Vehicle function failure evaluation method and system and vehicle

Abstract

The embodiment of the invention provides a vehicle function failure assessment method and system and a vehicle, and relates to the technical field of function failure probability assessment and automatic driving. The method comprises the steps that initial scene data corresponding to a vehicle are acquired, and the initial scene data are used for representing a real-time scene where the vehicle is located; performing parameter dimension reduction processing on the initial scene data to obtain target scene data; based on target failure probability distribution, adaptive sampling is carried out on the target scene data, a target sampling result is obtained, and the target failure probability distribution is used for representing scene information corresponding to a function failure event of the vehicle; and performing evaluation calculation according to the target sampling result to obtain the function failure probability of the vehicle in the real-time scene. According to the method and the device, the technical problems of low efficiency and low accuracy of a vehicle function failure evaluation method caused by dependence on a Monte Carlo method in related technologies are solved.

Description

Technical Field

[0001] The embodiments of the present application relate to the fields of functional failure probability assessment technology and autonomous driving technology, and in particular to a vehicle functional failure assessment method, system, and vehicle. Background Art

[0002] As vehicle functions become increasingly complex, assessing the probability of vehicle failure is becoming increasingly important. Assessing the probability of vehicle failure has become a pressing issue in related technical fields.

[0003] Related technologies typically rely on the Monte Carlo method to randomly combine parameters in vehicle driving scenarios, generating a large number of random scenario samples. These random scenario samples are then sampled to assess the probability of vehicle functional failure. However, due to the complexity and variability of driving scenarios and the large number of parameters involved, the Monte Carlo method requires simulating all possible parameter combinations, resulting in low efficiency in assessing vehicle functional failure. Furthermore, the Monte Carlo method uses a fixed sampling distribution, making it difficult to accurately simulate vehicle functional failure events, resulting in low accuracy.

[0004] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0005] The embodiments of the present application provide a vehicle functional failure assessment method, system, and vehicle, aiming to improve the technical problem in the related art of relying on the Monte Carlo method, which leads to low efficiency and low accuracy of the vehicle functional failure assessment method.

[0006] According to one aspect of an embodiment of the present application, a method for functional failure assessment of a vehicle is provided, including: obtaining initial scene data corresponding to the vehicle, wherein the initial scene data is used to characterize a real-time scene in which the vehicle is located; performing parameter dimensionality reduction processing on the initial scene data to obtain target scene data; based on a target failure probability distribution, adaptively sampling the target scene data to obtain a target sampling result, wherein the target failure probability distribution is used to characterize scene information corresponding to a functional failure event in the vehicle; performing an evaluation calculation based on the target sampling result to obtain a functional failure probability of the vehicle in the real-time scene, wherein the functional failure probability is used to evaluate the confidence of the target function of the vehicle in the real-time scene.

[0007] The functional failure assessment method for the above-mentioned vehicle provided in the embodiment of the present application achieves the following technical effects: by obtaining the initial scene data corresponding to the vehicle and performing parameter dimensionality reduction processing on the initial scene data, the initial scene data corresponding to the real-time scene in which the vehicle is located can be simplified to obtain target scene data with lower parameter dimensions, which helps to obtain target sampling results more quickly in the future; further, based on the target failure probability distribution, the target scene data is adaptively sampled, so that the sampling process can focus more on the functional failure events of the vehicle, and can find the functional failure events of the vehicle more quickly and accurately, thereby obtaining the target sampling results more quickly and accurately; further, the evaluation calculation is performed based on the target sampling results, which can improve the evaluation efficiency and accuracy of the functional failure probability. Therefore, the embodiment of the present application achieves the purpose of combining parameter dimensionality reduction processing and adaptive sampling to efficiently and accurately evaluate the functional failure probability of the vehicle in the real-time scene, achieves the technical effect of improving the evaluation efficiency and accuracy of the functional failure probability, and solves the technical problem of low efficiency and low accuracy of the vehicle functional failure assessment method in the related art due to reliance on the Monte Carlo method.

[0008] Optionally, the initial scenario data includes multiple scenario parameters, and parameter dimensionality reduction processing is performed on the initial scenario data to obtain the target scenario data, including: performing variance decomposition calculation on the multiple scenario parameters to obtain failure sensitive data corresponding to the multiple scenario parameters, wherein the failure sensitive data is used to evaluate the degree of influence of parameter changes of the multiple scenario parameters on the probability of functional failure; based on the failure sensitive data, selecting some scenario parameters from the multiple scenario parameters; and using the some scenario parameters to generate the target scenario data.

[0009] The above-mentioned optional embodiments of the present application can achieve the following technical effects: by performing variance decomposition calculations on multiple scenario parameters, the degree of influence of different scenario parameters on the probability of functional failure is quantified, and some scenario parameters that have a greater influence on the probability of functional failure are screened out. On the one hand, it solves the problem of waste of computing resources caused by too many parameters in related technologies, and improves the efficiency of functional failure assessment of vehicles. On the other hand, it screens out some scenario parameters that have a greater influence on the probability of functional failure, and only needs to perform functional failure assessment of the vehicle for these scenario parameters, avoiding the problem of low accuracy of evaluation results due to redundant analysis of irrelevant parameters, and improving the accuracy of functional failure assessment of vehicles.

[0010] Optionally, the failure sensitive data includes multiple sensitive indices, and variance decomposition calculation is performed on multiple scene parameters to obtain failure sensitive data corresponding to the multiple scene parameters respectively, including: calculating the multiple scene parameters based on the variance sensitive Sobol index method to obtain the overall variance and the decomposition variance corresponding to the multiple scene parameters respectively; for the target scene parameter among the multiple scene parameters, according to the overall variance and the decomposition variance corresponding to the target scene parameter, obtaining the target sensitive index corresponding to the target scene parameter among the multiple sensitive indices.

[0011] The above-mentioned optional embodiments of the present application can achieve the following technical effects: multiple scenario parameters are calculated based on the variance-sensitive Sobol index method. In the process of quantifying the degree of influence of scenario parameters on the probability of vehicle functional failure, not only the direct influence of a single scenario parameter on the probability of vehicle functional failure can be considered, but also the indirect influence of the interaction effect between the scenario parameter and other scenario parameters on the probability of vehicle functional failure can be considered. It can more comprehensively consider the degree of influence of parameter changes of scenario parameters on the probability of functional failure, obtain a more accurate sensitivity index, and thus improve the accuracy of failure-sensitive data.

[0012] Optionally, based on the target failure probability distribution, the target scene data is adaptively sampled to obtain the target sampling result, including: sampling the target scene data according to the initial sampling distribution to obtain the initial sampling result; adaptively iteratively updating the initial sampling result according to the target failure probability distribution to obtain the target sampling result, wherein the similarity between the sampling distribution of the target sampling result and the target failure probability distribution meets the target condition.

[0013] The above-mentioned optional embodiments of the present application can achieve the following technical effects: utilizing adaptive sampling technology, by iteratively updating the sampling distribution, dynamically adjusting the sampling strategy, so that during the sampling process, more attention can be paid to the functional failure events of the vehicle (especially, rare functional failure events), and the functional failure events of the vehicle can be captured more quickly and accurately, saving computing resources, speeding up the speed of obtaining the target sampling results, and improving the accuracy of the target sampling results, thereby laying a data foundation for the subsequent efficient and accurate calculation of the functional failure probability.

[0014] Optionally, the initial sampling results are adaptively iteratively updated according to the target failure probability distribution to obtain the target sampling results, including: in the current iteration round of the adaptive iterative update process, using the previous round sampling results corresponding to the previous iteration round and the target failure probability distribution, updating the previous round sampling distribution corresponding to the previous round sampling results to the current sampling distribution; sampling the target scene data according to the current sampling distribution to obtain the current sampling result corresponding to the current iteration round.

[0015] The above-mentioned optional embodiments of the present application can achieve the following technical effects: by iteratively optimizing the sampling distribution, the sampling results after each round of iteration can be closer to the expected results, thereby more quickly making the similarity between the sampling distribution of the target sampling results and the target failure probability distribution meet the target conditions, overcoming the defect of the mismatch between the sampling distribution and the failure probability distribution in the traditional sampling method, and realizing efficient and accurate evaluation of the vehicle function failure probability.

[0016] Optionally, using the previous round sampling results and the target failure probability distribution, updating the previous round sampling distribution to the current sampling distribution includes: using the target failure probability distribution to perform weight calculations on multiple sampling samples in the previous round sampling results to obtain sample weights corresponding to the multiple sampling samples; selecting target samples from the multiple sampling samples based on the multiple sample weights; and performing distribution fitting processing on the target samples to obtain the current sampling distribution corresponding to the current iteration round.

[0017] The above-mentioned optional embodiments of the present application can achieve the following technical effects: by calculating the sample weights corresponding to multiple sampling samples, the target samples that have a greater impact on the probability of calculation function failure can be accurately identified, so that the current sampling distribution gradually approaches the area that contributes more to the probability of calculation function failure, avoiding the problem of invalid sampling caused by using a fixed sampling distribution, thereby improving the calculation efficiency and accuracy of subsequent function failure probabilities.

[0018] Optionally, the vehicle functional failure assessment method further includes: obtaining test case data corresponding to the vehicle, wherein the test case data is used to characterize the test scenario in which the vehicle is located; performing scenario testing on the test case data to obtain functional performance data of the vehicle, wherein the functional performance data is used to assess the failure probability of the target function of the vehicle in the test scenario; and constructing a target failure probability distribution using the test case data and functional performance data.

[0019] The above-mentioned optional embodiments of the present application can achieve the following technical effects: obtaining test case data corresponding to the vehicle, which can be collected through actual scenarios or obtained through simulation scenarios. By combining actual scenarios and simulation scenarios, the vehicle's test scenarios can be considered more comprehensively, and then functional performance data can be obtained by performing scenario testing on the test case data, ensuring the reliability and accuracy of the constructed target failure probability distribution, providing a basis for subsequent adaptive sampling, and avoiding the problem of low accuracy of failure probability assessment results due to failure to consider rare failure events.

[0020] Optionally, performing an evaluation calculation based on the target sampling results to obtain the functional failure probability of the vehicle in a real-time scenario includes: using the functional performance data corresponding to the target failure probability distribution to determine the failure indication data corresponding to the target sampling results; and performing calculations using the failure indication data, the target failure probability distribution, and the sampling distribution of the target sampling results to obtain the functional failure probability.

[0021] The above-mentioned optional embodiments of the present application can achieve the following technical effects: by utilizing the functional performance data corresponding to the target failure probability distribution to determine the failure indication data corresponding to the target sampling result, it can more accurately reflect the actual performance of the vehicle's target function in various scenarios (especially rare failure scenarios); by combining the failure indication data, the target failure probability distribution and the sampling distribution of the target sampling results, it can more accurately evaluate and calculate the functional failure probability, thereby improving the accuracy of the functional failure probability.

[0022] According to another aspect of an embodiment of the present application, a vehicle functional failure assessment system is also provided, including: an acquisition unit for acquiring initial scene data corresponding to the vehicle, wherein the initial scene data is used to characterize the real-time scene in which the vehicle is located; a processing unit for performing parameter dimensionality reduction processing on the initial scene data to obtain target scene data; a sampling unit for adaptively sampling the target scene data based on a target failure probability distribution to obtain a target sampling result, wherein the target failure probability distribution is used to characterize the distribution of occurrence scene information corresponding to a functional failure event of the vehicle; an evaluation unit for performing evaluation calculations based on the target sampling results to obtain the functional failure probability of the vehicle in the real-time scene.

[0023] The functional failure assessment system for the above-mentioned vehicle provided in the embodiment of the present application achieves the following technical effects: utilizing an acquisition unit to acquire initial scene data corresponding to the vehicle, utilizing a processing unit to perform parameter dimensionality reduction processing on the initial scene data, thereby being able to simplify the initial scene data corresponding to the real-time scene in which the vehicle is located, and obtain target scene data with lower parameter dimensions, which helps to obtain target sampling results more quickly in the subsequent process; further, utilizing a sampling unit to adaptively sample target scene data based on a target failure probability distribution, so that the sampling process can focus more on functional failure events of the vehicle, and can find functional failure events of the vehicle more quickly and accurately, thereby obtaining target sampling results more quickly and accurately; further, utilizing an assessment unit to perform assessment calculations based on the target sampling results, thereby improving the assessment efficiency and accuracy of the functional failure probability, enhancing the system's performance in performing efficient and accurate functional failure assessments on vehicles, and providing better functional failure assessment services.

[0024] According to another aspect of an embodiment of the present application, a vehicle is also provided, including an on-board processor and an on-board memory, wherein the on-board memory is used to store computer programs; and the on-board processor is used to execute the computer programs stored in the memory to implement any one of the above-mentioned vehicle functional failure assessment methods.

[0025] The above-mentioned vehicle provided by the embodiment of the present application achieves the following technical effects: by obtaining the initial scene data corresponding to the vehicle and performing parameter dimensionality reduction processing on the initial scene data, the initial scene data corresponding to the real-time scene in which the vehicle is located can be simplified to obtain target scene data with lower parameter dimensions, which helps to obtain target sampling results more quickly in the future; further, based on the target failure probability distribution, the target scene data is adaptively sampled, so that the sampling process can focus more on the functional failure events of the vehicle, and can find the functional failure events of the vehicle more quickly and accurately, thereby obtaining the target sampling results more quickly and accurately; further, evaluation and calculation are performed based on the target sampling results, which can improve the evaluation efficiency and accuracy of the functional failure probability, enhance the safety of the vehicle, and thus improve the overall performance of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of a vehicle functional failure assessment method provided by an embodiment of the present application;

[0027] Figure 2 is a schematic diagram of a vehicle functional failure assessment method provided by an embodiment of the present application;

[0028] Figure 3 This is a structural block diagram of a vehicle function failure assessment system provided by an embodiment of the present application;

[0029] Figure 4 This is a structural block diagram of a vehicle provided by an embodiment of the present application;

[0030] Figure 5 This is a hardware structure block diagram of a computing terminal provided in one embodiment of the present application;

[0031] Figure 6 This is a structural block diagram of a vehicle function failure assessment device provided by an embodiment of the present application;

[0032] Figure 7 This is a structural block diagram of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to make the technical problems, technical solutions and beneficial effects solved by this application more clearly understood, this application is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0034] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0035] Example 1

[0036] This application embodiment provides a model training method, please refer to Figure 1 , including the following steps:

[0037] S10: Acquire initial scene data corresponding to the vehicle, wherein the initial scene data is used to represent the real-time scene in which the vehicle is located;

[0038] S20: Perform parameter dimensionality reduction processing on the initial scene data to obtain target scene data;

[0039] S30: Adaptively sampling target scenario data based on the target failure probability distribution to obtain a target sampling result, wherein the target failure probability distribution is used to characterize scenario information corresponding to a functional failure event of the vehicle;

[0040] S40: performing evaluation calculations based on the target sampling results to obtain a functional failure probability of the vehicle in a real-time scenario, wherein the functional failure probability is used to evaluate the confidence level of the target function of the vehicle in the real-time scenario.

[0041] The above-mentioned initial scene data can be obtained by analyzing and converting the driving data of the real-time scene. The above-mentioned driving data of the real-time scene may include vehicle data and environmental data. The above-mentioned vehicle data may include but is not limited to: vehicle type, vehicle speed, sensor accuracy data (such as perception delay information), vehicle visual range data, and control algorithm accuracy data (such as algorithm response speed). The above-mentioned environmental data may include but is not limited to: data of surrounding vehicles (such as the speed of surrounding vehicles, the acceleration of surrounding vehicles, the distance between surrounding vehicles, etc.), and climate information data (such as temperature, humidity, light intensity, etc.).

[0042] In an exemplary application scenario, Figure 2As shown, by analyzing the driving data of the real-time scene, the driving data under different reference dimensions can be converted to the same reference dimension to obtain the initial scene data corresponding to the vehicle. The initial scene data may include functional layer scene data, logical layer scene data, and specific layer scene data. The initial scene data is used to characterize the real-time scene in which the vehicle is located, which is helpful for the subsequent evaluation of the functional failure probability of the vehicle's target function in the real-time scene in which it is located from multiple reference dimensions.

[0043] The above-mentioned functional layer scenario data is used to describe the target function used by the vehicle in the driving scenario. The functional layer scenario data may include but is not limited to: the type of the target function, the functional description data of the target function. The above-mentioned logical layer scenario data is used to describe the decision-making process of the vehicle in executing the target function in the driving scenario. The logical layer scenario data may include but is not limited to: the execution logic data of the target function, the failure judgment logic data of the target function. The above-mentioned specific layer scenario data is used to describe the parameters of the vehicle in the driving scenario. The specific layer scenario data may include but is not limited to: data corresponding to the vehicle systems involved in the target function (such as sensor systems, automatic driving systems, communication systems, etc.), environmental data of the driving scene (such as temperature, humidity, slope, altitude, etc.), and the vehicle's own behavior data.

[0044] The target scene data can be obtained using parameter dimensionality reduction methods, which may include but are not limited to linear dimensionality reduction methods (e.g., principal component analysis methods), factor analysis methods, feature selection methods based on least squares and support vector machines, and sensitivity analysis methods.

[0045] Parameter dimensionality reduction methods can be used to select scenario parameters with the greatest impact on the probability of functional failure from the initial scenario data, thereby performing parameter dimensionality reduction on the initial scenario data and obtaining the target scenario data. Specifically, the parameter influence factor of a scenario parameter can be compared with an influence factor threshold. If the parameter influence factor of the scenario parameter is greater than the influence factor threshold, the scenario parameter is considered to have a greater impact on the probability of functional failure.

[0046] It is easy to understand that by performing parameter dimensionality reduction processing on the initial scene data corresponding to the acquired vehicle, the initial scene data can be simplified, and the scene parameters with a greater impact on the probability of functional failure in the initial scene data can be selected to obtain target scene data with lower parameter dimensions, which will help to perform adaptive sampling more quickly in the subsequent process and obtain target sampling results.

[0047] The scenario information corresponding to the vehicle functional failure event may include: the probability distribution of the scenario corresponding to the vehicle functional failure event, the scenario factors that lead to the vehicle functional failure event, and the target scenario data corresponding to the vehicle functional failure event. The target failure probability distribution can be obtained from a preset data packet, or by using a probability distribution generation method. The probability distribution generation method may include but is not limited to: a statistical regression analysis method, a simulation method, a Bayesian method, and a probability distribution generation method based on a machine learning model. The target sampling result can be obtained by an adaptive sampling method. The adaptive sampling method may include but is not limited to: an adaptive importance sampling method, a multiple adaptive importance sampling method, and a sequential Monte Carlo method. The target sampling result may include multiple target sampling samples and evaluation indicators corresponding to the multiple target sampling samples.

[0048] It is easy to understand that by using the adaptive sampling method, the target scene data is adaptively sampled, and the sampling distribution corresponding to the sampling results is continuously adjusted during the sampling process, so that the sampling distribution corresponding to the sampling results gradually approaches the target failure probability distribution, which can find the vehicle's functional failure events more quickly and accurately, thereby obtaining the target sampling results more quickly and accurately.

[0049] The aforementioned functional failure probability can be obtained using a probability assessment calculation method. These methods may include, but are not limited to, summation, arithmetic mean, weighted mean, and a combination of manual scoring. Using this probability assessment calculation method, multiple target sampling samples within the target sampling results are evaluated and calculated to obtain the sample functional failure probabilities corresponding to each of the target sampling samples. Using these multiple sample functional failure probabilities, the functional failure probability of the vehicle in a real-time scenario is then determined. The aforementioned confidence level may be either confidence level information or a confidence assessment value.

[0050] The functional failure assessment method for the above-mentioned vehicle provided in the embodiment of the present application achieves the following technical effects: by obtaining the initial scene data corresponding to the vehicle and performing parameter dimensionality reduction processing on the initial scene data, the initial scene data corresponding to the real-time scene in which the vehicle is located can be simplified to obtain target scene data with lower parameter dimensions, which helps to obtain target sampling results more quickly in the future; further, based on the target failure probability distribution, the target scene data is adaptively sampled, so that the sampling process can focus more on the functional failure events of the vehicle, and can find the functional failure events of the vehicle more quickly and accurately, thereby obtaining the target sampling results more quickly and accurately; further, the evaluation calculation is performed based on the target sampling results, which can improve the evaluation efficiency and accuracy of the functional failure probability. Therefore, the embodiment of the present application achieves the purpose of combining parameter dimensionality reduction processing and adaptive sampling to efficiently and accurately evaluate the functional failure probability of the vehicle in the real-time scene, achieves the technical effect of improving the evaluation efficiency and accuracy of the functional failure probability, and solves the technical problem of low efficiency and low accuracy of the vehicle functional failure assessment method in the related art due to reliance on the Monte Carlo method.

[0051] The vehicle functional failure assessment method provided in the embodiment of the present application combines parameter dimensionality reduction processing and adaptive sampling to efficiently and accurately assess the functional failure probability of the vehicle in real-time scenarios, and can be widely used in multiple application scenarios.

[0052] For example, in the application scenario of safety verification of vehicle autonomous driving functions, during the vehicle development and testing phase, the technical solution of this application can be applied to evaluate the failure probability of the vehicle in rare scenarios such as complex traffic, bad weather, and special road conditions, thereby guiding the improvement of the vehicle's autonomous driving system and enhancing the safety of the vehicle's autonomous driving system.

[0053] For example, in the application scenario of the intelligent transportation system, the intelligent transportation system relies on the communication function between vehicles and the communication function between vehicles and infrastructure to indicate traffic flow to alleviate traffic pressure. By applying the technical solution of this application, the failure probability of the intelligent transportation system under various complex traffic modes (such as network terminals, etc.) can be evaluated, thereby improving the intelligent transportation system and providing better intelligent transportation services.

[0054] For example, in the application scenario of driverless taxi services, before introducing driverless taxi services, a comprehensive assessment of their safety and passenger experience in complex urban traffic environments is required. The technical solution of this application can efficiently identify potential risk scenarios and assess the failure probability of driverless taxis in potential risk scenarios, thereby guiding the improvement of driverless taxis and providing safer driverless taxi services.

[0055] The vehicle functional failure assessment method provided in the embodiments of the present application can be applied to, but is not limited to, the application scenarios listed above. With the continuous evolution of technology, the above method can also be applied to a wider range of scenarios, such as drone transportation, remote-controlled vehicles, etc. By efficiently and accurately assessing the probability of functional failure, it assists in improving the system corresponding to the machine function, supports a variety of advanced functions and applications, and can improve the safety of the machine function.

[0056] Optionally, the initial scene data includes multiple scene parameters. In the above step S20, performing parameter dimensionality reduction processing on the initial scene data to obtain target scene data includes the following steps:

[0057] S201: performing variance decomposition calculation on multiple scenario parameters to obtain failure sensitive data corresponding to the multiple scenario parameters, wherein the failure sensitive data is used to evaluate the impact of parameter changes of the multiple scenario parameters on the probability of functional failure;

[0058] S202: Selecting some scenario parameters from multiple scenario parameters according to the failure sensitive data;

[0059] S203: Generate target scene data using some scene parameters.

[0060] The initial scenario data may also include a combination of multiple scenario parameters. The failure-sensitive data may be obtained using a variance decomposition method. The variance decomposition method may include, but is not limited to, a variance-sensitive Sobol index method (i.e., a Sobol index method), a method based on a structural equation model, and a Bayesian variance decomposition method.

[0061] By using the variance decomposition method, variance decomposition calculation can be performed on multiple scenario parameters contained in the initial scenario data. The degree of influence of each scenario parameter in the multiple scenario parameters on the probability of functional failure when the parameter change within the parameter change range corresponding to the scenario parameter can be determined, and the corresponding degree of influence of the scenario parameter can be quantified to obtain failure sensitive data corresponding to each scenario parameter in the multiple scenario parameters. The larger the numerical value corresponding to the failure sensitive data, the greater the influence of the scenario parameter on the probability of functional failure.

[0062] Based on the failure-sensitive data corresponding to each of the multiple scene parameters, a portion of the scene parameters is selected from the multiple scene parameters according to a parameter selection method. The above parameter selection method can be: based on the failure-sensitive data corresponding to each of the multiple scene parameters, the multiple scene parameters are sorted in descending order of the failure-sensitive data, and the scene parameters ranked in the top M in the sorting result are selected. The above parameter selection method can also be: setting a failure-sensitive data threshold, sequentially comparing the failure-sensitive data corresponding to each of the multiple scene parameters with the failure-sensitive data threshold, and if the failure-sensitive data corresponding to a certain scene parameter is higher than the failure-sensitive data threshold, then selecting the scene parameter into the portion of the scene parameters.

[0063] The target scene data may include multiple scene samples, which may be generated using multiple scene parameters and a combination relationship between the multiple scene parameters. The target scene data may be represented by a probability space.

[0064] In an exemplary application scenario, Figure 2 As shown in the figure, the selected scene parameters are combined with the combination relationship between the scene parameters to generate multiple scene samples. These scene samples are used to construct a probability space (i.e., target scene data). The probability space can be recorded as K = (k1, k2, ..., k d ), where each scene sample can be a point k in the probability space K, and k can be k1, k2, ..., k d , the probability space K is accompanied by a probability density function f1(·).

[0065] It should be noted that the aforementioned scenario samples can represent the real-time scenario of the vehicle, or they can represent virtual scenarios generated by randomly combining some scenario parameters corresponding to the real-time scenario the vehicle is in. By utilizing these virtual scenarios, a more comprehensive analysis can be conducted to determine whether a vehicle function failure event has occurred in the real-time scenario in which it is located.

[0066] The above-mentioned optional embodiments of the present application can achieve the following technical effects: by performing variance decomposition calculations on multiple scenario parameters, the degree of influence of different scenario parameters on the probability of functional failure is quantified, and some scenario parameters that have a greater influence on the probability of functional failure are screened out. On the one hand, it solves the problem of waste of computing resources caused by too many parameters in related technologies, and improves the efficiency of functional failure assessment of vehicles. On the other hand, it screens out some scenario parameters that have a greater influence on the probability of functional failure, and only needs to perform functional failure assessment of the vehicle for these scenario parameters, avoiding the problem of low accuracy of evaluation results due to redundant analysis of irrelevant parameters, and improving the accuracy of functional failure assessment of vehicles.

[0067] Optionally, the failure sensitive data includes multiple sensitivity indexes. In the above step S201, performing variance decomposition calculation on multiple scene parameters to obtain failure sensitive data corresponding to the multiple scene parameters includes the following steps:

[0068] S211: Calculate multiple scene parameters based on the variance-sensitive Sobol index method to obtain the overall variance and the decomposed variances corresponding to the multiple scene parameters;

[0069] S212: For a target scene parameter among the multiple scene parameters, obtain a target sensitivity index corresponding to the target scene parameter among the multiple sensitivity indices according to the overall variance and the decomposition variance corresponding to the target scene parameter.

[0070] The above-mentioned overall variance can be used to reflect the comprehensive impact of all scenario parameters included in the initial scenario data on the probability of functional failure. The above-mentioned decomposed variance can be used to reflect the degree of impact of changes in a single scenario parameter in the initial scenario data on the overall variance.

[0071] In an exemplary application scenario, the variance-sensitive Sobol index method (i.e., the Sobol index method) is used to perform variance decomposition calculation on multiple scene parameters. Assume that there are n scene parameters in the initial scene data, and the n scene parameters are sequentially recorded as x1, x2, ..., x n , the n scenario parameters are used as an input vector to calculate the probability of vehicle function failure, and the input vector is recorded as x=(x1,x2,……,x n ), and the calculated vehicle function failure probability is recorded as y = f(x). The above-mentioned overall variance and the decomposed variance corresponding to multiple scenario parameters can be obtained by the following process: Generate any scenario parameter x in the initial scenario data i Corresponding multiple (e.g., j) data points, the j data points are sequentially recorded as a i1 ,a i2 ,……,a ij , where a ij Characterize the scene parameter x i Corresponding to the jth data point, multiple scene parameters are combined to obtain multiple sets of experimental input vectors, for example, (a 17 ,a 21 ,……,a n9), based on the variance-sensitive Sobol index method (i.e., Sobol index method), the vehicle function failure probability corresponding to the multiple sets of experimental input vectors and the expected value of the vehicle function failure probability (denoted as E[f(x)]) are calculated. The overall variance (denoted as Var(y)) is calculated using the vehicle function failure probability corresponding to the multiple sets of experimental input vectors and the expected value of the vehicle function failure probability, as shown in formula (1); further, based on the Sobol index method, the above-mentioned overall variance is decomposed into any scenario parameter x i The corresponding partial variance (i.e., decomposed variance) Can represent only scene parameters x i The variance value obtained when the change occurs.

[0072] Var(y)=∫ x [f(x)-E[f(x)]] 2 dx formula (1)

[0073] Still in the above application scenario, multiple scene parameters are selected as target scene parameters in turn, and the target scene parameters are recorded as x obj , according to the overall variance and the target scene parameter x obj The corresponding decomposition variance Calculate the target sensitivity index S corresponding to the target scene parameters obj , as shown in formula (2).

[0074]

[0075] The above-mentioned optional embodiments of the present application can achieve the following technical effects: multiple scenario parameters are calculated based on the variance-sensitive Sobol index method. In the process of quantifying the degree of influence of scenario parameters on the probability of vehicle functional failure, not only the direct influence of a single scenario parameter on the probability of vehicle functional failure can be considered, but also the indirect influence of the interaction effect between the scenario parameter and other scenario parameters on the probability of vehicle functional failure can be considered. It can more comprehensively consider the degree of influence of parameter changes of scenario parameters on the probability of functional failure, obtain a more accurate sensitivity index, and thus improve the accuracy of failure-sensitive data.

[0076] Optionally, in the above step S30, adaptively sampling the target scene data based on the target failure probability distribution to obtain the target sampling result includes the following steps:

[0077] S301: Sampling target scene data according to an initial sampling distribution to obtain an initial sampling result;

[0078] S302: Adaptively iteratively update the initial sampling result according to the target failure probability distribution to obtain a target sampling result, wherein the similarity between the sampling distribution of the target sampling result and the target failure probability distribution meets a target condition.

[0079] The initial sampling distribution may be a preset sampling distribution. Selectable initial sampling distributions include, but are not limited to, uniform distribution, Poisson distribution, Bernoulli distribution, and normal distribution. The initial sampling results may include multiple initial sampling samples and evaluation indicators corresponding to each of the multiple initial sampling samples. The target conditions may be set based on sampling requirements. For example, the degree of similarity between the sampling distribution of the target sampling results and the target failure probability distribution may be greater than a preset degree of similarity.

[0080] In an exemplary application scenario, Figure 2 As shown, the functional failure probability of the target function of the vehicle in the scenario corresponding to the target scenario data is evaluated. The above-mentioned target sampling result is obtained by using an adaptive importance sampling method, and a uniform distribution is determined as the above-mentioned initial sampling distribution. According to the initial sampling distribution, multiple scene samples in the target scene data are sampled to obtain multiple initial sampling samples and evaluation indicators corresponding to the multiple initial sampling samples. The multiple initial sampling samples and the evaluation indicators corresponding to the multiple initial sampling samples are used as the initial sampling result, and the sampling distribution corresponding to the initial sampling result is determined. Furthermore, the similarity between the sampling distribution corresponding to the initial sampling result and the target failure probability distribution is calculated. If the similarity between the sampling distribution corresponding to the initial sampling result and the target failure probability distribution is greater than a preset similarity, the initial sampling result is used as the target sampling result. If the similarity between the sampling distribution corresponding to the initial sampling result and the target failure probability distribution is not greater than the preset similarity, the initial sampling result is adaptively iteratively updated according to the target failure probability distribution until the similarity between the sampling distribution of the target sampling result and the target failure probability distribution meets the target condition, thereby obtaining the target sampling result.

[0081] The above-mentioned optional embodiments of the present application can achieve the following technical effects: utilizing adaptive sampling technology, by iteratively updating the sampling distribution, dynamically adjusting the sampling strategy, so that during the sampling process, more attention can be paid to the functional failure events of the vehicle (especially, rare functional failure events), and the functional failure events of the vehicle can be captured more quickly and accurately, saving computing resources, speeding up the speed of obtaining the target sampling results, and improving the accuracy of the target sampling results, thereby laying a data foundation for the subsequent efficient and accurate calculation of the functional failure probability.

[0082] Optionally, in step S302 above, adaptively iteratively updating the initial sampling result according to the target failure probability distribution to obtain the target sampling result includes the following steps:

[0083] S321: In the current iteration round of the adaptive iterative update process, using the previous round sampling results corresponding to the previous iteration round and the target failure probability distribution, the previous round sampling distribution corresponding to the previous round sampling results is updated to the current sampling distribution;

[0084] S322: Sample the target scene data according to the current sampling distribution to obtain the current sampling result corresponding to the current iteration round.

[0085] The current iteration round can be used to represent the current state of the adaptive iterative update process. The previous iteration round can refer to the iteration round before the current iteration round in the adaptive iterative update process. The previous round sampling result can be the sampling result obtained in the previous iteration round. The previous round sampling distribution can be obtained by statistically analyzing and adjusting the previous round sampling results.

[0086] In an exemplary application scenario, the front wheel sampling result is the initial sampling result, and the current iteration round is the second iteration round. In the current iteration round of the adaptive iterative update process, the front wheel sampling result (i.e., the initial sampling result) is statistically analyzed, and the statistical analysis result is adjusted using the front wheel sampling result and the target failure probability distribution to obtain a front wheel sampling distribution corresponding to the front wheel sampling result; for example, if the combination of the front wheel sampling result and the target failure probability indicates that a specific driver behavior pattern (such as frequent lane changes) is highly correlated with the functional failure probability, the obtained front wheel sampling distribution will pay more attention to the behavior pattern, and in the current iteration round, it can more accurately capture the samples corresponding to the behavior pattern, thereby more accurately evaluating the functional failure probability.

[0087] Still in the above application scenario, determine whether the similarity between the front-wheel sampling distribution and the target failure probability distribution meets the target condition. If the similarity between the front-wheel sampling distribution and the target failure probability distribution does not meet the target condition, then the front-wheel sampling distribution corresponding to the front-wheel sampling result is used as the current sampling distribution corresponding to the current iteration round, and the target scene data is sampled according to the current sampling distribution to obtain the current sampling result corresponding to the current iteration round.

[0088] It should be noted that, in the above iterative process, an iteration round threshold may also be set. If the number of iteration rounds exceeds the iteration round threshold, the iteration is exited to avoid an infinite loop problem.

[0089] The above-mentioned optional embodiments of the present application can achieve the following technical effects: by iteratively optimizing the sampling distribution, each sampling result is made closer to the expected result, overcoming the defect of the mismatch between the sampling distribution and the failure probability distribution in the traditional sampling method, and being able to efficiently and accurately evaluate the probability of vehicle function failure.

[0090] Optionally, in step S321, updating the front wheel sampling distribution to the current sampling distribution using the front wheel sampling results and the target failure probability distribution includes the following steps:

[0091] S3211: Using the target failure probability distribution, weight calculation is performed on multiple sampling samples in the previous round sampling results to obtain sample weights corresponding to the multiple sampling samples;

[0092] S3212: Selecting a target sample from multiple sampling samples based on multiple sample weights;

[0093] S3213: Perform distribution fitting processing on the target sample to obtain the current sampling distribution corresponding to the current iteration round.

[0094] The above-mentioned sample weights are used to characterize the relative value of a sample in the process of assessing the probability of functional failure. The higher the sample weight of a sample, the higher the probability of the sample appearing in the target failure probability distribution compared to the probability of the sample appearing in the previous wheel sampling results. When the sample weight of a sample is 1, the probability of the sample appearing in the target failure probability distribution is the same as the probability of the sample appearing in the previous wheel sampling results. When the sample weight of a sample is greater than 1, the probability of the sample appearing in the target failure probability distribution is higher than the probability of the sample appearing in the previous wheel sampling results. In the process of assessing the probability of functional failure of the vehicle, more attention should be paid to the sample. When the sample weight of a sample is less than 1, the probability of the sample appearing in the target failure probability distribution is lower than the probability of the sample appearing in the previous wheel sampling results. In the process of assessing the probability of functional failure of the vehicle, less attention should be paid to the sample.

[0095] The above-mentioned target samples can be obtained according to the sample selection method. According to the sample weights corresponding to the multiple sampling samples, the target samples are selected from the multiple sampling samples according to the sample selection method. The above-mentioned sample selection method can be: according to the sample weight corresponding to each sampling sample in the multiple sampling samples, the multiple sample weights are sorted in order from high to low, and the sampling samples that are ranked in the top U (the specific number can be set according to the sampling requirements) in the sorting result and have a sample weight greater than 1 are selected as target samples. The above-mentioned sample selection method can also be: setting a weight threshold, comparing the sample weight corresponding to each sampling sample in the multiple sampling samples with the weight threshold in turn, and taking the multiple sampling samples whose sample weights corresponding to the sampling samples are higher than the weight threshold as target samples. The above-mentioned sample selection method can also be: according to the multiple sample weights, and in combination with the evaluation indicators corresponding to the multiple sampling samples, the sampling samples whose sample weights are greater than 1 and the evaluation indicators represent the functional failure event of the vehicle are determined as target samples.

[0096] In an exemplary application scenario, the target failure probability distribution is a multivariate normal distribution (denoted as p(z)), and the evaluation index corresponding to the sampling sample is denoted as g1(z). Statistical analysis is performed on multiple sampling samples in the previous round sampling results to obtain the statistical sampling distribution corresponding to the previous round sampling results (denoted as q(z)), and then the probability of any sampling sample in the multiple sampling samples included in the previous round sampling results appearing in the statistical sampling distribution q(z) can be obtained (denoted as q(z)). r )), using the target failure probability distribution, we can get the probability of the sample appearing in the target failure probability distribution p(z) (denoted as p(z) r )); Further, the probability q(z r ) and the probability p(z r ), calculate the sample weight corresponding to the sample (denoted as ω r ), as shown in formula (3).

[0097]

[0098] Still in the above application scenario, for any sample z r , when the sample z r The corresponding evaluation index g1(z r ) is less than or equal to 0, it indicates that the vehicle is in the sampling sample z r A functional failure event occurs in the corresponding scenario. According to the sample weight corresponding to each sampling sample in the multiple sampling samples, a sampling sample with a sample weight greater than 1 and a value of the evaluation index g1(z) less than or equal to 0 is selected from the multiple sampling samples as the target sample. Furthermore, the target sample is used to fit the current sampling distribution. Specifically, the weighted mean of the current iteration round is first calculated based on the target sample and the weight corresponding to the target sample. The weighted mean can be used to move the sampling distribution to the high weight area; then, the weighted covariance of the current round is calculated based on the weighted mean of the current iteration round, the target sample and the weight corresponding to the target sample. The weighted covariance can be used to reflect the degree of discreteness of the target sample; finally, the above-mentioned weighted mean and the above-mentioned weighted covariance are used to perform distribution fitting to obtain the current sampling distribution corresponding to the current iteration round (denoted as q new (z)), the current sampling distribution satisfies formula (4).

[0099] q new (z)∝q(z)·exp(α·ω(z)) Formula (4)

[0100] In formula (4), α represents a fixed value, which can be set according to sampling requirements.

[0101] The above-mentioned optional embodiments of the present application can achieve the following technical effects: by calculating the sample weights corresponding to multiple sampling samples, the target samples that have a greater impact on the probability of calculation function failure can be accurately identified, so that the current sampling distribution gradually approaches the area that contributes more to the probability of calculation function failure, avoiding the problem of invalid sampling caused by using a fixed sampling distribution, thereby improving the calculation efficiency and accuracy of subsequent function failure probabilities.

[0102] Optionally, the vehicle functional failure assessment method further includes the following steps:

[0103] S50: Acquire test case data corresponding to the vehicle, wherein the test case data is used to characterize the test scenario in which the vehicle is located;

[0104] S60: Performing scenario testing on the test case data to obtain functional performance data of the vehicle, wherein the functional performance data is used to evaluate the failure probability of the target function of the vehicle in the test scenario;

[0105] S70: Use test case data and functional performance data to construct a target failure probability distribution.

[0106] The above-mentioned test case data can be historical real-time scene data collected during the vehicle's driving process, or it can be simulated scene data obtained by simulating driving scenes using simulation technology. The test case data may include scene data corresponding to multiple scenes to be tested. In particular, simulation technology can be used to simulate rare driving scenes (such as extreme road conditions) to obtain simulated scene data corresponding to these rare driving scenes. These simulated scene data are used as test case data, thereby more comprehensively considering whether the vehicle has a functional failure event under various environments.

[0107] The functional performance data may be a continuous score value (e.g., any number between 0 and 1) or a binary value (e.g., 0 or 1). In particular, when the functional performance data is a binary value, a functional performance data value of 0 indicates that the target function of the vehicle has not failed in the test scenario, and a functional performance data value of 1 indicates that the target function of the vehicle has failed in the test scenario.

[0108] In an exemplary application scenario, simulation technology is used to simulate specific driving scenarios (such as rainy days, complex traffic roads, extreme environmental conditions, etc.), and the scenario data under these specific driving scenarios are used as test case data; further, the scenario data corresponding to each to-be-tested scenario in the test case data is subjected to scenario testing to obtain whether the target function of the vehicle has a functional failure event in the above-mentioned to-be-tested scenario, and obtain the functional performance data of the vehicle in the above-mentioned to-be-tested scenario; then, the test case data and the functional performance data corresponding to each to-be-tested scenario in the test case data are used to construct a target failure probability distribution. In particular, in the process of constructing the target failure probability distribution, more attention can be paid to rare to-be-tested scenarios, so that the probability of rare to-be-tested scenarios appearing in the target failure probability distribution is higher; when the functional performance data corresponding to a certain to-be-tested scenario is less than the rare failure probability threshold, it can be considered that the to-be-tested scenario belongs to a rare to-be-tested scenario.

[0109] In another exemplary application scenario, a deep learning algorithm can be used to take the historical real-time scene data collected by the vehicle during driving as input to generate unknown test scenarios. Furthermore, simulation technology is used to simulate the simulation scene data corresponding to the unknown test scenarios, and this part of the simulation scene data is also used as test case data, thereby achieving scenario generalization, so that a more comprehensive test scenario can be considered during the testing phase.

[0110] The above-mentioned optional embodiments of the present application can achieve the following technical effects: by combining actual scenarios and simulation scenarios, the vehicle's test scenarios can be considered more comprehensively, and then functional performance data can be obtained by performing scenario testing on the test case data, thereby ensuring the reliability and accuracy of the constructed target failure probability distribution, providing a basis for subsequent adaptive sampling, and avoiding the problem of low accuracy of failure probability assessment results due to failure to consider rare failure events.

[0111] Optionally, in the above step S40, performing evaluation calculation based on the target sampling result to obtain the functional failure probability of the vehicle in the real-time scenario includes the following steps:

[0112] S401: Determine failure indication data corresponding to a target sampling result using functional performance data corresponding to a target failure probability distribution;

[0113] S402: Calculate the functional failure probability using the failure indication data, the target failure probability distribution, and the sampling distribution of the target sampling results.

[0114] The failure indication data can be used to indicate whether a vehicle has experienced a functional failure in a real-time scenario. The failure indication data can be a continuous score value (e.g., any number between 0 and 1) or a binary value (e.g., 0 or 1).

[0115] In an exemplary application scenario, Figure 2 As shown, the functional performance data corresponding to the target failure probability distribution is used to determine the scene performance data corresponding to the multiple target sampling samples in the target sampling result, and the scene performance data corresponding to the arbitrary target sampling sample z is used. r Corresponding scene performance data, determine the failure indication data corresponding to the target sampling sample (recorded as f2(z r )). The above-mentioned determination of the scenario performance data corresponding to the target sampling sample can be performed by comparing and analyzing multiple scenario parameters corresponding to the target sampling sample with the test case data corresponding to the target failure probability distribution, determining the degree of similarity between the scenario corresponding to the target sampling sample and the scenario to be tested corresponding to a certain test case data in the target failure probability distribution, and determining the functional performance data corresponding to the scenario to be tested with the greatest degree of similarity as the scenario performance data corresponding to the target sampling sample.

[0116] Still in the above application scenario, for any target sampling sample z r , using the target sample z r The corresponding failure indication data f2(z r ), the target sample z r The probability of occurrence p(z r ) and the target sample z r The probability q(z r ) is calculated to obtain the target sampling sample z r The corresponding functional failure probability; then, the target sampling result contains multiple target sampling samples (e.g., a total of N num The functional failure probability corresponding to the target sampling samples is calculated by weighted average to obtain the functional failure probability (denoted as P fail ), calculate the functional failure probability P fail The process is shown in formula (5).

[0117]

[0118] In another exemplary application scenario, the target scene data is represented by a probability space, and the probability space corresponding to the target scene data can be K=(k1, k2, ..., k d ), where each scene sample contained in the target scene data can be represented as a point k in the probability space, and k can take values of k1, k2, ..., k d, the probability space K is accompanied by a probability density function f1(·), the corresponding evaluation index of the scene sample is marked as g2(z), and the functional failure probability of the vehicle's target function in the scene corresponding to the target scene data can be calculated by formula (6); further, if the average frequency of the scene corresponding to the target scene data within the target range (e.g., within H kilometers) is f c , then according to Bayes' theorem, the average failure probability of the vehicle's target function within the target range (denoted as λ) can be calculated by formula (7); then, the function failure probability is adjusted using this average failure probability.

[0119]

[0120] The above-mentioned optional embodiments of the present application can achieve the following technical effects: using the functional performance data corresponding to the target failure probability distribution to determine the failure indication data corresponding to the target sampling result, it can more accurately reflect the actual performance of the vehicle's target function in various scenarios (especially rare failure scenarios). By combining the failure indication data, the target failure probability distribution and the sampling distribution of the target sampling results, the calculated functional failure probability is more accurate.

[0121] Example 2

[0122] The present application also provides a vehicle failure assessment system 30, please refer to Figure 3 , including: an acquisition unit 310, used to acquire initial scene data corresponding to the vehicle, wherein the initial scene data is used to characterize the real-time scene in which the vehicle is located; a processing unit 320, used to perform parameter dimensionality reduction processing on the initial scene data to obtain target scene data; a sampling unit 330, used to adaptively sample the target scene data based on the target failure probability distribution to obtain a target sampling result, wherein the target failure probability distribution is used to characterize the distribution of occurrence scene information corresponding to the functional failure event of the vehicle; an evaluation unit 340, used to perform evaluation calculation based on the target sampling result to obtain the functional failure probability of the vehicle in the real-time scene.

[0123] The functional failure assessment system for the above-mentioned vehicle provided in the embodiment of the present application achieves the following technical effects: utilizing the acquisition unit 310 to acquire initial scene data corresponding to the vehicle, utilizing the processing unit 320 to perform parameter dimensionality reduction processing on the initial scene data, thereby simplifying the initial scene data corresponding to the real-time scene in which the vehicle is located, and obtaining target scene data with lower parameter dimensions, which helps to obtain target sampling results more quickly in the subsequent process; further, utilizing the sampling unit 330 to adaptively sample the target scene data based on the target failure probability distribution, so that the sampling process can focus more on functional failure events of the vehicle, and can find functional failure events of the vehicle more quickly and accurately, thereby obtaining target sampling results more quickly and accurately; further, utilizing the evaluation unit 340 to perform evaluation calculations based on the target sampling results, thereby improving the evaluation efficiency and accuracy of the functional failure probability, enhancing the system's performance in performing efficient and accurate functional failure assessments on vehicles, and providing better functional failure assessment services.

[0124] Example 3

[0125] The present application embodiment also provides a vehicle 40, please refer to Figure 4 , including an on-board memory 410 and an on-board processor 420, wherein the on-board memory 410 is used to store computer programs; the on-board processor 420 is used to execute the computer programs stored in the memory to implement the vehicle functional failure assessment method of any embodiment.

[0126] The above-mentioned vehicle provided by the embodiment of the present application achieves the following technical effects: by obtaining the initial scene data corresponding to the vehicle and performing parameter dimensionality reduction processing on the initial scene data, the initial scene data corresponding to the real-time scene in which the vehicle is located can be simplified to obtain target scene data with lower parameter dimensions, which helps to obtain target sampling results more quickly in the future; further, based on the target failure probability distribution, the target scene data is adaptively sampled, so that the sampling process can focus more on the functional failure events of the vehicle, and can find the functional failure events of the vehicle more quickly and accurately, thereby obtaining the target sampling results more quickly and accurately; further, evaluation and calculation are performed based on the target sampling results, which can improve the evaluation efficiency and accuracy of the functional failure probability, enhance the safety of the vehicle, and thus improve the overall performance of the vehicle.

[0127] Those skilled in the art will appreciate that, similarly, the above-mentioned vehicle may also be a computing terminal. Figure 5 This is a hardware structure block diagram of a computing terminal for implementing a vehicle functional failure assessment method according to an embodiment of the present application. Figure 5As shown, a computing terminal 50 (e.g., a computer terminal, a mobile smart terminal, a vehicle terminal, or a cloud computing virtual terminal, etc.) may include: one or more processors 502 (e.g., may include processors 502a, 502b, ..., 502n), a memory 504 for storing data, and a transmission device 506 for implementing a communication function, wherein the processor 502 may include but is not limited to processing components such as a microprocessor (Microcontroller Unit, abbreviated as MCU) or a programmable logic device (Field Programmable Gate Array, abbreviated as FPGA).

[0128] The above-mentioned computing terminal 50 may also include: a display, an input / output interface, a Universal Serial Bus (USB) port (the USB port can be used as one of the ports of the computer bus, not shown in the figure), a network interface (not shown in the figure), a power supply (not shown in the figure) and a camera (not shown in the figure).

[0129] It should be noted that the one or more processors 502 and / or other data processing circuits in the computing terminal 50 described above may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computing terminal 50 (or mobile device).

[0130] The memory 504 can be used to store software programs and modules of application software, such as program instructions and data storage devices corresponding to the path planning method in the embodiment of the present application. The processor 502 executes various functional applications and data processing by running the software programs and modules stored in the memory 504, that is, implementing the above-mentioned path planning method. The memory 504 may include a high-speed random access memory and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 504 may further include a memory remotely located relative to the processor 502, and these remote memories may be connected to the vehicle terminal 50 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0131] The transmission device 506 is used to receive or send data via a network. Specific examples of the aforementioned network may include a wireless network provided by the communication provider of the vehicle terminal 50. In one example, the transmission device 506 includes a network adapter (Network Interface Controller, abbreviated as NIC) and a network interface. The network adapter can be connected to other network devices through a base station so as to communicate with the Internet. The transmission device 506 can communicate data using a wired and / or wireless network connection. In one example, the transmission device 506 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0132] The input / output interface can be connected to corresponding input / output devices of the computing terminal 50 to implement input / output functions. Such input / output devices may include, but are not limited to, cursor control devices, keyboards, displays, etc. These input / output devices may be built into the computing terminal 50 or external devices connected to the computing terminal 50.

[0133] It can be understood by those skilled in the art that Figure 5 The structure of the computing terminal 50 shown is only for illustration and does not impose a strict limitation on the structure of the computing terminal 50. For example, the computing terminal 50 may also include Figure 5 More or fewer components than those shown in FIG, or the computing terminal 50 may have the same Figure 5 Different categories of components are shown.

[0134] Example 4

[0135] The present application also provides a vehicle function failure assessment device 60, please refer to Figure 6 , including: an acquisition module 601, used to obtain initial scene data corresponding to the vehicle, wherein the initial scene data is used to characterize the real-time scene in which the vehicle is located; a processing module 602, used to perform parameter dimensionality reduction processing on the initial scene data to obtain target scene data; a sampling module 603, used to adaptively sample the target scene data based on the target failure probability distribution to obtain a target sampling result, wherein the target failure probability distribution is used to characterize the scene information corresponding to the functional failure event of the vehicle; an evaluation module 604, used to perform evaluation calculation based on the target sampling result to obtain the functional failure probability of the vehicle in the real-time scene, wherein the functional failure probability is used to evaluate the confidence of the target function of the vehicle in the real-time scene.

[0136] The functional failure assessment device for the above-mentioned vehicle provided in the embodiment of the present application achieves the following technical effects: utilizing the acquisition module 601 to acquire initial scene data corresponding to the vehicle, utilizing the processing module 602 to perform parameter dimensionality reduction processing on the initial scene data, thereby simplifying the initial scene data corresponding to the real-time scene in which the vehicle is located, and obtaining target scene data with lower parameter dimensions, which helps to obtain target sampling results more quickly in the subsequent process; further, utilizing the sampling module 603 to adaptively sample the target scene data based on the target failure probability distribution, so that the sampling process can focus more on functional failure events of the vehicle, and can find functional failure events of the vehicle more quickly and accurately, thereby obtaining target sampling results more quickly and accurately; further, utilizing the evaluation module 604 to perform evaluation calculations based on the target sampling results, thereby improving the evaluation efficiency and accuracy of the functional failure probability, thereby enabling the functional failure assessment device of the vehicle to perform functional failure assessment on the vehicle more efficiently and accurately, thereby enhancing the performance of the functional failure assessment device of the vehicle.

[0137] It should be noted that the optional implementation methods of this embodiment can refer to the relevant description in Example 1 and will not be repeated here.

[0138] Example 5

[0139] The present application embodiment also provides an electronic device 70, please refer to Figure 7 , including a memory 710 and a processor 720, wherein the memory 710 is used to store computer programs; the processor 720 is used to execute the programs stored in the memory 710 to implement the vehicle functional failure assessment method introduced in any embodiment of the present application.

[0140] The above-mentioned electronic device provided in the embodiment of the present application achieves the following technical effects: by acquiring the initial scene data corresponding to the vehicle and performing parameter dimensionality reduction processing on the initial scene data, the initial scene data corresponding to the real-time scene in which the vehicle is located can be simplified to obtain target scene data with lower parameter dimensions, which helps to obtain target sampling results more quickly in the future; further, based on the target failure probability distribution, the target scene data is adaptively sampled, so that the sampling process can focus more on the functional failure events of the vehicle, and can find the functional failure events of the vehicle more quickly and accurately, thereby obtaining the target sampling results more quickly and accurately; further, evaluation and calculation are performed based on the target sampling results, which can improve the evaluation efficiency and accuracy of the functional failure probability, so that the electronic device can perform efficient and accurate functional failure evaluation on the target function of the vehicle, thereby improving the performance of the electronic device.

[0141] It can be understood by those skilled in the art that Figure 7The structure shown is for illustration only, and the electronic device may also be a terminal device such as a smart phone (eg, an Android phone, an iOS phone, etc.), a tablet computer, a PDA, or a mobile Internet device (MID). Figure 7 It does not limit the structure of the above electronic device. For example, the electronic device 70 may also include Figure 7 More or fewer components (e.g., network interface, display device, etc.) shown in, or with Figure 7 Different configurations shown.

[0142] Example 6

[0143] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the vehicle functional failure assessment method introduced in any embodiment of the present application is implemented.

[0144] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store computer programs.

[0145] The above-mentioned computer-readable storage medium provided in the embodiment of the present application achieves the following technical effects: by obtaining the initial scene data corresponding to the vehicle and performing parameter dimensionality reduction processing on the initial scene data, the initial scene data corresponding to the real-time scene in which the vehicle is located can be simplified to obtain target scene data with lower parameter dimensions, which helps to obtain target sampling results more quickly in the subsequent process; further, based on the target failure probability distribution, the target scene data is adaptively sampled, so that the sampling process can focus more on the functional failure events of the vehicle, and can find the functional failure events of the vehicle more quickly and accurately, thereby obtaining the target sampling results more quickly and accurately; further, evaluation calculation is performed based on the target sampling results, which can improve the evaluation efficiency and accuracy of the functional failure probability.

[0146] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0147] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0148] In this application, a plurality refers to two or more.

[0149] In this application, unless otherwise expressly defined, the terms "mounted," "connected," and "connected" should be interpreted broadly. For example, they can refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediary; and internal communication between two components. A person of ordinary skill in the art will understand the specific meanings of these terms in this application.

[0150] In this application, the terms "first," "second," "third," "fourth," etc. (if any) are used to distinguish similar objects and are not necessarily used to describe a particular sequential order.

[0151] The term "and / or" in this application simply describes an association between related objects, indicating that three possible relationships exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this application generally indicates that the related objects are in an "or" relationship.

[0152] Unless otherwise specified, all steps of the present application may be performed sequentially or randomly. For example, a statement that the method includes steps A and B indicates that the method may include steps A and B performed sequentially, or steps B and A performed sequentially. For example, a statement that the method may also include step C indicates that step C may be added to the method in any order, for example, the method may include steps A, B, and C, or steps A, C, and B, or steps C, A, and B, etc.

[0153] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for evaluating vehicle functional failure, characterized in that: include: Acquiring initial scene data corresponding to the vehicle, wherein the initial scene data is used to represent the real-time scene in which the vehicle is located; Performing parameter dimensionality reduction processing on the initial scene data to obtain target scene data; Adaptively sampling the target scenario data based on a target failure probability distribution to obtain a target sampling result, wherein the target failure probability distribution is used to characterize scenario information corresponding to a functional failure event of the vehicle; An evaluation calculation is performed based on the target sampling result to obtain a functional failure probability of the vehicle in the real-time scenario, wherein the functional failure probability is used to evaluate the confidence of the target function of the vehicle in the real-time scenario.

2. The method according to claim 1, characterized in that The initial scene data includes multiple scene parameters. The parameter dimensionality reduction process is performed on the initial scene data to obtain the target scene data including: Performing variance decomposition calculation on the multiple scenario parameters to obtain failure-sensitive data corresponding to the multiple scenario parameters, wherein the failure-sensitive data is used to evaluate the degree of influence of parameter changes of the multiple scenario parameters on the probability of functional failure; selecting, according to the failure-sensitive data, some scenario parameters from the plurality of scenario parameters; The target scene data is generated using the partial scene parameters.

3. The method according to claim 2, characterized in that The failure sensitive data includes a plurality of sensitivity indexes. The variance decomposition calculation is performed on the plurality of scenario parameters to obtain the failure sensitive data corresponding to the plurality of scenario parameters respectively, including: Calculating the multiple scene parameters based on a variance-sensitive Sobol index method to obtain an overall variance and decomposed variances corresponding to the multiple scene parameters respectively; For a target scene parameter among the multiple scene parameters, a target sensitivity index corresponding to the target scene parameter among the multiple sensitivity indices is obtained according to the overall variance and the decomposed variance corresponding to the target scene parameter.

4. The method according to claim 1, wherein Based on the target failure probability distribution, adaptively sampling the target scene data to obtain the target sampling result includes: Sampling the target scene data according to the initial sampling distribution to obtain an initial sampling result; The initial sampling result is adaptively and iteratively updated according to the target failure probability distribution to obtain the target sampling result, wherein the degree of similarity between the sampling distribution of the target sampling result and the target failure probability distribution meets a target condition.

5. The method according to claim 4, characterized in that Adaptively iteratively updating the initial sampling result according to the target failure probability distribution to obtain the target sampling result includes: In a current iteration round of the adaptive iterative update process, using a previous round sampling result corresponding to a previous iteration round and the target failure probability distribution, updating the previous round sampling distribution corresponding to the previous round sampling result to the current sampling distribution; The target scene data is sampled according to the current sampling distribution to obtain a current sampling result corresponding to the current iteration round.

6. The method according to claim 5, characterized in that Updating the front wheel sampling distribution to the current sampling distribution using the front wheel sampling result and the target failure probability distribution includes: Using the target failure probability distribution, weight calculation is performed on multiple sampling samples in the previous round sampling result to obtain sample weights corresponding to the multiple sampling samples respectively; Selecting a target sample from the plurality of sampling samples according to the plurality of sample weights; Perform distribution fitting processing on the target sample to obtain the current sampling distribution corresponding to the current iteration round.

7. The method according to claim 1, characterized in that The vehicle functional failure assessment method further includes: Acquire test case data corresponding to the vehicle, wherein the test case data is used to characterize the test scenario in which the vehicle is located; Performing scenario testing on the test case data to obtain functional performance data of the vehicle, wherein the functional performance data is used to evaluate the failure probability of the target function of the vehicle in the scenario to be tested; The target failure probability distribution is constructed using the test case data and the functional performance data.

8. The method according to claim 1, characterized in that The evaluation calculation is performed based on the target sampling result to obtain the functional failure probability of the vehicle in the real-time scenario, including: Determining failure indication data corresponding to the target sampling result using the functional performance data corresponding to the target failure probability distribution; The functional failure probability is obtained by performing calculations using the failure indication data, the target failure probability distribution, and the sampling distribution of the target sampling results.

9. A vehicle failure assessment system, characterized in that: include: an acquisition unit, configured to acquire initial scene data corresponding to the vehicle, wherein the initial scene data is used to represent a real-time scene in which the vehicle is located; a processing unit, configured to perform parameter dimensionality reduction processing on the initial scene data to obtain target scene data; a sampling unit, configured to adaptively sample the target scenario data based on a target failure probability distribution to obtain a target sampling result, wherein the target failure probability distribution is used to characterize the distribution of occurrence scenario information corresponding to a functional failure event of the vehicle; An evaluation unit is used to perform evaluation calculations based on the target sampling results to obtain a functional failure probability of the vehicle in the real-time scenario.

10. A vehicle, characterized in that: Including, on-board processor and on-board memory, wherein, The on-board memory is used to store computer programs; The on-board processor is used to execute the computer program stored in the memory to implement the vehicle functional failure assessment method according to any one of claims 1 to 8.

Images