A method and device for generating edge simulation scenes for autonomous driving tests

Abstract

The present application provides a method and device for generating edge simulation scenarios for autonomous driving tests, the method comprising: constructing a parameter probability space of the nominal behavior using the parameter value corresponding to each behavior parameter variable in the mathematical description model corresponding to the nominal behavior; for each behavior parameter variable, determining the edge scenario parameter interval corresponding to the behavior parameter variable based on the parameter probability space and a preset edge probability threshold, and sampling the parameter value in the edge scenario parameter interval to obtain multiple edge scenario parameters corresponding to the behavior parameter variable; and combining parameters using the edge scenario parameters corresponding to multiple behavior parameter variables to generate multiple target edge simulation scenarios corresponding to the nominal behavior. Through the method and device, edge scenarios with higher potential risks in natural driving can be identified, and the authenticity and rationality of the generated edge scenarios are improved.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving testing technology, and in particular to a method and device for generating edge simulation scenarios for autonomous driving testing. Background Art

[0002] In the development of autonomous vehicles, scenario-based testing is a key step in verifying their performance. The design of test scenarios should cover common driving situations as well as challenging edge scenarios. Although edge scenarios occur relatively infrequently, they place higher demands on the system's capabilities. The collected natural driving datasets can provide a large number of actual driving examples, from which edge cases of nominal scenarios can be mined.

[0003] Nominal scenarios refer to situations that frequently occur in daily driving, while edge scenarios refer to situations where the behavioral parameters of nominal scenarios are close to extreme values. This scenario can effectively expose potential problems in the autonomous driving system and help speed up defect discovery during the testing process. However, not all designed edge scenarios are reasonable. Their rationality and explainability directly depend on whether the corresponding scenarios can be found in real natural driving data sets. This means that manually set edge scenarios that lack a realistic basis are difficult to explain their practical significance and may not occur in real environments. Therefore, in order to provide more reasonable scenarios for the testing of autonomous driving systems, how to generate real edge simulation scenarios has become a technical issue that cannot be underestimated. Summary of the invention

[0004] In view of this, the purpose of this application is to provide a method and device for generating edge simulation scenarios for autonomous driving tests, which can accurately describe the statistical characteristics of nominal behavior and identify edge scenarios that occur less frequently in natural driving but have higher potential risks. Moreover, because the edge cases come from actual data, the generated edge simulation scenarios can better reflect the driving environment in the real world, thereby improving the authenticity and rationality of the generated edge scenarios.

[0005] In a first aspect, an embodiment of the present application provides an edge simulation scenario generation method for an autonomous driving test, the edge simulation scenario generation method comprising:

[0006] Establishing a mathematical description model corresponding to the nominal behavior of the autonomous driving system, and extracting parameter values ​​corresponding to the behavior parameter variables in the mathematical description model from the autonomous driving data set;

[0007] Constructing a marginal probability density function corresponding to each behavior parameter variable using a parameter value corresponding to each behavior parameter variable, and generating a parameter probability space of the nominal behavior based on a plurality of marginal probability density functions;

[0008] For each behavior parameter variable, determine the edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and a preset edge probability threshold, and sample the parameter values ​​in the edge scene parameter interval to obtain multiple edge scene parameters corresponding to the behavior parameter variable;

[0009] Parameter combination is performed using edge scenario parameters corresponding to a plurality of behavior parameter variables to generate a plurality of target edge simulation scenarios corresponding to the nominal behavior.

[0010] Furthermore, the method of constructing a marginal probability density function corresponding to each behavior parameter variable using the parameter value corresponding to each behavior parameter variable, and generating the parameter probability space of the nominal behavior based on multiple marginal probability density functions, includes:

[0011] For each behavior parameter variable, fitting the marginal distribution of the parameter value corresponding to the behavior parameter variable to obtain the marginal probability density function corresponding to the behavior parameter variable;

[0012] A joint probability density function between marginal probability density functions corresponding to a plurality of behavior parameter variables is established by using a Copula function, and the joint probability density function is used as the parameter probability space of the nominal behavior.

[0013] Furthermore, the determining of the edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and a preset edge probability threshold includes:

[0014] Determine the conditional probability distribution of each parameter value corresponding to the behavior parameter variable according to the parameter probability space;

[0015] Determining a parameter boundary value from a plurality of parameter values ​​corresponding to the behavior parameter variable based on the preset edge probability threshold and the conditional probability distribution of each parameter value;

[0016] Based on the parameter boundary value, an edge scene parameter interval corresponding to the behavior parameter variable is determined from multiple parameter values ​​corresponding to the behavior parameter variable.

[0017] Furthermore, the step of determining a parameter boundary value from a plurality of parameter values ​​corresponding to the behavior parameter variable based on the preset edge probability threshold and the conditional probability distribution of each parameter value includes:

[0018] Determine a first probability threshold and a second probability threshold based on the preset edge probability threshold;

[0019] Taking the parameter value whose conditional probability is equal to the first probability threshold among the multiple parameter values ​​as the first boundary value, and taking the parameter value whose conditional probability is equal to the second probability threshold among the multiple parameter values ​​as the second boundary value;

[0020] The parameter boundary value is determined from the first boundary value and the second boundary value based on the boundary value selection rule corresponding to the behavior parameter variable.

[0021] Further, the determining, based on the parameter boundary value, from a plurality of parameter values ​​corresponding to the behavior parameter variable to obtain an edge scene parameter interval corresponding to the behavior parameter variable includes:

[0022] Determine a plurality of target parameter values ​​adjacent to the parameter boundary value from the plurality of parameter values; wherein a ratio between the number of the plurality of target parameter values ​​and the number of the plurality of parameter values ​​reaches a preset threshold;

[0023] The smallest target parameter value among multiple target parameter values ​​is used as the first edge scene boundary parameter, the largest target parameter value among multiple target parameter values ​​is used as the second edge scene boundary parameter, and the edge scene parameter interval is generated according to the first edge scene boundary parameter and the second edge scene boundary parameter.

[0024] In a second aspect, an embodiment of the present application further provides an edge simulation scene generation device for an autonomous driving test, the edge simulation scene generation device comprising:

[0025] A parameter value acquisition module, used to establish a mathematical description model corresponding to the nominal behavior of the autonomous driving system, and extract parameter values ​​corresponding to the behavior parameter variables in the mathematical description model from the autonomous driving data set;

[0026] A parameter probability space construction module, used to construct a marginal probability density function corresponding to each behavior parameter variable using a parameter value corresponding to each behavior parameter variable, and generate a parameter probability space of the nominal behavior based on a plurality of marginal probability density functions;

[0027] An edge scene parameter determination module is used to determine, for each behavior parameter variable, an edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and a preset edge probability threshold, and sample parameter values ​​in the edge scene parameter interval to obtain a plurality of edge scene parameters corresponding to the behavior parameter variable;

[0028] The edge simulation scenario generation module is used to combine parameters using edge scenario parameters corresponding to multiple behavior parameter variables to generate multiple target edge simulation scenarios corresponding to the nominal behavior.

[0029] Furthermore, when the parameter probability space construction module is used to construct the marginal probability density function corresponding to each behavior parameter variable using the parameter value corresponding to each behavior parameter variable, and generate the parameter probability space of the nominal behavior based on multiple marginal probability density functions, the parameter probability space construction module is also used to:

[0030] For each behavior parameter variable, fitting the marginal distribution of the parameter value corresponding to the behavior parameter variable to obtain the marginal probability density function corresponding to the behavior parameter variable;

[0031] A joint probability density function between marginal probability density functions corresponding to a plurality of behavior parameter variables is established by using a Copula function, and the joint probability density function is used as the parameter probability space of the nominal behavior.

[0032] Furthermore, when the edge scene parameter determination module is used to determine the edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and the preset edge probability threshold, the edge scene parameter determination module is also used to:

[0033] Determine the conditional probability distribution of each parameter value corresponding to the behavior parameter variable according to the parameter probability space;

[0034] Determining a parameter boundary value from a plurality of parameter values ​​corresponding to the behavior parameter variable based on the preset edge probability threshold and the conditional probability distribution of each parameter value;

[0035] Based on the parameter boundary value, an edge scene parameter interval corresponding to the behavior parameter variable is determined from multiple parameter values ​​corresponding to the behavior parameter variable.

[0036] In the third aspect, an embodiment of the present application also provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the edge simulation scenario generation method for autonomous driving testing as described above are performed.

[0037] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the edge simulation scenario generation method for the autonomous driving test as described above are executed.

[0038] The embodiment of the present application provides a method and device for generating edge simulation scenarios for autonomous driving tests. First, a mathematical description model corresponding to the nominal behavior of the autonomous driving system is established, and the parameter values ​​corresponding to the behavior parameter variables in the mathematical description model are extracted from the autonomous driving data set; then, the edge probability density function corresponding to each behavior parameter variable is constructed using the parameter value corresponding to each behavior parameter variable, and the parameter probability space of the nominal behavior is generated based on multiple edge probability density functions; for each behavior parameter variable, the edge scene parameter interval corresponding to the behavior parameter variable is determined based on the parameter probability space and the preset edge probability threshold, and the parameter values ​​in the edge scene parameter interval are sampled to obtain multiple edge scene parameters corresponding to the behavior parameter variable; finally, the edge scene parameters corresponding to the multiple behavior parameter variables are used for parameter combination to generate multiple target edge simulation scenarios corresponding to the nominal behavior. The present application can accurately describe the statistical characteristics of the nominal behavior, identify those edge scenarios with low frequency of occurrence in natural driving but high potential risks, and because the edge cases come from actual data, the generated edge simulation scenarios can better reflect the driving environment in the real world.

[0039] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A flowchart of a method for generating an edge simulation scenario for an autonomous driving test provided in an embodiment of the present application;

[0042] Figure 2 A flowchart of a method for determining an edge scene parameter interval provided in an embodiment of the present application;

[0043] Figure 3 A schematic diagram of the structure of an edge simulation scene generation device for an autonomous driving test provided in an embodiment of the present application;

[0044] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application usually described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, each other embodiment obtained by those skilled in the art without making creative work belongs to the scope of protection of the present application.

[0046] First, the application scenarios to which this application is applicable are introduced. This application can be applied in the field of autonomous driving test technology.

[0047] In the development of autonomous vehicles, scenario-based testing is a key step in verifying their performance. The design of test scenarios should cover common driving situations as well as challenging edge scenarios. Although edge scenarios occur relatively infrequently, they place higher demands on the system's capabilities. The collected natural driving datasets can provide a large number of actual driving examples, from which edge cases of nominal scenarios can be mined.

[0048] Nominal scenarios refer to situations that frequently occur in daily driving, while edge scenarios refer to situations where the behavioral parameters of nominal scenarios are close to extreme values. This scenario can effectively expose potential problems in the autonomous driving system and help speed up defect discovery during the testing process. However, not all designed edge scenarios are reasonable. Their rationality and explainability directly depend on whether the corresponding scenarios can be found in real natural driving data sets. This means that manually set edge scenarios that lack a realistic basis are difficult to explain their practical significance and may not occur in real environments. Therefore, in order to provide more reasonable scenarios for the testing of autonomous driving systems, how to generate real edge simulation scenarios has become a technical issue that cannot be underestimated.

[0049] Based on this, an embodiment of the present application provides a method for generating edge simulation scenarios for autonomous driving tests, which improves the authenticity and rationality of the generated edge scenarios.

[0050] See also Figure 1 , Figure 1 This is a flow chart of a method for generating edge simulation scenarios for autonomous driving tests provided in an embodiment of the present application. Figure 1 As shown in, the edge simulation scene generation method provided by the embodiment of the present application includes:

[0051] S101, establishing a mathematical description model corresponding to the nominal behavior of the autonomous driving system, and extracting parameter values ​​corresponding to the behavior parameter variables in the mathematical description model from the autonomous driving data set.

[0052] Here, in the field of autonomous driving, nominal behavior refers to the operation and response of the autonomous driving system within normal, expected and design specifications. Nominal behavior is the expected performance of the autonomous driving system in situations that frequently occur in daily driving, such as the following behavior and lane change behavior of the autonomous driving system. The mathematical description model can describe the model of the above nominal behavior. According to the differences in the time and continuity characteristics of the nominal behavior, the behavior description needs to define the model one by one. In addition, each nominal behavior is described by a set of behavior parameter variables, and the behavior parameter variables include but are not limited to the speed, acceleration, distance, or other behavior characteristic parameters of the behavior participants. As an example, when the nominal behavior is the following behavior of the autonomous driving system, the distance between the front and rear vehicles, the speed and acceleration of the rear vehicle, and the speed of the front vehicle, four parameters can fully describe a following behavior, and these parameters can be defined as the behavior parameter variables of the nominal behavior. The number of behavior parameter variables should be as small as possible to fully and completely describe the nominal behavior of the nominal scene. The natural driving data set refers to the real driving data collected in the actual road environment, and the data is required to record the dynamic parameter sequence of the autonomous driving vehicle under various driving conditions.

[0053] For the above step S101, in the specific implementation, for the nominal behavior of the autonomous driving system, first establish a nominal mathematical description model, and then extract the parameter value corresponding to each behavior parameter variable in the mathematical description model from the autonomous driving data set. When extracting the behavior parameters, the data set needs to be reintegrated first. The integration is based on the perspective of the main vehicle or the responsible vehicle, that is, it is necessary to define the traffic participants in the data set as the main vehicle in turn, and extract all the interactive behaviors encountered by the main vehicle in the formal process based on the perspective of the main vehicle. For example, in a scene of a nominal behavior in which the main vehicle participates, first use the driving frame of the main vehicle as the time axis, traverse the data set time axis to find all behavior instances that meet the description model, and record the behavior duration of each instance and the time series of the characterization parameters of the behavior.

[0054] The essence of the scenario generalization that this application focuses on is the generalization of behavior, and the nominal behavior can be subdivided into continuous behavior and instantaneous behavior according to its timing characteristics and disassembly. Among them, continuous behavior means that the behavior can be disassembled in the time frame and the duration can be infinite, such as car-following behavior; instantaneous behavior refers to the behavior that cannot be disassembled in the time frame and has a limited duration, such as cutting-in behavior.

[0055] Here, we first take the continuous behavior as an example, take the following behavior of the autonomous driving system as the nominal behavior, and construct a mathematical description model of the following behavior, which is described by the following judgment formula:

[0056]

[0057] in, Indicates the lane of the following vehicle in the following behavior. Indicates the lane of the vehicle ahead in the following behavior. is a Boolean variable, representing whether the vehicle is in the lane-changing state. Indicates whether the following vehicle is changing lanes. Indicates whether the front vehicle is changing lanes. Indicates No. The behavior parameter variables of the following behavior are characterized by the following variables:

[0058]

[0059] in, For car following behavior, They represent the time series vector of the leading vehicle speed, the time series vector of the trailing vehicle speed, the time series vector of the trailing vehicle acceleration, and the time series vector of the trailing vehicle distance under the following behavior. For the The respective variable values ​​of the frames. These parameter variables need to be extracted from the natural driving dataset as the parameter values ​​of the behavior parameter variables.

[0060] Here, taking instantaneous behavior as an example, taking the cut-in behavior of the autonomous driving system as the nominal behavior, a mathematical description model of the cut-in behavior is constructed, which is described by the following judgment formula:

[0061]

[0062] The above judgment formula is interpreted as: the cutting vehicle is located in the adjacent lane of the cut-in vehicle, and the cutting vehicle is in the lane-changing state, while the cut-in vehicle is not in the lane-changing state, wherein: , Represents the lane where the cutting car and the cut-in car are located. represents the duration of the cut-in behavior. In addition, the trigger point for the cut-in behavior to be completed, i.e., the second row of discriminants, is when the cut-in vehicle and the cut-in vehicle are in the same lane and are in front of the cut-in vehicle. is the longitudinal position coordinate of the cutting vehicle and the cut vehicle. The behavior parameter variables of the cutting behavior are characterized by the following variables:

[0063]

[0064] in, For the cut-in behavior, since the cut-in behavior is instantaneous, its smallest behavior unit is one behavior instead of one frame, so are the characteristic values ​​of the speed of the cutting vehicle and the speed of the cut-in vehicle instead of the time series vector, are the initial and final values ​​of the longitudinal relative distance during the cutting process, , , For the t The respective variable values ​​under the frame, mean These parameter variables need to be extracted from the natural driving data set as the parameter values ​​of the behavior parameter variables.

[0065] S102, constructing a marginal probability density function corresponding to each behavior parameter variable using a parameter value corresponding to each behavior parameter variable, and generating a parameter probability space of the nominal behavior based on a plurality of marginal probability density functions.

[0066] Here, the parameter probability space describes the possibility of each behavior parameter combination occurring under a given nominal behavior scenario.

[0067] For the above step S102, in the specific implementation, a large-scale natural driving data set is used to analyze the behavior patterns of human drivers under various conditions in the real world. Through statistics and modeling of actual driving data, the marginal probability density function of each behavior parameter variable is constructed, thereby forming a complete logical scenario probability space. The logical scenario probability space is continuous in all dimensions. Its essence is the joint probability density function of the behavior parameter variables in the autonomous driving logic scenario, which includes the conditional probability of a certain behavior parameter variable, or the joint probability information between the behavior parameter variables.

[0068] With respect to the above step S102, the marginal probability density function corresponding to each behavior parameter variable is constructed using the parameter value corresponding to each behavior parameter variable, and the parameter probability space of the nominal behavior is generated based on multiple marginal probability density functions, including:

[0069] Step 1021 , for each behavior parameter variable, fitting the marginal distribution of the parameter value corresponding to the behavior parameter variable to obtain the marginal probability density function corresponding to the behavior parameter variable.

[0070] For the above step 1021, in the specific implementation, for each behavior parameter variable, the marginal distribution of the parameter value corresponding to the behavior parameter variable is fitted to obtain the marginal probability density function corresponding to the behavior parameter variable. Specifically, according to the embodiment provided by the present application, a GMM (Gaussian mixture model) can be used to fit the marginal distribution of the parameter value of each behavior parameter variable. Here, assume Behavior parameter variables The parameter value of each All obey mixed Gaussian distribution:

[0071]

[0072] in, is the optimal Gaussian mixture model of random variables, is the parameter value, is the number of Gaussian components, It is The mixture weights of the Gaussian model satisfy and , It is The probability density function of a Gaussian distribution.

[0073] After fitting the probability density function of the marginal distribution, the marginal probability density function corresponding to the behavior parameter variable is expressed as .

[0074] Here, when constructing the marginal probability density function, in addition to using the Gaussian mixture model (GMM), other probability models may also be considered, such as normal distribution, exponential distribution, Weibull distribution, etc., which are not specifically limited in this application. These distribution models can also capture the distribution characteristics of random variables and model the probability density function.

[0075] Step 1022: Use a Copula function to establish a joint probability density function between marginal probability density functions corresponding to a plurality of behavior parameter variables, and use the joint probability density function as the parameter probability space of the nominal behavior.

[0076] The Copula function is selected as the basic method for modeling the joint probability density function of random variables. The Copula function allows the probability density function of the marginal distribution to be separated from the probability density function of the joint distribution, so as to model the joint probability density function between different random variables when the joint relationship is unknown. , for any joint distribution function and the univariate cumulative distribution function (in ), according to Sklar's theorem, there must be a Copula function , so that:

[0077]

[0078] In the embodiment provided in the present application, a Gaussian Copula function is selected to model the joint probability density function between multiple behavior parameter variables. The form of the Gaussian Copula function is as follows:

[0079]

[0080] in, , represents the cumulative distribution function of the standard normal distribution, is the cumulative distribution function of the multidimensional normal distribution, and its covariance matrix is At the same time, based on the parameter values ​​in the natural driving data set, the maximum likelihood estimation method is used to determine the covariance matrix in the Copula function Finally, the marginal probability density functions corresponding to multiple behavior parameter variables are combined with the Copula function to obtain the final joint probability density function , which is the parameter probability space of the autonomous driving logic scenario.

[0081] Thus, according to the above steps 1021-1022, the present application selects the Gaussian copula function to model the joint distribution between the behavior parameter variables. Using the Gaussian copula function to model the joint probability density function between random variables is an effective method, but in some applicable cases, other types of copula functions, such as Archimedean Copulas or Vine Copulas, may be considered. Or other models or algorithms such as regression, neural networks, etc. may be used. These methods can also be used to model the correlation between different random variables, and may have better fitting effects under certain specific data distributions.

[0082] Thus, according to the above steps 1021-1022, the joint probability distribution model of the behavior parameter variables in the nominal behavior scenario is constructed by using the parameter values ​​of the behavior parameter variables in the natural driving data set. First, the parameter values ​​of the behavior parameter variables in the nominal behavior are extracted, and the marginal distribution of these parameters is fitted by the Gaussian mixture model, and then the joint distribution model is constructed using the Gaussian copula function. By modeling the joint probability density function between random variables, the relationship between variables in the nominal behavior can be accurately described.

[0083] S103, for each behavior parameter variable, determine the edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and the preset edge probability threshold, and sample the parameter values ​​in the edge scene parameter interval to obtain multiple edge scene parameters corresponding to the behavior parameter variable.

[0084] Here, the preset edge probability threshold defines which parameter combinations are considered to be non-nominal behaviors, i.e., edge behaviors. Here, the preset edge probability threshold can be set to 90%, which is not specifically limited in this application. When the preset edge probability threshold is 90%, it is defined that when the probability of a parameter falls outside the threshold, it is considered to be an edge parameter. The edge scene parameter interval refers to the value range of the edge scene parameter when constructing an edge simulation scene. The edge scene parameter is located within the edge scene parameter interval and is used to constitute the scene parameter of the edge scene.

[0085] For the above step S103, the parameters of edge behavior are mainly extracted based on a preset edge probability threshold. In the specific implementation, after the probability density function of the joint distribution of the behavior parameter variables is established in step S102, a reasonable preset edge probability threshold is determined through the distribution of parameter combinations in the autonomous driving data set, so that the parameter combination falling outside the preset edge probability threshold is identified as edge behavior. Specifically, the edge scene parameter interval corresponding to the behavior parameter variable is determined based on the parameter probability space and the preset edge probability threshold. Then, the parameter values ​​in the edge scene parameter interval are sampled to obtain multiple edge scene parameters corresponding to the behavior parameter variable. Here, you can choose to use Monte Carlo Methods, Importance Sampling or other random sampling techniques to sample the edge scene parameters, which is not specifically limited in this application.

[0086] Figure 2 Flow chart of a method for determining an edge scene parameter interval provided in an embodiment of the present application. Figure 2 As shown, for the above step S103, determining the edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and the preset edge probability threshold includes:

[0087] S201, determining the conditional probability distribution of each parameter value corresponding to the behavior parameter variable according to the parameter probability space.

[0088] Here, the conditional probability distribution describes the probability of a behavior parameter variable taking a certain parameter value given certain conditions.

[0089] For the above step S201, in the specific implementation, the conditional probability distribution of each parameter value corresponding to the behavior parameter variable is determined according to the parameter probability space. Here, as an example, if the behavior parameter variable has four parameter values , it is necessary to find the conditional probability distribution of the fourth parameter when the other three parameters are fixed. Continuing with the embodiment in the above steps, still taking the following behavior as an example, the parameters such as the speed and acceleration of the front and rear vehicles are fixed as conditions, and the conditional probability distribution of the following distance is determined based on the parameter probability space.

[0090] S202: Determine a parameter boundary value from a plurality of parameter values ​​corresponding to the behavior parameter variable based on the preset edge probability threshold and the conditional probability distribution of each parameter value.

[0091] Here, the parameter boundary value is used to represent that when the behavior parameter variable takes the parameter boundary value, the behavior parameter variable belongs to the marginal behavior.

[0092] Regarding the above step S202, in a specific implementation, a parameter boundary value is determined from a plurality of parameter values ​​corresponding to the behavior parameter variable according to a preset edge probability threshold and a conditional probability distribution of each parameter value.

[0093] As an optional embodiment, for the above step S202, determining the parameter boundary value from the multiple parameter values ​​corresponding to the behavior parameter variable based on the preset edge probability threshold and the conditional probability distribution of each parameter value includes:

[0094] Step 2021: Determine a first probability threshold and a second probability threshold based on the preset edge probability threshold.

[0095] For the above step 2021, in the specific implementation, the first probability threshold and the second probability threshold are determined according to the preset edge probability threshold. Here, as an example, when the preset edge probability threshold is 90%, it means that under certain conditions, only 10% of the behavior parameters that the drivers will perform will be used as the parameter boundary value of the edge parameter. When the preset edge probability threshold is 90%, the first probability threshold is 10% and the second probability threshold is 90%.

[0096] Step 2022: taking the parameter value among the multiple parameter values ​​whose conditional probability is equal to the first probability threshold as the first boundary value, and taking the parameter value among the multiple parameter values ​​whose conditional probability is equal to the second probability threshold as the second boundary value.

[0097] Regarding the above step 2022, in the specific implementation, the parameter value among the multiple parameter values ​​whose conditional probability is equal to the first probability threshold is used as the first boundary value, and the parameter value among the multiple parameter values ​​whose conditional probability is equal to the second probability threshold is used as the second boundary value.

[0098] Step 2023: Determine the parameter boundary value from the first boundary value and the second boundary value based on the boundary value selection rule corresponding to the behavior parameter variable.

[0099] Here, continuing with the above example, when the preset edge probability threshold is set to 90%, the basis for selecting the parameter boundary value can be explained as follows: if the probability of the parameter value of the behavior parameter variable of the autonomous driving vehicle exceeds 90% of the other parameter values ​​of the human driver under the behavior parameter variable in terms of aggressiveness, then the scene is considered to be an edge scene.

[0100] Here, the boundary value selection rule is determined according to the behavior property of the behavior parameter variable, and needs to be determined according to the change of the behavior parameter variable and the potential risk it brings. Here, as an example, when the behavior parameter variable is the cutting-in speed of the rear vehicle in the cutting-in behavior, the faster the cutting-in speed, the higher the potential risk. At this time, the boundary value selection rule is to select the smaller value from the first boundary value and the second boundary value as the parameter boundary value. When the driving parameter variable is the rear vehicle speed of the rear vehicle in the following behavior, the faster the rear vehicle speed, the higher the potential risk. At this time, the boundary value selection rule is to select the larger value from the first boundary value and the second boundary as the parameter boundary value. When the driving parameter variable is the following distance in the following behavior, the shorter the following distance, the higher the potential risk. At this time, the boundary value selection rule is to select the smaller value from the first boundary value and the second boundary value as the parameter boundary value.

[0101] Regarding the above step 2023, during the specific implementation, the boundary value corresponding to the behavior parameter variable is obtained, and the parameter boundary value of the behavior parameter variable is determined from the first boundary value and the second boundary value according to the boundary value selection rule corresponding to the behavior parameter variable.

[0102] S203: Determine, based on the parameter boundary value, an edge scene parameter interval corresponding to the behavior parameter variable from a plurality of parameter values ​​corresponding to the behavior parameter variable.

[0103] Regarding the above step S203, in a specific implementation, based on the parameter boundary value determined in the above step S202, the edge scene parameter interval corresponding to the behavior parameter variable is determined from multiple parameter values ​​corresponding to the behavior parameter variable.

[0104] As an optional embodiment, for the above step S203, determining the edge scene parameter interval corresponding to the behavior parameter variable from multiple parameter values ​​corresponding to the behavior parameter variable based on the parameter boundary value includes:

[0105] Step 2031: Determine a plurality of target parameter values ​​adjacent to the parameter boundary value from a plurality of parameter values.

[0106] The ratio between the number of the plurality of target parameter values ​​and the number of the plurality of parameter values ​​reaches a preset threshold value. Here, the preset threshold value may be 0.1, which is not specifically limited in this application.

[0107] For the above step 2031, in a specific implementation, a preset threshold is used to determine a plurality of target parameter values ​​adjacent to the parameter boundary value from the plurality of parameter values. Here, as an example, when the preset threshold is 0.1, the parameter values ​​in the first 5% of the parameter boundary values ​​and the parameter values ​​in the last 5% of the parameter boundary values ​​are taken as the target parameter values.

[0108] Step 2032: Use the smallest target parameter value among multiple target parameter values ​​as the first edge scene boundary parameter, use the largest target parameter value among multiple target parameter values ​​as the second edge scene boundary parameter, and generate the edge scene parameter interval according to the first edge scene boundary parameter and the second edge scene boundary parameter.

[0109] For the above step 2032, in the specific implementation, the smallest target parameter value among the multiple target parameter values ​​is used as the first edge scene boundary parameter, and the largest target parameter value among the multiple target parameter values ​​is used as the second edge scene boundary parameter, and an edge scene parameter interval is generated according to the first edge scene boundary parameter and the second edge scene boundary parameter, and the two boundary values ​​of the edge scene parameter interval are the first edge scene boundary parameter and the second edge scene boundary parameter.

[0110] Thus, according to the above steps, edge scenes are identified by setting a preset edge probability threshold under conditional probability. This method can effectively screen out potential high-risk scenes from a large amount of natural driving data. And edge parameters are sampled from the edge scene parameter interval to generalize the edge scene. This method realizes the generalization from small samples to large samples and ensures that the scene remains reasonable after generalization.

[0111] S104, performing parameter combination using edge scenario parameters corresponding to a plurality of behavior parameter variables to generate a plurality of target edge simulation scenarios corresponding to the nominal behavior.

[0112] For the above step S104, in the specific implementation, after determining the multiple edge scene parameters corresponding to each behavior parameter variable, the edge scene parameters corresponding to the multiple behavior parameter variables are combined into different scenarios to generate multiple target edge simulation scenarios corresponding to the nominal behavior. Here, since the parameter probability space modeled in step S102 is based on the joint probability density function, it has the characteristic of continuity in value, and the edge scene parameters are sampled from the edge scene parameter interval, and finally the edge simulation scene is generated based on the edge scene parameters. The target edge simulation scene obtained in this way represents a scene that is less common but potentially dangerous in actual driving.

[0113] The edge simulation scene generation method for autonomous driving test provided in an embodiment of the present application is as follows: first, a mathematical description model corresponding to the nominal behavior of the autonomous driving system is established, and parameter values ​​corresponding to the behavior parameter variables in the mathematical description model are extracted from the autonomous driving data set; then, an edge probability density function corresponding to each behavior parameter variable is constructed using the parameter value corresponding to each behavior parameter variable, and a parameter probability space of the nominal behavior is generated based on multiple edge probability density functions; for each behavior parameter variable, an edge scene parameter interval corresponding to the behavior parameter variable is determined based on the parameter probability space and a preset edge probability threshold, and the parameter values ​​in the edge scene parameter interval are sampled to obtain multiple edge scene parameters corresponding to the behavior parameter variable; finally, the edge scene parameters corresponding to the multiple behavior parameter variables are used to perform parameter combination to generate multiple target edge simulation scenes corresponding to the nominal behavior.

[0114] This application first defines and establishes a mathematical description model of nominal behavior, extracts nominal behavior and its behavior parameter variables in a natural driving data set, then models the joint probability density function of the behavior parameter variables of the nominal behavior, forms a parameter probability space of the joint distribution of parameters, sets the threshold boundary of the edge probability and identifies the edge scenes, and finally samples the edge parameters from the edge scene parameter interval to generate edge scenes using the edge parameters.

[0115] This application defines and establishes a mathematical description model of nominal behavior, combines the behavior characterization parameters extracted from the natural driving data set, and models the joint probability density function of the nominal behavior characterization parameters, thereby forming a probability space of joint distribution of parameters. This method can accurately describe the statistical characteristics of nominal behavior and identify those edge scenarios that have a low frequency of occurrence in natural driving but have higher potential risks. Compared with traditional methods, this application can more comprehensively consider the complex and changeable situations in actual driving. In particular, in the two typical scenarios of following and cutting-in behavior, by accurately modeling the following distance, speed, acceleration, and cut-in speed, time, and distance, edge cases that may cause harm can be effectively identified, providing more reasonable scenario support for the testing of autonomous driving systems.

[0116] After setting the threshold boundary of the edge probability, this application samples edge parameters from the parameter probability space, thereby achieving the generalization of edge scenarios. This method can improve the coverage of edge cases, and because the edge cases come from actual data, the generated edge simulation scenarios can better reflect the driving environment in the real world. Compared with traditional methods, this application effectively enhances the generalization ability of discrete real-world edge scenarios by reasonably setting the edge probability threshold and performing targeted conditional sampling nearby based on the boundary.

[0117] See also Figure 3 , Figure 3 This is a schematic diagram of the structure of an edge simulation scene generation device for an autonomous driving test provided in an embodiment of the present application. Figure 3 As shown in , the edge simulation scene generation device 300 includes:

[0118] A parameter value acquisition module 301 is used to establish a mathematical description model corresponding to the nominal behavior of the autonomous driving system, and extract parameter values ​​corresponding to the behavior parameter variables in the mathematical description model from the autonomous driving data set;

[0119] A parameter probability space construction module 302 is used to construct a marginal probability density function corresponding to each behavior parameter variable using a parameter value corresponding to each behavior parameter variable, and generate a parameter probability space of the nominal behavior based on a plurality of marginal probability density functions;

[0120] An edge scene parameter determination module 303 is used to determine, for each behavior parameter variable, an edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and a preset edge probability threshold, and sample parameter values ​​in the edge scene parameter interval to obtain a plurality of edge scene parameters corresponding to the behavior parameter variable;

[0121] The edge simulation scenario generation module 304 is used to combine parameters using edge scenario parameters corresponding to multiple behavior parameter variables to generate multiple target edge simulation scenarios corresponding to the nominal behavior.

[0122] Further, when the parameter probability space construction module 302 is used to construct the marginal probability density function corresponding to each behavior parameter variable using the parameter value corresponding to each behavior parameter variable, and generate the parameter probability space of the nominal behavior based on multiple marginal probability density functions, the parameter probability space construction module 302 is also used to:

[0123] For each behavior parameter variable, fitting the marginal distribution of the parameter value corresponding to the behavior parameter variable to obtain the marginal probability density function corresponding to the behavior parameter variable;

[0124] A joint probability density function between marginal probability density functions corresponding to a plurality of behavior parameter variables is established by using a Copula function, and the joint probability density function is used as the parameter probability space of the nominal behavior.

[0125] Further, when the edge scene parameter determination module 303 is used to determine the edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and the preset edge probability threshold, the edge scene parameter determination module 303 is also used to:

[0126] Determine the conditional probability distribution of each parameter value corresponding to the behavior parameter variable according to the parameter probability space;

[0127] Determining a parameter boundary value from a plurality of parameter values ​​corresponding to the behavior parameter variable based on the preset edge probability threshold and the conditional probability distribution of each parameter value;

[0128] Based on the parameter boundary value, an edge scene parameter interval corresponding to the behavior parameter variable is determined from multiple parameter values ​​corresponding to the behavior parameter variable.

[0129] Further, when the edge scene parameter determination module 303 is used to determine the parameter boundary value from the multiple parameter values ​​corresponding to the behavior parameter variable based on the preset edge probability threshold and the conditional probability distribution of each parameter value, the edge scene parameter determination module 303 is also used to:

[0130] Determine a first probability threshold and a second probability threshold based on the preset edge probability threshold;

[0131] Taking the parameter value whose conditional probability is equal to the first probability threshold among the multiple parameter values ​​as the first boundary value, and taking the parameter value whose conditional probability is equal to the second probability threshold among the multiple parameter values ​​as the second boundary value;

[0132] The parameter boundary value is determined from the first boundary value and the second boundary value based on the boundary value selection rule corresponding to the behavior parameter variable.

[0133] Further, when the edge scene parameter determination module 303 is used to determine the edge scene parameter interval corresponding to the behavior parameter variable from the multiple parameter values ​​corresponding to the behavior parameter variable based on the parameter boundary value, the edge scene parameter determination module 303 is also used to:

[0134] Determine a plurality of target parameter values ​​adjacent to the parameter boundary value from the plurality of parameter values; wherein a ratio between the number of the plurality of target parameter values ​​and the number of the plurality of parameter values ​​reaches a preset threshold;

[0135] The smallest target parameter value among multiple target parameter values ​​is used as the first edge scene boundary parameter, the largest target parameter value among multiple target parameter values ​​is used as the second edge scene boundary parameter, and the edge scene parameter interval is generated according to the first edge scene boundary parameter and the second edge scene boundary parameter.

[0136] See also Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 4 As shown in , the electronic device 400 includes a processor 410 , a memory 420 and a bus 430 .

[0137] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 communicates with the memory 420 via the bus 430. When the machine-readable instructions are executed by the processor 410, the above-mentioned Figure 1 as well as Figure 2 The steps of the edge simulation scene generation method for autonomous driving testing in the method embodiment shown in the method embodiment and the specific implementation method can be found in the method embodiment, which will not be repeated here.

[0138] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the computer program can execute the above-mentioned Figure 1 as well as Figure 2 The steps of the edge simulation scene generation method for autonomous driving testing in the method embodiment shown in the method embodiment and the specific implementation method can be found in the method embodiment, which will not be repeated here.

[0139] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0140] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0141] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0142] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0143] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0144] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for generating edge simulation scenarios for autonomous driving tests, characterized in that: The edge simulation scene generation method comprises: Establishing a mathematical description model corresponding to the nominal behavior of the autonomous driving system, and extracting parameter values ​​corresponding to the behavior parameter variables in the mathematical description model from the autonomous driving data set; Constructing a marginal probability density function corresponding to each behavior parameter variable using a parameter value corresponding to each behavior parameter variable, and generating a parameter probability space of the nominal behavior based on a plurality of marginal probability density functions; For each behavior parameter variable, determine the edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and a preset edge probability threshold, and sample the parameter values ​​in the edge scene parameter interval to obtain multiple edge scene parameters corresponding to the behavior parameter variable; Combining parameters using edge scenario parameters corresponding to a plurality of behavior parameter variables to generate a plurality of target edge simulation scenarios corresponding to the nominal behavior; The determining, based on the parameter probability space and a preset edge probability threshold, an edge scene parameter interval corresponding to the behavior parameter variable comprises: Determine the conditional probability distribution of each parameter value corresponding to the behavior parameter variable according to the parameter probability space; Determining a parameter boundary value from a plurality of parameter values ​​corresponding to the behavior parameter variable based on the preset edge probability threshold and the conditional probability distribution of each parameter value; Based on the parameter boundary value, an edge scene parameter interval corresponding to the behavior parameter variable is determined from multiple parameter values ​​corresponding to the behavior parameter variable.

2. The edge simulation scene generation method according to claim 1, characterized in that: The method of constructing a marginal probability density function corresponding to each behavior parameter variable by using the parameter value corresponding to each behavior parameter variable, and generating a parameter probability space of the nominal behavior based on a plurality of marginal probability density functions, comprises: For each behavior parameter variable, fitting the marginal distribution of the parameter value corresponding to the behavior parameter variable to obtain the marginal probability density function corresponding to the behavior parameter variable; A joint probability density function between marginal probability density functions corresponding to a plurality of behavior parameter variables is established by using a Copula function, and the joint probability density function is used as the parameter probability space of the nominal behavior.

3. The edge simulation scene generation method according to claim 1, characterized in that: The step of determining the parameter boundary value from the multiple parameter values ​​corresponding to the behavior parameter variable based on the preset edge probability threshold and the conditional probability distribution of each parameter value includes: Determine a first probability threshold and a second probability threshold based on the preset edge probability threshold; Taking the parameter value whose conditional probability is equal to the first probability threshold among the multiple parameter values ​​as the first boundary value, and taking the parameter value whose conditional probability is equal to the second probability threshold among the multiple parameter values ​​as the second boundary value; The parameter boundary value is determined from the first boundary value and the second boundary value based on the boundary value selection rule corresponding to the behavior parameter variable.

4. The edge simulation scene generation method according to claim 1, characterized in that: The step of determining, based on the parameter boundary value, an edge scene parameter interval corresponding to the behavior parameter variable from a plurality of parameter values ​​corresponding to the behavior parameter variable comprises: Determine a plurality of target parameter values ​​adjacent to the parameter boundary value from the plurality of parameter values; wherein a ratio between the number of the plurality of target parameter values ​​and the number of the plurality of parameter values ​​reaches a preset threshold; The smallest target parameter value among multiple target parameter values ​​is used as the first edge scene boundary parameter, the largest target parameter value among multiple target parameter values ​​is used as the second edge scene boundary parameter, and the edge scene parameter interval is generated according to the first edge scene boundary parameter and the second edge scene boundary parameter.

5. An edge simulation scene generation device for autonomous driving test, characterized in that: The edge simulation scene generating device comprises: A parameter value acquisition module, used to establish a mathematical description model corresponding to the nominal behavior of the autonomous driving system, and extract parameter values ​​corresponding to the behavior parameter variables in the mathematical description model from the autonomous driving data set; A parameter probability space construction module, used to construct a marginal probability density function corresponding to each behavior parameter variable using a parameter value corresponding to each behavior parameter variable, and generate a parameter probability space of the nominal behavior based on a plurality of marginal probability density functions; An edge scene parameter determination module is used to determine, for each behavior parameter variable, an edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and a preset edge probability threshold, and sample parameter values ​​in the edge scene parameter interval to obtain a plurality of edge scene parameters corresponding to the behavior parameter variable; An edge simulation scenario generation module, used to combine parameters using edge scenario parameters corresponding to a plurality of behavior parameter variables to generate a plurality of target edge simulation scenarios corresponding to the nominal behavior; When the edge scene parameter determination module is used to determine the edge scene parameter interval corresponding to the behavior parameter variable based on the parameter probability space and the preset edge probability threshold, the edge scene parameter determination module is further used to: Determine the conditional probability distribution of each parameter value corresponding to the behavior parameter variable according to the parameter probability space; Determining a parameter boundary value from a plurality of parameter values ​​corresponding to the behavior parameter variable based on the preset edge probability threshold and the conditional probability distribution of each parameter value; Based on the parameter boundary value, an edge scene parameter interval corresponding to the behavior parameter variable is determined from multiple parameter values ​​corresponding to the behavior parameter variable.

6. The edge simulation scene generation device according to claim 5, characterized in that: When the parameter probability space construction module is used to construct the marginal probability density function corresponding to each behavior parameter variable using the parameter value corresponding to each behavior parameter variable, and generate the parameter probability space of the nominal behavior based on multiple marginal probability density functions, the parameter probability space construction module is also used to: For each behavior parameter variable, fitting the marginal distribution of the parameter value corresponding to the behavior parameter variable to obtain the marginal probability density function corresponding to the behavior parameter variable; A joint probability density function between marginal probability density functions corresponding to a plurality of behavior parameter variables is established by using a Copula function, and the joint probability density function is used as the parameter probability space of the nominal behavior.

7. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate through the bus. When the machine-readable instructions are run by the processor, the steps of the edge simulation scenario generation method for autonomous driving testing as described in any one of claims 1 to 4 are executed.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the edge simulation scenario generation method for autonomous driving testing as described in any one of claims 1 to 4 are executed.

Images