0s; Medium risk: 2.5s <TTC≤5.0s; High risk: 1.5s <TTC≤2.5s; Extremely high risk: TTC≤1.5s.

3. The method for determining the weights of autonomous driving risk avoidance decisions that integrate driving styles according to claim 1, characterized in that, In S4, the clustering algorithm used is K-means, the distance metric is Euclidean distance, and the cluster centers are iteratively updated until the position change is less than a preset threshold or the number of iterations reaches the maximum value.

4. The method for determining the weights of autonomous driving risk avoidance decisions that integrate driving styles according to claim 1, characterized in that, In S6, the weight allocation model formula is: ； In the formula, To represent autonomous vehicles in In the scene The comprehensive decision-making evaluation index for driving style-based risk avoidance paths has a value range of [0,1]. To represent factors of fairness; To represent factors related to the safety of drivers and passengers; As a representative of safety factors for vulnerable groups; , , For respectively , , The weighting coefficients.

5. The method for determining the weights of autonomous driving risk avoidance decisions that integrate driving styles according to claim 1, characterized in that, In S7, the formula for calculating the triangular fuzzy number is: ； In the formula, ,2,3, Representing fairness factors, Represents factors related to the safety of drivers and passengers. Safety factors representing vulnerable groups; For the other two categories of factors; Representative factors Relative factors The importance of, and satisfying: ； ； ； In the formula, The fuzzy interval adjustment coefficient is calculated using the following formula: ； In the formula, This represents the highest risk coefficient among all known conflict scenarios. 1.0 is the base blur width; 0.5 represents the risk impact gain coefficient; The risk coefficient is calculated using the following formula: ； Fuzzy judgment matrix , represented as: ； In the formula, all diagonal elements are (1,1,1), indicating that each factor is of equal importance to itself. For the first The first influencing factor and the second Fuzzy comparison number of influencing factors.

6. The method for determining the risk avoidance decision weights of autonomous driving systems that integrate driving styles according to claim 1, characterized in that, S9 specifically refers to: Traversing a clear positive and negative matrix For each column, calculate the sum of all elements in that column using the following formula: ； In the formula, For clear positive and negative matrices The Middle The sum of all elements in the column; based on For the matrix Perform column normalization to obtain the normalized matrix. The calculation formula is: ； In the formula, The new matrix obtained after column normalization The Middle line, number Column elements; For normalized matrix Summing each row yields a row sum vector, using the formula: ； ； In the formula, For column normalized matrix The Middle The sum of all elements in the row; For rows and vectors; Let be the order of the matrix; For rows and vectors After normalization, the subjective weight vector is finally obtained. The formula is: ； In the formula, For the corresponding number The row elements of each influencing factor; The number of influencing factors; For the first The final subjective weight of each influencing factor.

7. The method for determining the risk avoidance decision weights of autonomous driving systems that integrate driving styles according to claim 1, characterized in that, In S11, calculate the correlation coefficient between any two influencing factors. for: ； In the formula, For the first The number of valid samples in the th... Normalized scores for each influencing factor; For the first The sample mean of the normalized scores of all valid samples under each influencing factor; For the first The number of valid samples in the th... Normalized scores for each influencing factor; For the first The sample mean of the normalized scores of all valid samples under each influencing factor; The total number of valid samples; The correlation coefficient between factors; For the first Sample standard deviation of normalized scores for each influencing factor; For the first The sample standard deviation of the normalized scores of each influencing factor.

8. The method for determining the weights of autonomous driving risk avoidance decisions that integrate driving styles according to claim 1, characterized in that, S15, the fusion coefficient is calculated based on the driving style coefficient and the scenario risk coefficient, and the formula is: ； In the formula, The original fusion coefficient; Style coefficient; This represents the scenario risk coefficient.

9. A computer device, characterized in that, include: A processor configured to be coupled to memory, read and execute instructions and / or program code in the memory to perform the method as described in any one of claims 1-8.

10. A computer-readable medium, characterized in that, The computer-readable medium stores computer program code that, when executed on a computer, causes the computer to perform the method as described in any one of claims 1-8.

Images