Method and System for Analyzing the Safety of an Autonomous Driving System Based on Scenario Parameters

Abstract

The present application provides a method and system for analyzing the safety of an autonomous driving system based on scenario parameters, relating to the field of artificial intelligence technology. The method includes: obtaining scenario parameter data; based on the scenario parameter data, obtaining parameter vector instances and the feature quantities of the autonomous driving system under the parameter vector instances; training a linear model to obtain a target model; the target model determines the parameter vector instances and feature quantities with higher rankings according to the constraint conditions to obtain target parameters; combining any two of the target parameters to obtain a target parameter combination; dividing the scenario parameter space into parameter sub-spaces; based on the parameter sub-spaces, constructing an objective function according to the target model and the scenario parameter safety threshold; generating a safety analysis result according to the objective function, so as to solve the problem that the current safety analysis method of the autonomous driving system cannot perform safety analysis on the autonomous driving system based on key parameters, resulting in a low credibility of the analysis result.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular, to a method and system for analyzing the safety of an autonomous driving system based on scenario parameters. Background Art

[0002] With the continuous development of artificial intelligence theories and methods, autonomous driving technology has gradually matured and been applied. Since the behavior of an autonomous driving system is closely related to the scenarios in which it operates, that is, different parameters of the scenario will cause the autonomous driving system to exhibit different behaviors therein.

[0003] To ensure the safety of an autonomous driving system during use, currently, safety analysis of the autonomous driving system is carried out based on simulation scenario parameters, aiming to evaluate the safety of the autonomous driving system within the scenario parameter space by simulating the operation of the autonomous driving system. By obtaining scenario parameters such as rainfall, light intensity, road width, etc. in the vehicle's automatic scenario, the autonomous driving system is run in a simulator for simulation, so as to analyze the safety of the autonomous driving system in this scenario parameter space.

[0004] However, the currently adopted method for analyzing the safety of an autonomous driving system conducts unified analysis based on all scenario parameters, and it is impossible to obtain key parameters (i.e., the parameters that have the greatest impact on safety) from numerous scenario parameters. Since the influence degree of each scenario parameter on the autonomous driving system is different, it is impossible to conduct safety analysis of the autonomous driving system based on key parameters, resulting in a relatively low credibility of the analysis results. Summary of the Invention

[0005] This application provides a method and system for analyzing the safety of an autonomous driving system based on scenario parameters to solve the technical problem that the existing method for analyzing the safety of an autonomous driving system cannot conduct safety analysis of the autonomous driving system based on key parameters, resulting in a relatively low credibility of the analysis results.

[0006] The first aspect of this application provides a method for analyzing the safety of an autonomous driving system based on scenario parameters, which is applied to the autonomous driving system of a vehicle; it includes:

[0007] Obtain scenario parameter data; the scenario parameter data includes: a scenario parameter vector, a scenario parameter space, and a scenario parameter safety threshold; the scenario parameter vector is composed of several scenario parameters when the vehicle is in an autonomous driving scenario, and the scenario parameter space is a set of the scenario parameter ranges;

[0008] Based on the scenario parameter data, obtain a parameter vector instance and the characteristic quantity of the autonomous driving system under the parameter vector instance; the parameter vector instance includes: the scenario parameter vector and the corresponding scenario parameter range;

[0009] Train a linear model using the parameter vector instance and the feature quantity to obtain a target model;

[0010] Input the constraint conditions into the target model, and the target model sorts the parameter vector instance and the feature quantity in descending order according to the Shapley value according to the constraint conditions, and determines the first preset number of the parameter vector instances and the feature quantity with the top ranking to obtain target parameters;

[0011] Combine any two parameters in the target parameters to obtain a target parameter combination;

[0012] Divide the scenario parameter space into a second preset number of parameter subspaces according to the target parameter combination;

[0013] Based on the parameter subspace, construct an objective function according to the target model and the scenario parameter safety threshold;

[0014] Generate a security analysis result according to the objective function.

[0015] In some embodiments, the step of obtaining a parameter vector instance and the feature quantity of the autonomous driving system under the parameter vector instance based on the scenario parameter data includes:

[0016] Use the uniform distribution method to obtain a third preset number of parameter vector instances in the scenario parameter space;

[0017] Input the parameter vector instance into a simulator to construct a scenario instance corresponding to the parameter vector instance;

[0018] Use the simulator to run the autonomous driving system, and perform simulation training on the autonomous driving system under the scenario instance to obtain the feature quantity of the autonomous driving system.

[0019] In some embodiments, the step of sorting the parameter vector instance and the feature quantity in descending order according to the Shapley value according to the constraint conditions, and determining the first preset number of the parameter vector instances and the feature quantity with the top ranking to obtain target parameters includes:

[0020] According to the constraint conditions, use the SHAP method to calculate the Shapley values of the parameter vector instance and the feature quantity;

[0021] Sort the Shapley values in descending order to determine the first preset number of target Shapley values with the top ranking;

[0022] Determine the parameter vector instance and the feature quantity corresponding to the target Shapley value to obtain target parameters.

[0023] In some embodiments, the target parameter combination includes: a first target parameter and a second target parameter; the step of dividing the scenario parameter space into a second preset number of parameter subspaces according to the target parameter combination includes:

[0024] Determine the scenario parameter range of the target parameter combination in the scenario parameter space;

[0025] Based on the scenario parameter range, divide the first target parameter and the second target parameter according to a fourth preset number to obtain a first parameter subspace and a second parameter subspace;

[0026] Combine the first parameter subspace and the second parameter subspace to obtain a second preset number of parameter subspaces; the second preset number is:

[0027] N = L×L;

[0028] where L is the fourth preset number.

[0029] In some embodiments, the target model is configured to:

[0030] Fit the parameter vector instance and the feature quantity to generate a target feature quantity.

[0031] In some embodiments, the objective function is:

[0032] ;

[0033] where δ ij is the quantization security index of the i , j th parameter subspace; T is the scenario parameter safety threshold; r ij represents the i , j th parameter subspace; f ( X ) is the target feature quantity.

[0034] In some embodiments, the step of generating a security analysis result according to the objective function includes:

[0035] According to the objective function, use the mixed integer linear programming method to calculate the quantization security index of the parameter subspace;

[0036] Generate a security analysis result according to the quantization security index.

[0037] In some embodiments, the step of generating a security analysis result according to the quantified security metric includes:

[0038] If the quantified security metric is less than or equal to 0, generate a first piece of information;

[0039] If the quantified security metric is greater than 0, generate a second piece of information;

[0040] Wherein, the first piece of information represents that the autonomous driving system is safe, and the second piece of information represents that the autonomous driving system is unsafe.

[0041] In some embodiments, the method further includes:

[0042] Generate a visualization result based on the security analysis result; the visualization result is a heat map; the heat map is configured with a second preset number of squares for displaying the first piece of information or the second piece of information.

[0043] A second aspect of the present application provides a security analysis system for an autonomous driving system based on scenario parameters, including:

[0044] An acquisition module, configured to:

[0045] Acquire scenario parameter data; the scenario parameter data includes: a scenario parameter vector, a scenario parameter space, and a scenario parameter safety threshold; the scenario parameter vector is composed of several scenario parameters when the vehicle is in an autonomous driving scenario, and the scenario parameter space is a set of the scenario parameter ranges;

[0046] Based on the scenario parameter data, acquire a parameter vector instance and a feature quantity of the autonomous driving system under the parameter vector instance; the parameter vector instance includes: the scenario parameter vector and the corresponding scenario parameter range;

[0047] A training module, configured to:

[0048] Train a linear model using the parameter vector instance and the feature quantity to obtain a target model;

[0049] An analysis result determination module, configured to:

[0050] Input a constraint condition into the target model, and the target model sorts the parameter vector instance and the feature quantity in descending order according to the Shapley value according to the constraint condition, determines the first preset number of the parameter vector instance and the feature quantity with the highest ranking to obtain target parameters;

[0051] Combine any two parameters in the target parameters to obtain a target parameter combination;

[0052] Divide the scene parameter space into a second preset number of parameter sub - spaces according to the target parameter combination;

[0053] Based on the parameter sub - spaces, construct an objective function according to the target model and the scene parameter safety threshold;

[0054] Generate a safety analysis result according to the objective function.

[0055] This application provides a method and system for safety analysis of an autonomous driving system based on scene parameters, which is applied to the autonomous driving system of a vehicle; the method includes: obtaining scene parameter data; the scene parameter data includes: a scene parameter vector, a scene parameter space, and a scene parameter safety threshold; the scene parameter vector is composed of several scene parameters when the vehicle is in an autonomous driving scene, and the scene parameter space is a set of the scene parameter ranges; based on the scene parameter data, obtain a parameter vector instance and a feature quantity of the autonomous driving system under the parameter vector instance; the parameter vector instance includes: the scene parameter vector and the corresponding scene parameter range; use the parameter vector instance and the feature quantity to train a linear model to obtain a target model; input a constraint condition into the target model, and the target model sorts the parameter vector instance and the feature quantity according to the Shapley value from large to small according to the constraint condition, and determines the first preset number of the parameter vector instances and the feature quantity with the top ranking to obtain target parameters; combine any two parameters in the target parameters to obtain a target parameter combination; divide the scene parameter space into a second preset number of parameter sub - spaces according to the target parameter combination; based on the parameter sub - spaces, construct an objective function according to the target model and the scene parameter safety threshold; generate a safety analysis result according to the objective function, so as to enable the safety analysis method of the autonomous driving system to perform safety analysis on the autonomous driving system based on key parameters, thereby improving the credibility of the safety analysis result of the autonomous driving system. Description of the Drawings

[0056] In order to more clearly illustrate the technical solutions of this application, the drawings required for the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0057] Figure 1 It is a flow chart of the method for safety analysis of an autonomous driving system based on scene parameters in this application;

[0058] Figure 2 It is a process diagram of the method for safety analysis of an autonomous driving system based on scene parameters in this application;

[0059] Figure 3 Process diagram of the target model in this application

[0060] Figure 4 Schematic diagram of the visualization result in this application Detailed implementation manners

[0061] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this application

[0062] Since in some technologies, the safety analysis method of the autonomous driving system cannot perform safety analysis on the autonomous driving system based on key parameters, resulting in a low credibility of the analysis results. To solve this technical problem, this application provides a safety analysis method and system for the autonomous driving system based on scenario parameters. The safety analysis method and system for the autonomous driving system based on scenario parameters will be described below

[0063] Exemplarily, with the continuous development of artificial intelligence theories and methods, autonomous driving technology has gradually matured and been applied. Since the behavior of the autonomous driving system is closely related to the scenarios in which it operates, that is, different parameters of the scenario will cause the autonomous driving system to exhibit different behaviors in them. To ensure the safety of the autonomous driving system during use, currently, safety analysis of the autonomous driving system is performed based on simulation scenario parameters, aiming to evaluate the safety of the autonomous driving system in the scenario parameter space by simulating the operation of the autonomous driving system. By obtaining scenario parameters in the vehicle automatic scenario, such as rainfall, light intensity, road width, etc., the autonomous driving system is run in the simulator for simulation, so as to analyze the safety of the autonomous driving system in this scenario parameter space

[0064] Exemplarily, the current test method for the autonomous driving system based on parameterized scenarios is as follows

[0065] Considering that a certain safety of the autonomous driving system can be represented by the constraint t≥T, where t is a characteristic quantity and T is a safety threshold (for example, for collision safety, the characteristic quantity is the minimum distance t between the vehicle under autonomous driving and other objects, and the threshold can be set to 0.1 meters, that is, the constraint t≥0.1 is formed)

[0066] Suppose a certain autonomous driving scenario is composed of a parameter vector X=(x1, x2,..., x mdetermined by (such as rainfall, light intensity, road width, initial distance of other vehicles, maximum braking speed of other vehicles, etc.). The ranges of the above parameters are [a1, b1], [a2, b2],..., [a m , b m , constituting the parameter space R = [a1, b1] × [a2, b2] ×... × [a m , b m . Then, according to a certain probability distribution, K sets of numerical values can be randomly selected from the parameter space R to form K sets of parameter vector instances (i.e., specific parameter vectors with numerical values) X i (i = 1, 2,..., K). Each X i can form a specific scenario instance in the simulator. Then, for the above scenario instances, the autonomous driving system is run in the simulator for K simulations, and the characteristic quantity t i (i = 1, 2,...., K) is measured during the simulation and it is judged whether t i ≥ T can hold, so as to analyze the safety of the autonomous driving system in the scenario parameter space.

[0067] Exemplarily, there are the following problems with the current test method for autonomous driving systems based on parameterized scenarios:

[0068] 1. Unable to reflect the relationship between the parameter space and the degree of safety

[0069] The results obtained by the current test method for autonomous driving systems based on parameterized scenarios are a series of parameter vectors and whether the autonomous driving system is safe under the said parameters. These results can only show whether the system violates the safety constraints under these discrete specific parameters, lacking both the quantification of the degree of safety and the effective reflection of the influence of different values of the parameters on the degree of safety of the system.

[0070] 2. Lack of safety analysis of key parameters

[0071] Different scenario parameters have different degrees of influence on the autonomous driving system. The current test method for autonomous driving systems based on parameterized scenarios treats all parameters equally and cannot obtain the key parameters (i.e., the parameters with the greatest influence on safety) and conduct safety analysis for the said key parameters.

[0072] 3. Discreteness and locality of testing

[0073] The current test method for autonomous driving systems based on parameterized scenarios directly uses randomly sampled scenario parameters to construct scenarios for simulation testing. These parameters are discrete sample points distributed in the parameter space. Therefore, the test results cannot effectively reflect the safety of the autonomous driving system in the entire parameter space.

[0074] Such asFigure 1 As shown, it is a flowchart of a method for analyzing the safety of an autonomous driving system based on scenario parameters in this application.

[0075] The first aspect of this application provides a method for analyzing the safety of an autonomous driving system based on scenario parameters, which is applied to the autonomous driving system of a vehicle; it includes the following steps:

[0076] As Figure 2 shown, it is a process diagram of a method for analyzing the safety of an autonomous driving system based on scenario parameters in this application.

[0077] S100: Obtain scenario parameter data; the scenario parameter data includes: scenario parameter vector X = (x1, x2,..., x m )(such as rainfall, light intensity, road width, initial distance of other vehicles, maximum braking speed of other vehicles, etc.), scenario parameter space. The ranges of the parameters in the above scenario parameter vector are [a1, b1], [a2, b2],..., [a m , b m , which constitute the parameter space R = [a1, b1] × [a2, b2] ×... × [a m , b m , and scenario parameter safety thresholds; the scenario parameter vector is composed of several scenario parameters when the vehicle is in an autonomous driving scenario, and the scenario parameter space is the set of the scenario parameter ranges.

[0078] S200: Based on the scenario parameter data, obtain a parameter vector instance and the characteristic quantity of the autonomous driving system under the parameter vector instance; the parameter vector instance includes: the scenario parameter vector and the corresponding scenario parameter range.

[0079] The step of obtaining a parameter vector instance and the characteristic quantity of the autonomous driving system under the parameter vector instance based on the scenario parameter data includes the following sub-steps:

[0080] S210: Use the uniform distribution method to obtain the third preset number of parameter vector instances in the scenario parameter space; sample K groups of parameter vector instances X1, X2,..., X K in the parameter space R with a certain probability distribution (such as uniform distribution, normal distribution method, etc.). By using the uniform distribution method or other probability distribution methods, the randomly sampled parameter vector instances can be made more random, thus laying a foundation for constructing scenario instances later.

[0081] S220: Input the parameter vector instance into the simulator to construct the scenario instance corresponding to the parameter vector instance; through the above K groups of parameter vector instances X1, X2,..., X KConstruct corresponding scenario instances in the emulator.

[0082] S230: Run the autonomous driving system using the emulator, enabling the autonomous driving system to perform simulation training under the scenario instance, and obtain the characteristic quantities of the autonomous driving system. Run the autonomous driving system through the emulator for K simulations, measure and record the corresponding characteristic quantities t1, t2,..., t K .

[0083] S300: Train a linear model using the parameter vector instance and the characteristic quantity to obtain a target model; use the data pairs (X1, t1), (X2, t2),..., (X K , t K ) formed by the above K parameter vector instances and corresponding characteristic quantities to train a linear or piecewise linear surrogate model (such as an affine function, a neural network with a ReLU activation function, etc.) to fit the above K sets of data.

[0084] As Figure 3 shown, it is a process diagram of the method for the target model in the application.

[0085] Exemplarily, the target model is configured to: fit the parameter vector instance and the characteristic quantity to generate a target characteristic quantity.

[0086] S400: Input the constraint conditions into the target model, and the target model sorts the parameter vector instance and the characteristic quantity in descending order according to the Shapley value according to the constraint conditions, determines the first preset number of the parameter vector instance and the characteristic quantity with a higher ranking, and obtains the target parameters. Among them, the first preset number can be determined according to actual analysis. The constraint condition is the safety test condition of the autonomous driving system, such as: collision test, etc.

[0087] In this embodiment, the target model first obtains the required corresponding parameter vector instance and characteristic quantity through the constraint condition, secondly calculates the Shapley value of the above parameter vector instance and characteristic quantity using the SHAP method, and finally sorts the parameter vector instance and characteristic quantity in descending order according to the size of the Shapley value, so as to determine several parameter vector instances and the characteristic quantity with a larger Shapley value as the target parameters.

[0088] Exemplarily, the SHAP method is a method for explaining model predictions based on game theory, and the Shapley value (SHAP) is used to measure the contribution of each feature to the model prediction result. Through the SHAP method, the parameter vector instance and the characteristic quantity are sorted in descending order according to the Shapley value according to the constraint condition, and the parameter vector instance and the characteristic quantity with a higher ranking represent the parameter vector instance and the characteristic quantity with a larger contribution degree to the target model prediction result.

[0089] The step of sorting the parameter vector instances and the feature quantities according to the Shapley value from large to small according to the constraint conditions, and determining the first preset number of the parameter vector instances and the feature quantities with the highest ranking to obtain the target parameters includes the following sub-steps:

[0090] S410: Calculate the Shapley values of the parameter vector instances and the feature quantities according to the constraint conditions by using the SHAP method; S420: Sort the Shapley values from large to small, and determine the first preset number of target Shapley values with the highest ranking; S430: Determine the parameter vector instances and the feature quantities corresponding to the target Shapley values to obtain the target parameters. Calculate the Shapley values of the input vector of f(X), that is, the data pair, by using the standard SHAP method, sort the Shapley values, and select the h data pairs with the largest Shapley values as the key parameters.

[0091] S500: Combine any two parameters in the target parameters to obtain a target parameter combination; the target parameter combination includes: a first target parameter and a second target parameter; the first target parameter and the second target parameter are any two parameters in the target parameters.

[0092] The step of dividing the scenario parameter space into a second preset number of parameter sub-spaces according to the target parameter combination includes the following sub-steps:

[0093] S510: Determine the scenario parameter range of the target parameter combination in the scenario parameter space; first determine the scenario parameter ranges [a1, b1] and [a2, b2] of the first target parameter x1 and the second target parameter x2 in the entire scenario parameter space.

[0094] S520: Based on the scenario parameter range, divide the first target parameter and the second target parameter according to the fourth preset number to obtain a first parameter sub-space and a second parameter sub-space; for the first target parameter x1 and the second target parameter x2, divide their ranges into L sub-intervals, and the specific steps are as follows:

[0095] For the first target parameter x1, divide its range [a1, b1] into L sub-intervals, and the width of each sub-interval is: ;

[0096] For the second target parameter x2, divide its range [a2, b2] into L sub-intervals, and the width of each sub-interval is: 。

[0097] S530: Combine the first parameter subspace and the second parameter subspace to obtain a second preset number of parameter subspaces; the second preset number is:

[0098] N = L × L;

[0099] In the formula, L is the fourth preset number.

[0100] Specifically, for each sub - interval combination of the first target parameter x1 and the second target parameter x2, generate a parameter subspace. For the i - th sub - interval [a1+(i - 1)Δ1, a1 + i - Δ1] of the first target parameter x1 and the j - th sub - interval [a2+(j - 1)Δ2, a2 + j - Δ2] of the second target parameter x2, the generated parameter subspace is:

[0101] [a1+(i - 1)Δ1, a1 + i - Δ1]×[a2+(j - 1)Δ2, a2 + j - Δ2]×[a3, a3]×...×[a m , b m .

[0102] S600: Divide the scenario parameter space into a second preset number of parameter subspaces according to the target parameter combination.

[0103] Specifically, assume an analysis of any first target parameter x1 and second target parameter x2: Divide the entire parameter space [a1, b1]×[a2, b2]×...×[a m , b m into L×L = L 2 parameter subspaces equally according to the first target parameter x1 and the second target parameter x2. Use r ij to represent the i, j - th parameter subspace, that is, r ij =[a1+(b1 - a1)i / L, a1+(b1 - a1)(i + 1) / L]×[a2+(b2 - a2)i / l, a2+(b2 - a2)(i + 1) / L]×[a3, b3]...×[a m , b m , where i, j = 0, 1,..., L - 1.

[0104] S700: Based on the parameter subspace, construct an objective function according to the target model and the scenario parameter safety threshold; the objective function is:

[0105] ;

[0106] In the formula, δ ij is the i , jQuantization security index of a parameter subspace; T is the scenario parameter safety threshold; r ij Denote the i , j th parameter subspace; f ( X ) is the target feature quantity. f ( X ) represents the target model.

[0107] S800: Generate a security analysis result according to the objective function.

[0108] The step of generating a security analysis result according to the objective function includes the following sub-steps:

[0109] S810: Calculate the quantization security index of the parameter subspace according to the objective function by using the mixed-integer linear programming method; S820: Generate a security analysis result according to the quantization security index. Based on the constructed objective function above, solve for the value of δ by using the standard mixed-integer linear programming method. ij The value of δ ij is the quantization security index for the i, j-th parameter subspace r ij . δ ij ≤0 indicates safety, δ ij >0 indicates insecurity, and the larger δ ij is, the more insecure it is.

[0110] The step of generating a security analysis result according to the quantization security index includes the following sub-steps:

[0111] S821: If the quantization security index is less than or equal to 0, generate the first information; S822: If the quantization security index is greater than 0, generate the second information; where, the first information represents the safety of the autonomous driving system, and the second information represents the insecurity of the autonomous driving system. Among them, the quantization security index corresponds to each parameter subspace, and the safety of the autonomous driving system can be judged by the quantity of the first information and the second information.

[0112] As Figure 4 shown, the schematic diagram of the visualization result in this application.

[0113] S900: Generate a visualization result based on the security analysis result; the visualization result is a heat map; the heat map is configured with a second preset number of squares, and the squares are used to display the first information or the second information. To make it more convenient for inspectors to observe the security analysis result, this application uses a heat map to display the security analysis result.

[0114] In this embodiment, a heat map is used to visualize the safety analysis results, that is, the heat map contains L 2 squares corresponding in sequence to the equal division of the parameter subspace, and the color of each square corresponds to the safety index δ ij of the parameter subspace. Among them, the darker the color, the safer the autonomous driving system, and the brighter the color, the more dangerous the autonomous driving system.

[0115] This application provides a method for analyzing the safety of an autonomous driving system based on scenario parameters. The specific implementation is as follows:

[0116] For example, the scenario parameter vector includes: fog density, precipitation, precipitation sediment, solar altitude angle, solar azimuth angle, humidity and wind intensity, initial speed, trigger distance; the constraint condition is a collision safety test; the scenario parameter safety threshold is 0.2m; according to the constraint condition, 3 target parameters are determined through the target model, such as: FOG-DENS (fog density), (PREC-DEP) precipitation sediment, (SUN-ALT) solar altitude angle; any two of the target parameters are combined to obtain a target parameter combination, and the scenario parameter space is divided into several parameter subspaces according to the target parameter combination, and a target function is constructed by using the target model and the scenario parameter safety threshold; finally, according to the target function, the safety analysis results are generated and displayed in the form of a heat map, as Figure 4 shown.

[0117] Specifically, as shown in (a) in Figure 4 , it is the safety analysis result under the combination of fog density and precipitation sediment as the target parameter. Through the safety analysis result, the safety of fog density and precipitation sediment under specific parameters can be obtained; as shown in (b) in Figure 4 , it is the safety analysis result under the combination of fog density and solar altitude angle as the target parameter. Through the safety analysis result, the safety of fog density and solar altitude angle under specific parameters can be obtained; as shown in (c) in Figure 4 , it is the safety analysis result under the combination of precipitation sediment and solar altitude angle as the target parameter. Through the safety analysis result, the safety of precipitation sediment and solar altitude angle under specific parameters can be obtained.

[0118] In this embodiment, the safety of the autonomous driving system can be observed through the heat map within the range of the target parameters, as shown in Figure 4 . Among them, the horizontal and vertical coordinates of the heat map respectively represent the ranges of the two target parameters. Through the heat map, the safety of the two target parameters in any range can be observed. Among them, the darker the color, the safer the autonomous driving system, and the brighter the color, the more dangerous the autonomous driving system.

[0119] The present application provides a method for analyzing the safety of an autonomous driving system based on scenario parameters. By providing an autonomous driving system, a parameter vector X = (x1, x2,..., x m ) of a simulation scenario, and a parameter space R = [a1, b1] × [a2, b2] ×... × [a m , b m , and a safety constraint t ≥ T, h key parameters with the greatest impact on safety are obtained through analysis. These h parameters are combined in pairs, and the entire parameter range R is equally divided into L 2 parameter subspaces R ij based on every two of these key parameters, and the quantitative safety index δ ij of each parameter subspace is calculated and visualized.

[0120] The second aspect of the present application provides a system for analyzing the safety of an autonomous driving system based on scenario parameters, including:

[0121] An acquisition module, configured to:

[0122] Acquire scenario parameter data; the scenario parameter data includes: a scenario parameter vector, a scenario parameter space, and a scenario parameter safety threshold; the scenario parameter vector is composed of several scenario parameters when the vehicle is in an autonomous driving scenario, and the scenario parameter space is a set of the scenario parameter ranges;

[0123] Based on the scenario parameter data, acquire a parameter vector instance and the feature quantity of the autonomous driving system under the parameter vector instance; the parameter vector instance includes: the scenario parameter vector and the corresponding scenario parameter range;

[0124] A training module, configured to:

[0125] Train a linear model using the parameter vector instance and the feature quantity to obtain a target model;

[0126] An analysis result determination module, configured to:

[0127] Input a constraint condition into the target model, and the target model sorts the parameter vector instance and the feature quantity from largest to smallest according to the Shapley value according to the constraint condition, determines the first preset number of the parameter vector instances and the feature quantity with a higher ranking to obtain target parameters;

[0128] Combine any two parameters in the target parameters to obtain a target parameter combination;

[0129] Divide the scenario parameter space into a second preset number of parameter subspaces according to the target parameter combination;

[0130] Based on the parameter subspace, construct an objective function according to the target model and the scenario parameter safety threshold;

[0131] Generate a security analysis result according to the objective function.

[0132] It should be noted that for the effects during the operation of the above system embodiments, reference may be made to the effects of the above method embodiments, which will not be elaborated here.

[0133] The above specific implementation manners further elaborate on the purpose, technical solutions, and beneficial effects of the embodiments of the present application. It should be understood that the above are only the specific implementation manners of the embodiments of the present application and are not used to limit the protection scope of the embodiments of the present application. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solutions of the embodiments of the present application shall be included in the protection scope of the embodiments of the present application.

Claims

1. A safety analysis method for an autonomous driving system based on scenario parameters, applied to the autonomous driving system of a vehicle; characterized in that, Including: Obtain scene parameter data; The scene parameter data includes: a scene parameter vector, a scene parameter space, and a scene parameter safety threshold; the scene parameter vector is composed of several scene parameters when the vehicle is in an autonomous driving scene, and the scene parameter space is a set of the scene parameter ranges; Based on the scene parameter data, obtain a parameter vector instance and a feature quantity of the autonomous driving system under the parameter vector instance; the parameter vector instance includes: the scene parameter vector and the corresponding scene parameter range; Use the parameter vector instance and the feature quantity to train a linear model to obtain a target model; Input a constraint condition into the target model, and the target model sorts the parameter vector instance and the feature quantity from large to small according to the Shapley value according to the constraint condition, determines the first preset number of the parameter vector instance and the feature quantity with the front ranking, and obtains target parameters; Combine any two parameters in the target parameters to obtain a target parameter combination; Divide the scene parameter space into a second preset number of parameter subspaces according to the target parameter combination; Based on the parameter subspaces, construct an objective function according to the target model and the scene parameter safety threshold; Generate a safety analysis result according to the objective function; Among them, the step of obtaining a parameter vector instance and a feature quantity of the autonomous driving system under the parameter vector instance based on the scene parameter data includes: Use the uniform distribution method to obtain a third preset number of parameter vector instances in the scene parameter space; Input the parameter vector instance into a simulator to construct a scene instance corresponding to the parameter vector instance; Use the simulator to run the autonomous driving system, enable the autonomous driving system to perform simulation training under the scene instance, and obtain the feature quantity of the autonomous driving system; Among them, the target parameter combination includes: a first target parameter and a second target parameter; the step of dividing the scene parameter space into a second preset number of parameter subspaces according to the target parameter combination includes: Determine the scene parameter range of the target parameter combination in the scene parameter space; Based on the scene parameter range, divide the first target parameter and the second target parameter according to a fourth preset number to obtain a first parameter subspace and a second parameter subspace; Combine the first parameter subspace and the second parameter subspace to obtain a second preset number of parameter subspaces; the second preset number is: N = L×L; In the formula, L is the fourth preset number.

2. The method for analyzing the safety of an autonomous driving system based on scenario parameters according to claim 1, wherein The step of sorting the parameter vector instance and the feature quantity from large to small according to the Shapley value according to the constraint condition, determining the first preset number of the parameter vector instance and the feature quantity with the front ranking, and obtaining target parameters includes: According to the constraint condition, use the SHAP method to calculate the Shapley values of the parameter vector instance and the feature quantity; Sort the Shapley values from large to small to determine the first preset number of target Shapley values with the front ranking; Determine the parameter vector instance corresponding to the target Shapley value and the feature quantity to obtain the target parameter.

3. The method for analyzing the safety of an autonomous driving system based on scenario parameters according to claim 1, wherein The target model is configured to: Fit the parameter vector instance and the feature quantity to generate a target feature quantity.

4. The method for analyzing the safety of an autonomous driving system based on scenario parameters according to claim 3, wherein The target function is: ； In the formula, δ ij is the quantization security index of the i -th j parameter subspace; T is the security threshold of the scenario parameter; r ij represents the i -th j parameter subspace; f ( X ) is the target feature quantity.

5. The method for analyzing the safety of an autonomous driving system based on scenario parameters according to claim 4, wherein The step of generating a security analysis result according to the target function includes: According to the target function, use the mixed-integer linear programming method to calculate the quantization security index of the parameter subspace; Generate a security analysis result according to the quantization security index.

6. The method for analyzing the safety of an autonomous driving system based on scenario parameters according to claim 5, wherein The step of generating a security analysis result according to the quantization security index includes: If the quantization security index is less than or equal to 0, generate a first piece of information; If the quantization security index is greater than 0, generate a second piece of information; Wherein, the first piece of information represents that the autonomous driving system is safe, and the second piece of information represents that the autonomous driving system is unsafe.

7. The method for analyzing the safety of an autonomous driving system based on scenario parameters according to claim 6, wherein The method further includes: Generate a visualization result based on the security analysis result; the visualization result is a heat map; the heat map is configured with a second preset number of squares, and the squares are used to display the first piece of information or the second piece of information.

8. An autonomous driving system safety analysis system based on scenario parameters, characterized in that, Includes: An acquisition module, configured to: Acquire scenario parameter data; The scenario parameter data includes: a scenario parameter vector, a scenario parameter space, and a scenario parameter safety threshold; the scenario parameter vector is composed of several scenario parameters when the vehicle is in an autonomous driving scenario, and the scenario parameter space is a set of the scenario parameter ranges; Based on the scenario parameter data, acquire a parameter vector instance and the feature quantity of the autonomous driving system under the parameter vector instance; the parameter vector instance includes: the scenario parameter vector and the corresponding scenario parameter range; A training module, configured to: Use the parameter vector instance and the feature quantity to train a linear model to obtain a target model; An analysis result determination module, configured to: Input constraint conditions into the target model, and the target model sorts the parameter vector instance and the feature quantity from largest to smallest according to the Shapley value according to the constraint conditions, and determines the first preset number of the parameter vector instance and the feature quantity with the highest ranking to obtain the target parameter; Combine any two parameters in the target parameter to obtain a target parameter combination; Divide the scenario parameter space into a second preset number of parameter subspaces according to the target parameter combination; Based on the parameter subspace, construct a target function according to the target model and the scenario parameter safety threshold; Generate a security analysis result according to the target function; Wherein, the acquisition module is further configured to: Use the uniform distribution method to acquire a third preset number of parameter vector instances in the scenario parameter space; Input the parameter vector instance into a simulator to construct a scenario instance corresponding to the parameter vector instance; Use the simulator to run the autonomous driving system, and make the autonomous driving system perform simulation training under the scenario instance to acquire the feature quantity of the autonomous driving system. Wherein, the target parameter combination includes: a first target parameter and a second target parameter; the analysis result determination module is further configured to: Determine the scenario parameter range of the target parameter combination in the scenario parameter space; Based on the scenario parameter range, divide the first target parameter and the second target parameter according to a fourth preset quantity to obtain a first parameter subspace and a second parameter subspace; Combine the first parameter subspace and the second parameter subspace to obtain a second preset quantity of parameter subspaces; the second preset quantity is: N = L × L; In the formula, L is the fourth preset quantity.

Images