Method, apparatus, device and medium for estimating failure probability of autonomous driving system

Abstract

The present invention relates to the field of autonomous driving testing, and particularly to a method, apparatus, device and medium for estimating the failure probability of an autonomous driving system. The method includes: optimizing the probability density function of the scenario parameters with the optimization objective that the proportion of the initial scenario parameter samples with an estimated failure probability greater than the first preset failure probability threshold is greater than the preset proportion, to obtain the target probability density function; the estimated failure probability corresponding to the initial scenario parameter sample refers to the estimated probability of the target autonomous driving system failing under the initial scenario parameter sample; sampling based on the target probability density function within the preset range of scenario parameter values to obtain the target scenario parameter samples; estimating the failure probability of the target autonomous driving system based on the target probability density function and the estimated failure probabilities corresponding to all the target scenario parameter samples. This application can reduce the number of scenario parameter samples while ensuring the estimation accuracy, reduce the computational cost, and improve the estimation efficiency.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving tests. Specifically, it relates to a method, device, equipment and medium for estimating the failure probability of an autonomous driving system. Background Art

[0002] The failure probability of an autonomous driving system refers to the probability that the autonomous driving system fails to complete the predetermined tasks normally or malfunctions in a specific test scenario or actual operating environment. It is an important indicator to measure the safety and reliability of an autonomous driving system. Generally, the Monte Carlo method is used to estimate the failure probability of an autonomous driving system.

[0003] Due to the very strict requirements for the failure probability of an autonomous driving system, the estimation accuracy requirement for the failure probability is very high. And the estimation accuracy of the Monte Carlo method is inversely proportional to the square root of the sample size. Therefore, the number of sample scenario parameters required when using this method to estimate the failure probability is extremely large, resulting in too high a computational cost and low efficiency. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method, device, equipment and medium for estimating the failure probability of an autonomous driving system, which can sample based on the optimized probability density function, estimate the failure probability of the target autonomous driving system based on the obtained target scenario parameter samples, reduce the number of scenario parameter samples while ensuring the estimation accuracy, reduce the computational cost, and improve the estimation efficiency.

[0005] In a first aspect, an embodiment of this application provides a method for estimating the failure probability of an autonomous driving system, and the method includes:

[0006] Taking the proportion of the initial scenario parameter samples with an estimated failure probability greater than the first preset failure probability threshold being greater than the preset proportion as the optimization goal, optimize the probability density function of the scenario parameters to obtain the target probability density function; the estimated failure probability corresponding to the initial scenario parameter samples refers to the estimated probability that the target autonomous driving system fails under the initial scenario parameter samples;

[0007] Sample within the preset value range of the scenario parameters based on the target probability density function to obtain the target scenario parameter samples;

[0008] Estimate the failure probability of the target autonomous driving system based on the target probability density function and the estimated failure probabilities corresponding to all the target scenario parameter samples.

[0009] In a possible implementation manner, the step of sampling within the preset value range of the scenario parameters based on the target probability density function to obtain the target scenario parameter samples includes:

[0010] Taking the relative error of the first failure probability of the target autonomous driving system being less than a preset error threshold as the sampling target, sampling is performed based on the target probability density function within the value range of the scenario parameters to obtain the first scenario parameter sample in the target scenario parameter samples;

[0011] Sampling is performed based on the target probability density function within the value range of the scenario parameters to obtain the second scenario parameter sample in the target scenario parameter samples; the estimated failure probability corresponding to any of the second scenario parameter samples is greater than or equal to the random failure probability threshold corresponding to the second scenario parameter sample.

[0012] In a possible implementation manner, the estimating the failure probability of the target autonomous driving system based on the target probability density function and the estimated failure probabilities corresponding to all target scenario parameter samples includes:

[0013] Determining the first failure probability of the target autonomous driving system under the first scenario parameter sample set obtained after sampling as the second failure probability of the target autonomous driving system; the first scenario parameter sample set refers to the set of the first scenario parameter samples in the target scenario parameter samples;

[0014] Calculating a correction coefficient corresponding to the second failure probability according to the target probability density function and the estimated failure probabilities corresponding to all the second scenario parameter samples in the target scenario parameter samples;

[0015] Correcting the second failure probability of the target autonomous driving system based on the correction coefficient to obtain the target failure probability of the target autonomous driving system.

[0016] In a possible implementation manner, the calculating the correction coefficient corresponding to the second failure probability according to the target probability density function and the estimated failure probabilities corresponding to all second scenario parameter samples includes:

[0017] Substituting the target probability density function and the estimated failure probabilities corresponding to all second scenario parameter samples into the following formula to obtain the correction coefficient corresponding to the second failure probability;

[0018] ;

[0019] ;

[0020] where, is the correction coefficient, is the number of second scenario parameter samples, is the i-th second scenario parameter sample is the weight of, is the indicator function, The simulation value of the evaluation index under the i-th second scenario parameter sample obtained by simulating the target autonomous driving system under, The estimated failure probability corresponding to the i-th second scenario parameter sample is, The actual probability density function of the scenario parameters is the target probability density function.

[0021] In a possible implementation, the estimated failure probability corresponding to any initial scenario parameter sample, the estimated failure probability corresponding to any first scenario parameter sample, or the estimated failure probability corresponding to any second scenario parameter sample is calculated by the following formula:

[0022] ;

[0023] wherein, is the estimated failure probability corresponding to any initial scenario parameter sample, the estimated failure probability corresponding to any first scenario parameter sample, or the estimated failure probability corresponding to any second scenario parameter sample, is any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample, is the cumulative distribution function of the standard normal distribution function; is the predicted evaluation index distribution function of the target autonomous driving system under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample of the normal distribution expectation, is the predicted evaluation index distribution function of the target autonomous driving system under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample of the standard deviation.

[0024] In a possible implementation, the predicted evaluation index distribution function of the target autonomous driving system under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample is determined by the following steps:

[0025] The predicted evaluation index distribution function under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample is determined through a pre-constructed prediction model.

[0026] In a possible implementation, the prediction model is constructed by the following steps:

[0027] Hyper-covering sampling is performed within the preset scenario parameter value range through space-filling Latin hypercube sampling to obtain a set of simulation scenario parameter samples;

[0028] Simulate the target autonomous driving system according to the sample set of simulation scenario parameters to obtain the estimated evaluation index of each simulation scenario parameter sample in the sample set of simulation scenario parameters;

[0029] Perform Gaussian process regression modeling based on all simulation scenario parameter samples and the corresponding simulation scenario parameter samples to obtain a prediction model; the prediction model is used to determine the estimated evaluation index distribution function of the target autonomous driving system under any scenario parameter sample.

[0030] In a second aspect, an embodiment of the present application further provides a failure probability estimation device for an autonomous driving system, and the device includes:

[0031] An optimization module, configured to optimize the probability density function of the scenario parameters with the goal that the proportion of the initial scenario parameter samples with a failure probability greater than the first preset failure probability threshold is greater than the preset proportion, to obtain a target probability density function; the estimated failure probability corresponding to the initial scenario parameter sample refers to the probability that the target autonomous driving system fails under the initial scenario parameter sample;

[0032] A sampling module, including sampling based on the target probability density function within the preset range of scenario parameter values to obtain target scenario parameter samples;

[0033] An estimation module, including estimating the failure probability of the target autonomous driving system based on the target probability density function and the estimated failure probabilities corresponding to all target scenario parameter samples.

[0034] In a third aspect, an embodiment of the present application further provides an electronic device, including: a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the electronic device runs, the processor communicates with the storage medium through the bus, and the processor executes the machine-readable instructions to perform the steps of the failure probability estimation method for the autonomous driving system according to any item in the first aspect.

[0035] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it performs the steps of the failure probability estimation method for the autonomous driving system according to any item in the first aspect.

[0036] The embodiments of the present application provide a method, apparatus, device and medium for estimating the failure probability of an autonomous driving system. The method includes: optimizing the probability density function of scene parameters with the optimization objective that the proportion of initial scene parameter samples with an estimated failure probability greater than the first preset failure probability threshold is greater than the preset proportion, to obtain a target probability density function; the estimated failure probability corresponding to the initial scene parameter sample refers to the estimated probability that the target autonomous driving system fails under the initial scene parameter sample; sampling based on the target probability density function within the preset value range of scene parameters to obtain target scene parameter samples; estimating the failure probability of the target autonomous driving system based on the target probability density function and the estimated failure probabilities corresponding to all target scene parameter samples. The present application can sample based on the optimized probability density function, estimate the failure probability of the target autonomous driving system based on the sampled target scene parameter samples, reduce the number of scene parameter samples while ensuring the estimation accuracy, reduce the calculation cost, and improve the estimation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application, and thus should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0038] Figure 1 The flowchart of a method for estimating the failure probability of an autonomous driving system provided by an embodiment of the present application is shown;

[0039] Figure 2 The sampling flowchart of the target scene parameter samples provided by an embodiment of the present application is shown;

[0040] Figure 3 The structural schematic diagram of an apparatus for estimating the failure probability of an autonomous driving system provided by an embodiment of the present application is shown;

[0041] Figure 4 The structural schematic diagram of an electronic device provided by an embodiment of the present application is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, 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. It should be understood that the accompanying drawings in this application are only for the purpose of illustration and description, and are not used to limit the protection scope of this application. Additionally, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate the operations implemented according to some embodiments of this application. It should be understood that the operations in the flowchart may not be implemented in sequence, and steps without a logical context relationship may be reversed or implemented simultaneously. Moreover, those skilled in the art can add one or more other operations to the flowchart or remove one or more operations from the flowchart under the guidance of the content of this application.

[0043] In addition, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. The components of the embodiments of this application generally described and illustrated in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of this application claimed, but merely represents the selected embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of this application.

[0044] To enable those skilled in the art to use the content of this application, the following implementation manners are given in combination with a specific application scenario, the "autonomous driving test field". For those skilled in the art, without departing from the spirit and scope of this application, the general principles defined here can be applied to other embodiments and application scenarios. Although this application is mainly described around the "autonomous driving test field", it should be understood that this is only an exemplary embodiment.

[0045] It should be noted that the term "including" will be used in the embodiments of this application to indicate the existence of the features stated thereafter, but does not exclude the addition of other features.

[0046] The following provides a detailed description of a method for estimating the failure probability of an autonomous driving system provided by the embodiments of this application.

[0047] Refer to Figure 1 As shown, it is a flowchart of a method for estimating the failure probability of an autonomous driving system provided by the embodiments of this application. The following explains each step exemplarily in the embodiments of this application:

[0048] S101. Optimize the probability density function of the scenario parameters with the objective of the proportion of the initial scenario parameter samples with an estimated failure probability greater than the first preset failure probability threshold being greater than the preset proportion, to obtain the target probability density function.

[0049] In the embodiments of the present application, the estimated failure probability corresponding to the initial scenario parameter sample refers to the estimated probability of the target autonomous driving system failing under the initial scenario parameter sample. Each initial scenario parameter sample corresponds to an estimated failure probability, which is calculated based on a pre-constructed prediction model of the estimated evaluation index distribution function for determining the target autonomous driving system under any scenario parameter sample (such as the initial scenario parameter sample). The scenario parameters are variables defined to describe the driving scenarios of the target autonomous driving system, such as formal speed, whether there are pedestrians, vehicle position, etc. The scenario parameter sample is obtained by assigning values to the scenario parameters. The probability density function (PDF) of the scenario parameters refers to a function that describes the probability distribution of the scenario parameters near a certain determined value point in the continuous probability space.

[0050] Specifically, the probability density function of the scenario parameters is optimized through the following steps: sample based on the latest probability density function within the preset value range of the scenario parameters to obtain an initial scenario parameter sample set; determine whether the proportion of the initial scenario parameter samples in the initial scenario parameter sample set with an estimated failure probability greater than the first preset failure probability threshold (such as 0.5) is greater than the preset proportion; if the proportion of the initial scenario parameter samples with an estimated failure probability greater than the first preset failure probability threshold is less than the preset proportion (such as 20%), then update the latest probability density function based on the cross-entropy optimization method, and jump to "sample based on the latest probability density function within the preset value range of the scenario parameters to obtain an initial scenario parameter sample set" and continue to execute; if the proportion of the initial scenario parameter samples with an estimated failure probability greater than the first preset failure probability threshold is greater than or equal to the preset proportion, then determine the latest probability density function as the target probability density function.

[0051] Here, the probability density function of the initial scenario parameters is the actual probability density function of the scenario parameters. The probability density function of the initial scenario parameters conforms to the standard normal distribution. Therefore, in the normal distribution family, the probability density function of the scenario parameters can be optimized based on the cross-entropy optimization method. During the optimization process, only the expectation and variance of the probability density function are optimized, and the covariance is not updated to avoid local optimization and affect the optimization effect.

[0052] In addition, the estimated failure probability corresponding to the initial scenario parameter sample is determined through the following steps: calculate the estimated failure probability corresponding to the initial scenario parameter sample through the calculation formula of the estimated failure probability corresponding to the scenario parameter sample; when the number of initial scenario parameter samples with an estimated failure probability of 0 exceeds the preset number, use the normal distribution expectation of the estimated evaluation index distribution function of the target autonomous driving system under the initial scenario parameter sample as the final estimated failure probability corresponding to the initial scenario parameter sample.

[0053] S102. Sample based on the target probability density function within the preset range of scenario parameter values to obtain target scenario parameter samples.

[0054] Refer to Figure 2 As shown, it is the sampling flow chart of the target scenario parameter sample provided by the embodiment of the present application:

[0055] S201. With the relative error of the first failure probability of the target autonomous driving system being less than the preset error threshold as the sampling target, sample based on the target probability density function within the range of scenario parameter values to obtain the first scenario parameter sample in the target scenario parameter samples.

[0056] In the embodiment of the present application, the first failure probability of the target autonomous driving system refers to the failure probability of the target autonomous driving system under the "set of first scenario parameter samples after any sampling based on the sampling target", and is calculated through the estimated failure probabilities corresponding to all the first scenario parameter samples after any sampling. The set of first scenario parameter samples refers to the set of first scenario parameter samples. The relative error of the first failure probability of the target autonomous driving system is calculated based on the first failure probability of the target autonomous driving system and the estimated failure probabilities corresponding to all the first scenario parameter samples.

[0057] Specifically, the first scenario parameter sample in the target scenario parameter sample is obtained by sampling based on the target probability density function within the range of scenario parameter values through the following steps: sample based on the target probability density function within the range of scenario parameter values to obtain a set of first scenario parameter samples; calculate the relative error of the first failure probability of the target autonomous driving system based on this set of first scenario parameter samples; if the relative error is greater than or equal to the preset error threshold, continue to sample based on the target probability density function within the range of scenario parameter values to increase the number of first scenario parameter samples in the set of first scenario parameter samples; if the relative error is less than the preset error threshold, stop sampling.

[0058] Example: First, sample 100 first scenario parameter samples based on the target probability density function, calculate the relative error of the first failure probability of the target autonomous driving system based on the set composed of the 100 first scenario parameter samples. If the relative error is greater than or equal to the preset error threshold, sample 100 first scenario parameter samples again based on the target probability density function. Then calculate the relative error of the first failure probability of the target autonomous driving system based on the set composed of the 200 sampled first scenario parameter samples. If the relative error is less than the preset error threshold, stop sampling.

[0059] Furthermore, the calculation formulas for the first failure probability and the relative error of the first failure probability of the target autonomous driving system under the set of first scenario parameter samples after any sampling based on the sampling target are as follows:

[0060] ;

[0061] ;

[0062] Wherein, is the first failure probability of the target autonomous driving system, is the number of first scenario parameter samples , is the estimated failure probability corresponding to the i-th first scenario parameter sample , is the relative error of the first failure probability of the target autonomous driving system, is the variance of the estimated failure probabilities corresponding to all first scenario parameter samples .

[0063] Here, in the process of sampling the first scenario parameter samples in the embodiments of the present application, it does not rely on simulation. Therefore, the increase in the amount of target scenario parameter samples will not lead to an increase in the calculation cost, reducing the estimation complexity of the failure probability of the target autonomous driving system.

[0064] S202. Sample within the value range of the scenario parameters based on the target probability density function to obtain second scenario parameter samples in the target scenario parameter samples; the estimated failure probability corresponding to any second scenario parameter sample is greater than or equal to the random failure probability threshold corresponding to the second scenario parameter sample.

[0065] In the embodiment of the present application, a to-be-verified scenario parameter sample is obtained by sampling within the value range of scenario parameters based on the target probability density function; a random failure probability threshold is randomly matched for the to-be-verified scenario parameter sample within the preset failure probability value range ([0, 1]) to obtain the random failure probability threshold corresponding to the to-be-verified scenario parameter sample; if the estimated failure probability corresponding to the to-be-verified scenario parameter sample is less than the random failure probability threshold, the to-be-verified scenario parameter sample is discarded; if the estimated failure probability corresponding to the to-be-verified scenario parameter sample is greater than or equal to the random failure probability threshold, the to-be-verified scenario parameter sample is determined as the second scenario parameter sample; if the number of second scenario parameter samples is less than the preset sample number, sampling continues within the value range of scenario parameters based on the target probability density function; if the number of second scenario parameter samples is equal to the preset sample number, sampling stops.

[0066] Here, in the process of sampling the second scenario parameter sample in the embodiment of the present application, it does not depend on simulation, so the sampling of the target scenario parameter sample size will not cause an increase in the calculation cost, reducing the estimation complexity of the failure probability of the autonomous driving system. Sampling the second scenario parameter sample by the acceptance-rejection sampling method ensures the unbiasedness of the estimation. Optimizing the probability density function results in the failure probability corresponding to the sample not being too low, thereby improving the sampling efficiency.

[0067] S103. Estimate the failure probability of the target autonomous driving system based on the target probability density function and the estimated failure probabilities corresponding to all target scenario parameter samples.

[0068] Step 1. Determine the first failure probability of the target autonomous driving system under the first scenario parameter sample set obtained after sampling as the second failure probability of the target autonomous driving system; the first scenario parameter sample set refers to the set of first scenario parameter samples in the target scenario parameter samples.

[0069] Step 2. Calculate the correction coefficient corresponding to the second failure probability according to the target probability density function and the estimated failure probabilities corresponding to all second scenario parameter samples in the target scenario parameter samples.

[0070] Specifically, substitute the target probability density function and the estimated failure probabilities corresponding to all second scenario parameter samples into the following formula to obtain the correction coefficient corresponding to the second failure probability;

[0071] ;

[0072] ;

[0073] Among them, is the correction coefficient, is the number of second scenario parameter samples, is the weight of the i-th second scenario parameter sample , is an indicator function is the simulation value of the evaluation index obtained by simulating the target autonomous driving system under the i-th second scenario parameter sample . is the estimated failure probability corresponding to the i-th second scenario parameter sample . is the actual probability density function of the scenario parameter is the target probability density function

[0074] Here, since the target autonomous driving system needs to be simulated based on the second scenario parameter sample when calculating the correction coefficient, in order to avoid the increase in calculation cost caused by simulation, the number of second scenario parameter samples is reasonably designed as the preset sample number

[0075] Step 3: Correct the second failure probability of the target autonomous driving system based on the correction coefficient to obtain the target failure probability of the target autonomous driving system

[0076] In the embodiment of the present application, the product of the correction coefficient and the second failure probability is determined as the target failure probability of the target autonomous driving system

[0077] Furthermore, the embodiment of the present application provides a calculation formula for the estimated failure probability of the scenario parameter sample. Specifically, the estimated failure probability corresponding to any initial scenario parameter sample, the estimated failure probability corresponding to any first scenario parameter sample, or the estimated failure probability corresponding to any second scenario parameter sample is calculated through the following formula

[0078] ;

[0079] where is the estimated failure probability corresponding to any initial scenario parameter sample, the estimated failure probability corresponding to any first scenario parameter sample, or the estimated failure probability corresponding to any second scenario parameter sample is any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample is the cumulative distribution function of the standard normal distribution function is the expected value of the normal distribution of the predicted evaluation index distribution function of the target autonomous driving system under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample is the standard deviation of the predicted evaluation index distribution function of the target autonomous driving system under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample ​​

[0080] Here, the cumulative distribution function refers to the probability that a random variable is less than or equal to a certain value, which can be understood as the probability that a standard normal distribution random variable is less than or equal to .

[0081] Furthermore, the predicted evaluation index distribution function of the target autonomous driving system under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample is determined through the following steps: The predicted evaluation index distribution function under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample is determined through a pre-constructed prediction model.

[0082] In the embodiment of the present application, overcovering sampling is performed within the preset range of scenario parameter values through the space-filling Latin hypercube sampling method to obtain a set of simulation scenario parameter samples; the target autonomous driving system is simulated according to the set of simulation scenario parameter samples to obtain the predicted evaluation index of each simulation scenario parameter sample in the set of simulation scenario parameter samples; Gaussian process regression modeling is performed based on all the simulation scenario parameter samples and the corresponding predicted evaluation indexes to obtain a prediction model; the prediction model is used to determine the predicted evaluation index distribution function of the target autonomous driving system under any scenario parameter sample.

[0083] Furthermore, the embodiment of the present application can also calculate the accuracy of the target failure probability. Specifically, the accuracy of the target failure probability is calculated through the following formula:

[0084] ;

[0085] where, is the second failure probability, is the variance of the estimated failure probabilities corresponding to all the first scenario parameter samples when calculating the second failure probability , is the variance of the estimated failure probabilities corresponding to all the second scenario parameter samples when calculating the correction coefficient .

[0086] Furthermore, the process of estimating the target failure probability in the embodiment of the present application is elaborated in detail through the foregoing content. Next, the principle of the target failure probability in the embodiment of the present application will be elaborated through the following content:

[0087] (1) The return value of the prediction model obtained through Gaussian process regression modeling is the predicted evaluation index distribution function of the target autonomous driving system under any scenario parameter sample , that is, the normal distribution expectation and the standard deviation . At this time, obeys At this time, the estimated failure probability corresponding to any scenario parameter sample can be defined:

[0088] ;

[0089] where, is used to represent the situation where the estimated target autonomous driving system fails under , is the probability that the estimated target autonomous driving system fails under .

[0090] (2) Let be a random variable independent of X and uniformly distributed on 0 - 1. Let be the probability density function of the scenario parameter, and is equivalent to . Define the first failure probability under :

[0091] ;

[0092] where the random matching failure probability threshold is 's value. represents the probability expectation under the probability density function . Let the weight function be:

[0093]

[0094] Then the correction coefficient can be defined as:

[0095] ;

[0096] where represents 's conditional expectation under the condition when follows the density function

[0097]

[0098]

[0099]

[0100]

[0101]

[0102] ;

[0103] Therefore: . Among them, is the scenario parameter sample for the scenario parameter X and the value range of

[0104] Here, (1) the independent variable follows the standard normal distribution and is independent among all dimensions, that is The density function of is the standard normal density function of the specified dimension. For the logical scenarios in general cases, through distribution transformation, the normal distribution can be transformed into the real distribution space, so this solution has generality. (2) Taking as the failure scenario, that is, to estimate , represents the probability of the autonomous driving system under the probability density function of . represents the estimated value of the evaluation index of the autonomous driving system under (which can be the safety index TTC). (3) In addition to the Gaussian process regression model, Bayesian neural networks, probabilistic support vector machines, etc. can also be used as prediction models. In the embodiments of this application, the Gaussian process regression model is used, that is, through to estimate , is the simulation failure probability of the target autonomous driving system obtained by simulation under .

[0105] The embodiments of this application provide a method for estimating the failure probability of an autonomous driving system. The method includes: taking the proportion of the initial scenario parameter samples with the estimated failure probability greater than the first preset failure probability threshold being greater than the preset proportion as the optimization goal, optimizing the probability density function of the scenario parameters to obtain the target probability density function; the estimated failure probability corresponding to the initial scenario parameter sample refers to the estimated probability of the target autonomous driving system failing under the initial scenario parameter sample; sampling based on the target probability density function within the preset value range of the scenario parameters to obtain the target scenario parameter samples; estimating the failure probability of the target autonomous driving system based on the target probability density function and the estimated failure probabilities corresponding to all the target scenario parameter samples. This application can sample based on the optimized probability density function, estimate the failure probability of the target autonomous driving system based on the obtained target scenario parameter samples, while ensuring the estimation accuracy, reducing the number of scenario parameter samples, reducing the calculation cost, and improving the estimation efficiency.

[0106] Based on the same inventive concept, an embodiment of the present application further provides a failure probability estimation device for an autonomous driving system corresponding to the failure probability estimation method of the autonomous driving system. Since the principle of the device in the embodiment of the present application to solve problems is similar to the above-mentioned failure probability estimation method of the autonomous driving system in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be elaborated.

[0107] Referring Figure 3 as shown in the figure, it is a schematic diagram of a failure probability estimation device for an autonomous driving system provided by an embodiment of the present application. The device includes:

[0108] An optimization module 301, configured to optimize the probability density function of the scenario parameters with the proportion of the initial scenario parameter samples with a failure probability greater than the first preset failure probability threshold being greater than the preset proportion, to obtain a target probability density function; the estimated failure probability corresponding to the initial scenario parameter samples refers to the probability that the target autonomous driving system fails under the initial scenario parameter samples;

[0109] A sampling module 302, including sampling based on the target probability density function within a preset value range of the scenario parameters to obtain target scenario parameter samples;

[0110] An estimation module 303, including estimating the failure probability of the target autonomous driving system based on the target probability density function and the estimated failure probabilities corresponding to all target scenario parameter samples.

[0111] Through the failure probability estimation device for an autonomous driving system provided by the present application, it is possible to sample based on the optimized probability density function, estimate the failure probability of the target autonomous driving system based on the obtained target scenario parameter samples, reduce the number of scenario parameter samples while ensuring the estimation accuracy, reduce the calculation cost, and improve the estimation efficiency.

[0112] As Figure 4 shown, an electronic device 400 provided by an embodiment of the present application includes: a processor 401, a memory 402, and a bus. The memory 402 stores machine-readable instructions executable by the processor 401. When the electronic device runs, the processor 401 communicates with the memory 402 through the bus, and the processor 401 executes the machine-readable instructions to perform the steps of the failure probability estimation method of the autonomous driving system as described above.

[0113] Specifically, the above-mentioned memory 402 and processor 401 can be general memories and processors, which are not specifically limited here. When the processor 401 runs the computer program stored in the memory 402, it can execute the above-mentioned failure probability estimation method of the autonomous driving system.

[0114] Corresponding to the method for estimating the failure probability of the above-mentioned autonomous driving system, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it executes the steps of the method for estimating the failure probability of the above-mentioned autonomous driving system.

[0115] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems and devices described above can refer to the corresponding processes in the method embodiments, which will not be elaborated in this application. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the modules is only a logical function division, and there can be other division methods in actual implementation. For another example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed with each other can be through some communication interfaces. The indirect couplings or communication connections of the devices or modules can be in electrical, mechanical, or other forms.

[0116] The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0117] In addition, in each embodiment of the present application, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0118] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the information processing method described in each embodiment of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, ROM, RAM, magnetic disks, or optical discs that can store program codes.

[0119] The above are only the specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A method for estimating the failure probability of an autonomous driving system, characterized in that, The method includes: Optimizing the probability density function of the scenario parameters with the objective of making the proportion of initial scenario parameter samples with an estimated failure probability greater than a first preset failure probability threshold greater than a preset proportion, to obtain a target probability density function; the estimated failure probability corresponding to the initial scenario parameter sample refers to the estimated probability of the target autonomous driving system failing under the initial scenario parameter sample; Sampling based on the target probability density function within a preset range of scenario parameter values to obtain target scenario parameter samples; Determining the first failure probability of the target autonomous driving system under a first set of scenario parameter samples obtained after sampling as the second failure probability of the target autonomous driving system; the first set of scenario parameter samples refers to the set of first scenario parameter samples among the target scenario parameter samples; calculating a correction coefficient corresponding to the second failure probability according to the target probability density function and the estimated failure probabilities corresponding to all second scenario parameter samples in the target scenario parameter samples; correcting the second failure probability of the target autonomous driving system based on the correction coefficient to obtain the target failure probability of the target autonomous driving system.

2. The method for estimating the failure probability of the autonomous driving system according to claim 1, wherein The sampling based on the target probability density function within a preset range of scenario parameter values to obtain target scenario parameter samples includes: With the relative error of the first failure probability of the target autonomous driving system being less than a preset error threshold as the sampling objective, sampling based on the target probability density function within the range of scenario parameter values to obtain the first scenario parameter samples in the target scenario parameter samples; Sampling based on the target probability density function within the range of scenario parameter values to obtain the second scenario parameter samples in the target scenario parameter samples; the estimated failure probability corresponding to any one of the second scenario parameter samples is greater than or equal to the random failure probability threshold corresponding to the second scenario parameter sample.

3. The method for estimating the failure probability of the autonomous driving system according to claim 1, wherein The calculating the correction coefficient corresponding to the second failure probability according to the target probability density function and the estimated failure probabilities corresponding to all second scenario parameter samples includes: Substituting the target probability density function and the estimated failure probabilities corresponding to all second scenario parameter samples into the following formula to obtain the correction coefficient corresponding to the second failure probability; ； ； Among them, is a correction coefficient, is the number of second scenario parameter samples, is the i-th second scenario parameter sample weight, is an indicator function, is the simulation value of the evaluation index obtained by simulating the target autonomous driving system under the i-th second scenario parameter sample below, is the i-th second scenario parameter sample corresponding estimated failure probability, is the actual probability density function of the scenario parameter, is the target probability density function.

4. The method for estimating the failure probability of the autonomous driving system according to claim 1, wherein Calculating the estimated failure probability corresponding to any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample through the following formula: ； wherein, is the estimated failure probability corresponding to any initial scenario parameter sample, the estimated failure probability corresponding to any first scenario parameter sample, or the estimated failure probability corresponding to any second scenario parameter sample, is any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample, is the cumulative distribution function of the standard normal distribution function; is the estimated evaluation index distribution function of the target autonomous driving system under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample is the normal distribution expectation of is the estimated evaluation index distribution function of the target autonomous driving system under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample is the standard deviation of 5. The method for estimating the failure probability of the automatic driving system according to claim 4, wherein Determining the predicted evaluation index distribution function of the target autonomous driving system under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample through the following steps: Determining the predicted evaluation index distribution function of the target autonomous driving system under any initial scenario parameter sample, any first scenario parameter sample, or any second scenario parameter sample through a pre-constructed prediction model.

6. The method for estimating the failure probability of the automatic driving system according to claim 5, wherein, Constructing the prediction model through the following steps: Performing over-coverage sampling within a preset range of scenario parameter values by means of space-filling Latin hypercube sampling to obtain a set of simulation scenario parameter samples; Simulate the target autonomous driving system according to the simulation scenario parameter sample set to obtain the estimated evaluation index of each simulation scenario parameter sample in the simulation scenario parameter sample set; Based on all simulation scenario parameter samples and the corresponding simulation scenario parameter samples, perform Gaussian process regression modeling to obtain a prediction model; the prediction model is used to determine the estimated evaluation index distribution function of the target autonomous driving system under any scenario parameter sample.

7. A failure probability estimation device for an autonomous driving system, characterized in that, The device includes: An optimization module, configured to optimize the probability density function of the scenario parameters with the proportion of the initial scenario parameter samples with a failure probability greater than the first preset failure probability threshold being greater than the preset proportion, to obtain a target probability density function; the estimated failure probability corresponding to the initial scenario parameter sample refers to the probability that the target autonomous driving system fails under the initial scenario parameter sample; A sampling module, including sampling based on the target probability density function within the preset value range of the scenario parameters to obtain target scenario parameter samples; An estimation module, including determining the first failure probability of the target autonomous driving system under the first scenario parameter sample set obtained after sampling as the second failure probability of the target autonomous driving system; the first scenario parameter sample set refers to the set of the first scenario parameter samples in the target scenario parameter samples; calculating a correction coefficient corresponding to the second failure probability according to the target probability density function and the estimated failure probabilities corresponding to all the second scenario parameter samples in the target scenario parameter samples; and correcting the second failure probability of the target autonomous driving system based on the correction coefficient to obtain the target failure probability of the target autonomous driving system.

8. An electronic device, characterized in that, Includes: A processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the electronic device runs, the processor communicates with the storage medium through the bus, and the processor executes the machine-readable instructions to perform the steps of the method for estimating the failure probability of the autonomous driving system according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is run by the processor, it performs the steps of the method for estimating the failure probability of the autonomous driving system according to any one of claims 1 to 6.

Images