Advanced driving assistance system safety assessment method, device, medium and product

Abstract

The invention discloses an advanced driving assistance system safety assessment method and device, a medium and a product, and relates to the technical field of safety assessment, and the method comprises the steps: obtaining an observation data point of an advanced driving assistance system, building an agent model of the advanced driving assistance system based on the observation data point and combining a logit function and a logistic function based on a Gaussian process, and carrying out the safety assessment of the advanced driving assistance system. According to the method, a proxy model is constructed, a predicted value of the proxy model is obtained, three acquisition functions are adopted, an optimal candidate scene is determined based on a trust region search algorithm, and a safety evaluation result is determined according to all the optimal candidate scenes and corresponding test results. According to the method, the test efficiency of a safety evaluation algorithm can be improved, and exploration of a risk scene set, safe, convenient and accurate identification of a system and estimation of a fault probability can be completed at the same time.

Description

Technical Field

[0001] The present application relates to the technical field of safety assessment, and particularly to a method, device, medium and product for safety assessment of an advanced driver assistance system. Background Art

[0002] With the rapid development of advanced driver assistance system technology, the degree of vehicle automation and intelligence has been continuously improved, and the safety of vehicles has attracted much attention. However, ensuring the safety of vehicles is a task with great challenges. For example, even in seemingly simple scenarios such as a vehicle driving at high speed, a vehicle equipped with an adaptive cruise control system is easily affected by the driving behavior of the vehicle in front, and various acceleration and deceleration behaviors of the vehicle in front may bring serious traffic accidents.

[0003] Currently, the methods for evaluating whether a vehicle equipped with an advanced driver assistance system meets certain standards usually use a large number of test cases and traffic scenarios to conduct simulation tests on the vehicle under test. However, such methods have a significant drawback, that is, a large number of test cases or traffic scenarios may miss some unsafe test conditions. For example, various test cases or traffic scenarios can be quickly and automatically generated by using Monte Carlo simulation, but such methods are not specifically designed to search for scenarios that cause the vehicle to enter an unsafe state. In contrast, formal verification methods can provide strong safety guarantees by verifying whether each action of the vehicle meets the safety specifications. However, it is difficult to extend formal verification methods to continuous and complex advanced driver assistance systems. For example, model checking can only have efficient test performance in simple and discrete systems.

[0004] In order to efficiently expose the unsafe states existing in advanced driver assistance systems, currently, a safety assessment algorithm optimized based on black-box sampling is widely used in the safety assessment of advanced driver assistance systems, including fault exploration based on evolutionary search and surrogate models. In the safety assessment algorithm optimized based on black-box sampling, metric temporal logic or signal temporal logic formulas are usually used to describe the safety specifications of advanced driver assistance systems. In addition, the robustness of temporal logic defines the degree to which an advanced driver assistance system meets the safety specifications. The falsification algorithm optimized based on black-box sampling searches for optimal test cases by minimizing the system robustness value. In this way, the system robustness value can guide the falsification algorithm to forge test cases that force the system to fail, so that the state where the advanced driver assistance system does not meet the safety specifications can be detected more effectively and automatically. However, the existing safety assessment algorithm based on evolutionary search needs to randomly generate a large number of offspring during each iteration optimization process and test them using a real autonomous driving intelligent system, which requires extremely high computing resources. In addition, in the safety assessment algorithm based on surrogate models, a large number of candidate test scenarios generated by grid search or uniform distribution sampling still require a large amount of computing power when selecting the optimal candidate test scenario. In addition, the existing methods cannot be used to accurately identify the safety boundaries of intelligent vehicles and estimate the failure probability of the system under test.

[0005] Therefore, to solve the above technical problems, it is urgent to provide a new safety assessment method for advanced driver assistance systems, which can improve the test efficiency of the safety assessment algorithm and simultaneously complete the exploration of the risk scenario set, the accurate identification of the system safety boundary, and the estimation of the failure probability. Summary of the Invention

[0006] The purpose of this application is to provide a safety assessment method, device, medium, and product for advanced driver assistance systems, which can improve the test efficiency of the safety assessment algorithm and simultaneously complete the exploration of the risk scenario set, the accurate identification of the system safety boundary, and the estimation of the failure probability.

[0007] To achieve the above purpose, this application provides the following solutions:

[0008] In the first aspect, this application provides a safety assessment method for advanced driver assistance systems, including:

[0009] Obtain the observation data points of the advanced driver assistance system; the observation data points include: test scenarios and corresponding test results;

[0010] Based on the observation data points, establish a surrogate model of the advanced driver assistance system by combining a Gaussian process with a logit function and a logistics function;

[0011] According to the predicted values of the surrogate model, three acquisition functions are used to determine the optimal candidate scenarios based on the trust-region search algorithm; the three acquisition functions are: the coverage exploration acquisition function, the boundary recognition acquisition function, and the fault sampling acquisition function; the predicted values of the surrogate model are: the mapping results of the Gaussian process predicted values using the logistic function;

[0012] Test the advanced driver assistance system according to all the optimal candidate scenarios to obtain the test results;

[0013] Determine the safety assessment results according to the optimal candidate scenarios and the corresponding test results; the safety assessment results include: risk scenarios and their corresponding fault modes, the scenario with the highest probability, and the fault probability of the advanced driver assistance system.

[0014] Optionally, the establishment of the surrogate model of the advanced driver assistance system based on the Gaussian process combined with the logit function and the logistics function according to the observed data points specifically includes:

[0015] Use the formula to determine the kernel function matrix K(X 1:n ,X 1:n ) of the observed data points;

[0016] Use the formula to determine the elements K(X i , X j ) in the kernel function matrix;

[0017] Use the formula to map the test results y i of the observed data points to the real number interval;

[0018] Update the hyperparameter η of the surrogate model by maximizing the likelihood function of the mapped observed data points ;

[0019] where X 1:n is the test scenario set vector, X 1:n = [X1, X2, X3,..., X n T , the superscript T is the transpose, n is the number of test scenarios, X i is the i-th test scenario, X j is the j-th test scenario, exp() is a mathematical function, X is the parameter space of the test scenario, y i is the test result corresponding to the i-th test scenario, z i is the logit function value, is used to clearly define the logit function, the parameter ε = 10 -5 , the parameter s is used to adjust the steepness of the s-shaped curve, and the parameter s = 10-1​ And ρ are the parameters of the kernel function, representing the amplitude and scale of the kernel function respectively.

[0020] Optionally, after establishing the surrogate model of the advanced driver assistance system based on the Gaussian process combined with the logit function and the logistics function according to the observed data points, it further includes:

[0021] Using the formula To determine the predicted value of the surrogate model; where, Is the prediction result of the surrogate model, Is the logistic function, used to map the prediction of the Gaussian process To [0, 1], Is the predicted value of the Gaussian process, GP is the Gaussian process, X i Is the test scenario, Is the predicted variance of the Gaussian process, and x′ is the test scenario to be predicted.

[0022] Optionally, according to the predicted value of the surrogate model, using three acquisition functions, based on the trust region search algorithm, to determine the optimal candidate scenario, specifically including:

[0023] Initializing the trust regions of the three acquisition functions and the corresponding trust region center points based on the trust region search algorithm;

[0024] Determining the three acquisition functions according to the trust region center points respectively;

[0025] Determining the optimal candidate scenarios in the trust regions corresponding to the three acquisition functions according to the predicted value of the surrogate model and the trust region center points respectively;

[0026] Updating the trust region length according to the trust region center point and the current optimal candidate scenario.

[0027] Optionally, determining the three acquisition functions according to the trust region center points respectively, specifically including:

[0028] Using the formula To determine the coverage exploration acquisition function;

[0029] Using the formula To determine the boundary recognition acquisition function;

[0030] Using the formula To determine the fault sampling acquisition function;

[0031] Where, x1 is the coverage exploration acquisition function, x2 is the boundary recognition acquisition function, and x3 is the fault sampling acquisition function, Denote the predicted standard deviation of the surrogate model, max is the maximum value of the predicted standard deviation of the surrogate model, min is the minimum value of the predicted standard deviation of the surrogate model, X is the parameter space of the test scenario, λ is used to coordinate the exploration and exploitation of the boundary scenario, h'(X i ) is the predicted mean of the Gaussian process, p(X i ) is the probability model of the test scenario, t is the iteration number of the test algorithm, X i is the test scenario, argmin and argmax respectively represent taking the minimum value and the maximum value.

[0032] Optionally, updating the trust region length according to the trust region center point and the optimal candidate scenario specifically includes:

[0033] If the current optimal candidate scenario is better than the trust region center point then use the formula l i = min(L max , 2l) to increase the trust region length, and the expression of the trust region center point is

[0034] If the trust region center point is better than the current optimal candidate scenario then use the formula l i = l / 2 to reduce the trust region length, and the expression of the trust region center point is

[0035] where, l i is the current trust region length, L max is the maximum length of the trust region, m is the iteration number of the trust region search, i is the subscript of the trust region, and l is the length of the trust region.

[0036] Optionally, determining the safety assessment result according to the optimal candidate scenario and the corresponding test result specifically includes:

[0037] Use the formula X fail = {x i |x i ∈X, y i ∈Y, δ{y i > 0.5}}, calculate all risk scenarios and the scenario with the maximum probability; where, X represents the parameter space of the test scenario, Y represents the set of test results, X i represents the test scenario, y i represents the test result, X fail represents the set of risk scenarios, X * represents the scenario with the maximum probability, p(x i ) represents the probability distribution of the scenario, δ(·) is the indicator function, when the test result yi > 0.5, X fail has a value of 1, otherwise X fail has a value of 0;

[0038] Using the formula to calculate the failure probability of the advanced driver assistance system; where represents the failure probability of the advanced driver assistance system, represents the predicted value of the surrogate model, n is the number of test scenarios, q(x i ) is the probability of the test scenario under the proposed distribution, when has a value of 1, otherwise has a value of 0.

[0039] In a second aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the steps of the above-mentioned method for evaluating the safety of an advanced driver assistance system.

[0040] In a third aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the above-mentioned method for evaluating the safety of an advanced driver assistance system.

[0041] In a fourth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the steps of the above-mentioned method for evaluating the safety of an advanced driver assistance system.

[0042] According to the specific embodiments provided by the present application, the present application has the following technical effects:

[0043] The present application provides a method, device, medium, and product for evaluating the safety of an advanced driver assistance system. According to the test scenarios and corresponding test results of the advanced driver assistance system, a surrogate model of the advanced driver assistance system is established based on a Gaussian process combined with a logit function and a logistics function, and then the predicted value of the surrogate model is obtained, which can predict the test results of the advanced driver assistance system. According to the predicted value of the surrogate model, a coverage exploration acquisition function, a boundary recognition acquisition function, and a failure acquisition function are used, and the optimal candidate scenario is determined through a trust region search algorithm, which can improve the test efficiency of the system; the advanced driver assistance system is tested through all the optimal candidate scenarios to obtain the test results, and the safety evaluation result is determined according to the optimal candidate scenarios and the corresponding test results. The present application can be applied to a variety of different systems under test, can improve the test efficiency of the safety evaluation algorithm, and at the same time complete the exploration of the risk scenario set, the accurate identification of system safety and convenience, and the estimation of failure probability. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0045] Figure 1 It is a flowchart of a method for safety assessment of an advanced driver assistance system in an embodiment of the present application;

[0046] Figure 2 It is a framework diagram of a method for safety assessment of an advanced driver assistance system provided in an embodiment of the present application;

[0047] Figure 3 It is a schematic structural diagram of a computer device provided in an embodiment of the present application. Detailed implementation manners

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0049] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the drawings and specific implementation manners.

[0050] As Figure 1 shown, a method for safety assessment of an advanced driver assistance system provided in an embodiment of the present application includes:

[0051] S101: Obtain the observation data points of the advanced driver assistance system; the observation data points include: test scenarios and corresponding test results.

[0052] S102: Based on the observation data points, establish a surrogate model of the advanced driver assistance system by combining the Gaussian process with the logit function and the logistics function.

[0053] S103: According to the predicted values of the surrogate model, use three acquisition functions and based on the trust region search algorithm, determine the optimal candidate scenarios; the three acquisition functions are: coverage exploration acquisition function, boundary recognition acquisition function, and fault sampling acquisition function; the predicted values of the surrogate model are: the mapping results of the predicted values of the Gaussian process using the logistics function. The optimal candidate test scenario search algorithm based on the trust region is based on the predicted values of the surrogate model Select the optimal candidate test scenarios (x1, x2, x3) with three acquisition functions. The main process includes initializing the trust region, surrogate model prediction, trust region center selection, and trust region length update.

[0054] S104: Test the advanced driver assistance system according to all the optimal candidate scenarios to obtain test results.

[0055] S105: Determine the safety assessment results based on the optimal candidate scenarios and the corresponding test results; the safety assessment results include: risk scenarios and their corresponding failure modes, the scenario with the highest probability, and the failure probability of the advanced driver assistance system.

[0056] In an exemplary embodiment, before the above S102, the method may further include:

[0057] Initialize the observed data points; randomly generate k test scenarios x i , and test the system under test to obtain test results y i , i = 1, 2, 3,..., k.

[0058] Specifically, S102 includes:

[0059] S201: Use the formula to determine the kernel function matrix K(X 1:n , X 1:n ) of the observed data points.

[0060] S202: Use the formula to determine the elements K(X i , X j ) in the kernel function matrix.

[0061] S203: Use the formula to map the test results y i of the observed data points to the real number interval.

[0062] S204: Update the hyperparameter η of the surrogate model by maximizing the likelihood function of the mapped observed data points .

[0063] Among them, X 1:n is the test scenario set vector, X 1:n = [X1, X2, X3,..., X n T , the superscript T is the transpose, n is the number of test scenarios, X i is the i-th test scenario, X j is the j-th test scenario, exp() is a mathematical function, X is the parameter space of the test scenario, y i ​is the test result corresponding to the i-th test scenario, z i is the logit function value, used to clearly define the logit function, with parameter ε = 10 -5 , parameter s is used to adjust the steepness of the sigmoid curve, parameter s = 10-1, and ρ are the parameters of the kernel function, representing the amplitude and scale of the kernel function respectively.

[0064] In an exemplary embodiment, first, define the probability distribution of a function f ~ GP(f; m(X 1:n ), K(X 1:n , X 1:n )); where X 1:n = [X1, X2, X3,..., X n T represents the test scenario, m(X 1:n ) is the mean function, and K(x 1:n , x 1:n ) is the kernel function matrix, expressed as The elements in the matrix are determined by the squared exponential kernel function K(X , X i , X j ), 1 ≤ i, j ≤ n, as follows: The squared exponential kernel function can represent the correlation between data points. The hyperparameters of the kernel function can be obtained by maximizing the likelihood p(y |η) of the observed data points 1:n , where y 1:n represents the test result, is the hyperparameter of the kernel function. Among them, y i is the observed value and also the test result of the system under test. Secondly, use the logit function to map the y in the observed data points of the system under test i to the real number interval, as follows: Then, this application updates the hyperparameter η of the surrogate model by maximizing the likelihood function of the observed data points .

[0065] In another exemplary embodiment of this application, before the above S103, the method may further include:

[0066] Use the formula to determine the predicted value of the surrogate model; where is the predicted result of the surrogate model, is the logistic function, used to map the prediction of the Gaussian process to [0, 1], ​is the predicted value of the Gaussian process, GP is the Gaussian process, and X i is the test scenario, is the predicted variance of the Gaussian process, and x′ is the test scenario to be predicted.

[0067] Among them, S103 specifically includes:

[0068] S301: Initialize the trust regions of the three acquisition functions and the corresponding trust region center points based on the trust region search algorithm. The trust region search algorithm initializes three trust regions, including initializing the center point and lengths l1, l2, l3, as well as the iteration number m = 1 and the maximum iteration number M. Among them, the initial center point of the trust region is determined by multiple randomly initialized candidate scenarios and the prediction results of the surrogate model to determine.

[0069] S302: Determine the three acquisition functions according to the trust region center points respectively.

[0070] S303: Determine the optimal candidate scenarios in the trust regions corresponding to the three acquisition functions according to the predicted values of the surrogate model and the trust region center points respectively. Since the test algorithm contains three acquisition functions, therefore, the trust region search algorithm correspondingly also includes three corresponding trust regions.

[0071] S304: Update the trust region length according to the trust region center point and the current optimal candidate scenario.

[0072] Specifically, S302 includes:

[0073] S3021: Use the formula to determine the coverage exploration acquisition function.

[0074] S3022: Use the formula to determine the boundary recognition acquisition function.

[0075] S3023: Use the formula to determine the fault sampling acquisition function.

[0076] Among them, x1 is the coverage exploration acquisition function, x2 is the boundary recognition acquisition function, x3 is the fault sampling acquisition function, represents the predicted standard deviation of the surrogate model, max is the maximum value of the predicted standard deviation of the surrogate model, min is the minimum value of the predicted standard deviation of the surrogate model, X is the parameter space of the test scenario, λ is used to coordinate the exploration and exploitation of the boundary scenarios, h'(X i ) is the predicted mean of the Gaussian process, p(X i ) is the probability model of the test scenario, t is the iteration number of the test algorithm, X iFor the test scenario, argmin and argmax respectively represent taking the minimum value and the maximum value.

[0077] In an exemplary embodiment, when selecting the trust region center, the first two trust region center points are selected according to the coverage exploration and boundary recognition acquisition functions. The coverage exploration acquisition function is expressed as: The boundary recognition acquisition function is expressed as: In the formula, is the differential form of the logistics function and also the analytical form of the system safety boundary. λ coordinates the exploration and exploitation of the boundary scenario. p(X i ) represents the true probability of the test scenario, p(X i ) 1 / t represents first selecting the boundary scenarios with high probabilities, and as the number of iteration steps t increases, gradually exploring all possible boundary scenarios. Since the failure sampling acquisition function is distributed sampling, the failure sampling acquisition function is expressed as: In the formula, represents the upper bound of the confidence predicted by the surrogate model. δ(·) represents the indicator function. When is greater than 0.5, x is a risk scenario and is sampled with a certain natural probability p(X i ), and p(X i ) ensures the rationality of the risk scenario. Otherwise, x is a safe scenario and the sampling probability is 0, so it will not be sampled. There is no selection of the optimal value for this acquisition function. Here, the scenario that makes the standard deviation of the surrogate model prediction the smallest is used as the third trust region center point, expressed as:

[0078] Specifically, S304 specifically includes:

[0079] If the current optimal candidate scenario is better than the trust region center point then the trust region length is increased using the formula l i = min(L max , 2l), and the expression of the trust region center point is

[0080] If the trust region center point is better than the current optimal candidate scenario then the trust region length is reduced using the formula l i = l / 2, and the expression of the trust region center point is

[0081] Among them, l i is the current trust region length, L max is the maximum length of the trust region, m is the number of iterations of the trust region search, i is the subscript of the trust region, and l is the length of the trust region.

[0082] The trust region search algorithm enters iterative search. In the iterative search, the search algorithm first randomly generates q candidate scenarios in each trust region. Then, according to the predicted values of their corresponding surrogate models and the selection rules of the trust region center, the optimal candidate scenario in the trust region is selected. Finally, according to the trust region center point and the current optimal candidate scenario Update the trust region length l i , if the new optimal candidate scenario is better than the trust region center point, increase the trust region length, l i = min(L max , 2l), and use these optimal candidate scenarios as the new trust region center Otherwise, reduce the trust region length, l i = l / 2, and keep the original trust region center. After M iterations of the entire search process, the third trust region center generates a risk scenario according to the fault sampling acquisition function as the third optimal candidate scenario, and finally outputs the final optimal candidate scenario {x1, x2, x3}.

[0083] In an exemplary embodiment, S105 includes the following steps S1051 to S1052:

[0084] S1051: Use the formula X fail = {x i |x i ∈ X, y i ∈ Y, δ{y i > 0.5}}, Calculate all risk scenarios and the scenario with the maximum probability; where, X represents the parameter space of the test scenario, Y represents the test result set, X i represents the test scenario, y i represents the test result, X fail represents the risk scenario set, X * represents the scenario with the maximum probability, p(x i ) represents the probability distribution of the scenario, δ(·) is the indicator function, when the test result y i > 0.5, the value of X fail is 1, otherwise the value of X fail is 0.

[0085] S1052: Use the formula Calculate the failure probability of the advanced driver assistance system; where, represents the failure probability of the advanced driver assistance system, represents the predicted value of the surrogate model, n is the number of test scenarios, q(xi ) is the probability of the test scenario under the proposed distribution. When the value is 1, otherwise the value is 0.

[0086] In an exemplary embodiment, the Bayesian safety assessment framework based on trust region search is as Figure 2 shown. First, the test algorithm initializes some observed data points and uses them to update the surrogate model. Second, based on the predicted values of the surrogate model and three acquisition functions, the test algorithm iteratively uses the trust region search algorithm to find three optimal candidate scenarios (x1, x2, x3) corresponding to the acquisition functions, where i represents the number of acquisition functions, and tests the system under test. Finally, after multiple iterations of the above process, three key tasks for the safety assessment of the system under test are completed according to the observed data points, including falsification, analysis of the most likely risk scenarios, and estimation of the unsafe probability of the intelligent vehicle based on importance sampling. In this framework, the surrogate model combines Gaussian processes, logit, and logistics functions, where the logit and logistics functions can obtain the analytical form of the safety boundary of the system under test. Additionally, the three acquisition functions are coverage exploration, boundary identification, and failure region sampling, to achieve rapid and sufficient exploration of the entire scenario parameter space, accurate identification of the safety boundary of the system under test, and generation of risk scenarios.

[0087] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal, and its internal structure diagram can be as Figure 3 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. The computer program, when executed by the processor, implements a method for safety assessment of an advanced driver assistance system.

[0088] Those skilled in the art can understand that Figure 3The structure shown is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0089] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0090] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0091] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0092] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0093] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0094] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0095] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. An advanced driver assistance system safety assessment method, characterized in that, The advanced driver assistance system safety assessment method includes: Obtaining the observation data points of the advanced driver assistance system; the observation data points include: test scenarios and corresponding test results; Based on the observation data points, establishing a surrogate model of the advanced driver assistance system by combining Gaussian process with logit function and logistics function; According to the predicted values of the surrogate model, using three acquisition functions and based on the trust region search algorithm, determining the optimal candidate scenarios; the three acquisition functions are: coverage exploration acquisition function, boundary recognition acquisition function, and fault sampling acquisition function; the predicted values of the surrogate model are: the mapping results of the Gaussian process predicted values using the logistics function; Testing the advanced driver assistance system according to all the optimal candidate scenarios to obtain test results; Determining the safety assessment results according to the optimal candidate scenarios and the corresponding test results; the safety assessment results include: risk scenarios and their corresponding failure modes, the scenario with the highest probability, and the failure probability of the advanced driver assistance system.

2. The safety evaluation method for an advanced driver assistance system according to claim 1, wherein The step of establishing a surrogate model of the advanced driver assistance system by combining Gaussian process with logit function and logistics function based on the observation data points specifically includes: Using the formula to determine the kernel function matrix K(X 1:n , X 1:n ); Using the formula to determine the element K(X i , X j ) in the kernel function matrix; Using the formula map the test result y of the observed data points i to the real number interval; Using the observed data points after maximizing the mapping to update the hyperparameter η of the surrogate model with the likelihood function; Among them, X 1:n is the test scenario set vector, and X 1:n = [X1, X2, X3,..., X n T , where the superscript T represents transpose, n is the number of test scenarios, X i is the i-th test scenario, X j is the j-th test scenario, exp() is a mathematical function, X is the parameter space of the test scenario, y i is the test result corresponding to the i-th test scenario, z i is the logit function value, which is used to clearly define the logit function, the parameter ε = 10 -5 , the parameter s is used to adjust the steepness of the sigmoid curve, the parameter s = 10-1, and ρ are the parameters of the kernel function, representing the amplitude and scale of the kernel function respectively.​ 3. The safety assessment method for an advanced driver assistance system according to claim 2, wherein After establishing a surrogate model of the advanced driver assistance system by combining Gaussian process with logit function and logistics function based on the observation data points, it further includes: Using the formula to determine the predicted value of the surrogate model; where is the prediction result of the surrogate model, is the logistic function, which is used to map the prediction of the Gaussian process to [0, 1], is the predicted value of the Gaussian process, GP is the Gaussian process, and X i is the test scenario, is the predicted variance of the Gaussian process, and x′ is the test scenario to be predicted.

4. The safety assessment method for an advanced driver assistance system according to claim 1, wherein The step of determining the optimal candidate scenarios using three acquisition functions and based on the trust region search algorithm according to the predicted values of the surrogate model specifically includes: Initializing the trust regions of the three acquisition functions and the corresponding trust region center points based on the trust region search algorithm; Determining the three acquisition functions respectively according to the trust region center points; Determining the optimal candidate scenarios in the trust regions corresponding to the three acquisition functions respectively according to the predicted values of the surrogate model and the trust region center points; Updating the trust region length according to the trust region center points and the current optimal candidate scenarios.

5. The safety assessment method for an advanced driver assistance system according to claim 4, wherein, The step of determining the three acquisition functions respectively according to the trust region center points specifically includes: Use the formula to determine the coverage exploration acquisition function; Use the formula to determine the boundary recognition acquisition function; Use the formula to determine the fault sampling acquisition function; Among them, x1 is the coverage exploration acquisition function, x2 is the boundary recognition acquisition function, and x3 is the fault sampling acquisition function. represents the prediction standard deviation of the surrogate model, max is the maximum value of the prediction standard deviation of the surrogate model, min is the minimum value of the prediction standard deviation of the surrogate model, X is the parameter space of the test scenario, λ is used to coordinate the exploration and exploitation of the boundary scenario, h'(X i ) is the predicted mean of the Gaussian process, p(X i ) is the probability model of the test scenario, t is the iteration number of the test algorithm, X i is the test scenario, argmin and argmax respectively represent taking the minimum value and the maximum value.

6. The safety assessment method for an advanced driver assistance system according to claim 4, characterized in that, The step of updating the trust region length according to the trust region center points and the optimal candidate scenarios specifically includes: If the current optimal candidate scenario is better than the trust region center point then use the formula l i = min(L max , 2l) to increase the trust region length. The expression for the trust region center point is If the trust region center point is better than the current optimal candidate scenario then use the formula l i = l / 2 to reduce the trust region length, and the trust region center point expression is where l i is the current trust region length, L max is the maximum length of the trust region, m is the number of iterations of the trust region search, i is the subscript of the trust region, and l is the length of the trust region.

7. The safety assessment method for an advanced driver assistance system according to claim 1, wherein Determining the safety assessment results according to the optimal candidate scenarios and the corresponding test results specifically includes: Using formula X fail ={x i |x i ∈X, y i ∈Y, δ{y i >0.5}}, calculate all risk scenarios and the scenario with the maximum probability; where X represents the parameter space of the test scenarios, Y represents the set of test results, X i represents a test scenario, y i represents a test result, X fal represents the set of risk scenarios, x * represents the scenario with the maximum probability, p(x i ) represents the probability distribution of the scenario, δ(·) is an indicator function, when the test result y i >0.5, X fail has a value of 1, otherwise X fail has a value of 0; Use the formula to calculate the failure probability of the advanced driver assistance system; where represents the failure probability of the advanced driver assistance system, represents the predicted value of the surrogate model, n is the number of test scenarios, q(x i ) is the probability of the test scenario under the proposed distribution. When the value is 1, otherwise the value is 0.

8. A computer device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the advanced driver assistance system safety assessment method according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the advanced driver assistance system safety assessment method according to any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the advanced driver assistance system safety assessment method according to any one of claims 1-7.

Images