Novel AEB site test case selection method

Abstract

The invention discloses a novel AEB site test case selection method, and relates to the technical field of intelligent driving, and the method comprises the steps: building a test scene set through scene elements based on a test case set, data expansion, index expansion, case empowerment and case division, supplementing the scene set with the contents of accident information, user attention and the like, and forming a test case library; according to a user scene and an accident scene, test cases are selected, cases more emphasizing practice are provided, the incidence relation between the test cases and product requirements is defined, a test case table can be generated in a customized mode, disjunction of testing and development is avoided, the coverage degree of the test cases is improved, the test cases are made to fit the actual driving scene of a user, multiple use case selection dimensions are achieved, and the user experience is improved. And site testing, process management and product development are facilitated.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and in particular to a novel method for selecting AEB site test cases. Background Art

[0002] In the early 1980s, with the advancement of artificial intelligence technology and the upgrading of hardware equipment such as sensors, smart driving cars gradually transformed from a forward-looking research discipline to an applied discipline. As the market evolves, specific requirements are put forward for technological development. First, users are getting younger, and their attention to technologies such as smart driving functions is increasing, requiring high quality of product development and development process quality; second, the penetration rate of smart driving function models has increased, and the testing tasks have become heavier, which has brought great challenges to the design of test plans and the implementation of field tests on schedule. The existing verification model is mainly achieved through the design of typical test conditions, the expansion of industry specification conditions, the manual extraction of use cases, the construction of field scenarios, and real vehicle testing. The existing verification model has the following problems:

[0003] 1. Typical test conditions collect road data, cluster scene elements such as vehicle and target motion status, extract elements with high frequency, and reconstruct the scene. Although they are representative, there are few extreme scenes and occasional scenes, and there are no clustering conditions. In addition, the vehicle speed and target relationship between different working conditions are coupled by exhaustive method, which does not conform to the actual situation of users and causes waste of test resources.

[0004] 2. The existing method of extracting use cases requires multiple screening and comparison of known content, which is inefficient and easily affected by subjective factors. The extracted use case set is relatively scattered, without selection rules that are adaptable to different needs, and is mostly based on regulations. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention conceives a novel AEB field test case selection method, which constructs a test scenario set through scenario elements, supplements the scenario set with accident information, user concerns and other contents to form a test case library, and determines the case selection method through demand input analysis.

[0006] The technical solution used to implement the present invention is: a new AEB site test case selection method, which is characterized in that it includes the following steps:

[0007] 1) Build a test scenario set:

[0008] According to the logical scenarios listed by the scenario elements, complete the test scenario collection; exhaustively arrange the collected test scenario elements and list the test scenario elements completely; formulate evaluation indicators for the test scenario, and the test scenario indicators include: whether there is a collision, motion relationship, distance, triggering TTC, and collision relative speed;

[0009] 2) Build a test case library:

[0010] On the basis of the test scenario set, user data analysis results, accident data analysis results, regulatory item tags, use case indicator calculations, and use case comprehensive scoring logic are added to form a test case library, and feature columns are added to the test case set, and the feature columns include user tags, accident tags, regulatory tags, indicator tags, and use case scoring tags;

[0011] 3) Determine the test case selection logic:

[0012] For the millimeter wave radar solution, the environmental factors and features in the scene elements are filtered, and the database tool is used to remove the duplication and then select to form a test case table. The selection requirements are logically classified as follows:

[0013] 3.1) Verification: no explicit input, meeting regulatory requirements, better than the benchmark vehicle, better than the original model;

[0014] 3.2) Development category: similar use case search and boundary case testing.

[0015] Furthermore, in step 1), the scene elements include: driving behavior of the vehicle, number, characteristics, motion relationship of target objects, environmental factors, and road facilities.

[0016] Further, in step 2), the user data analysis result is obtained by installing a data acquisition device on a road test vehicle simulating a user, collecting data containing road scene element information, including video data, bus data, and Ethernet data, performing unified time settings, and segmenting and packaging the data according to time. The packaged data is used as a scene input, and the use cases are counted according to the features in the test case set to obtain the probability of occurrence of the test case scenario.

[0017] Further, in step 2), the accident data analysis is to check the accident list input according to key fields, and the key fields include: degree of casualties, property losses, and on-site scene description. According to the on-site scene description, the degree of casualties is statistically analyzed according to the features in the corresponding test case set. If a test case in the test case set has major casualties and property losses in the accident scene set, and these occur frequently, then the accident severity of the test case should be high.

[0018] Furthermore, in step 2), the use case indicator is calculated by merging the historical test results of each use case to obtain a result set of each indicator under each test case.

[0019] Further, in step 3.1), the verification class selection logic is as follows:

[0020] a) No explicit input: After removing duplicates from the self-vehicle driving behavior and motion relationship in the scenario elements, a list of scenarios that are unrelated to the self-vehicle speed and the speed of the preceding vehicle is obtained. A round of testing is performed according to the scenario list. For the passed items, the test cases are supplemented according to the self-vehicle driving behavior and motion relationship in the test case library to form a final test case table;

[0021] b) Meet regulatory requirements: In the Regulatory tab, select the corresponding regulations, such as C-NCAP (2024) and C-IASI, to directly form the final test case table;

[0022] c) Better than the benchmark vehicle: Only benchmark vehicles with the same hardware solutions are discussed, otherwise, the selection is based on step a);

[0023] d) Better than the original model: Only the optimization of the hardware solution is discussed. Otherwise, follow the steps in a) to select the failed test cases in the test case table of the original model. If the failed test case passes, select the test case with lower priority in the driving behavior and motion relationship of the vehicle in the scene elements; if the failed test case fails, select all the test cases in the scene where it is located.

[0024] Further, in step 3.2), the development category selection logic is as follows:

[0025] e) Search for similar use cases:

[0026] Same scenario search: compare the labels of test cases one by one, and require that all scenario elements are consistent;

[0027] Same type search: The scene elements only need to meet one or more of the same elements to be considered as similar use cases;

[0028] f) Boundary case testing: Use cases that are in the top 10% of accident severity and the bottom 10% of user scenario probability are defined as boundary cases and are screened according to the corresponding labels.

[0029] The beneficial effects of the novel AEB site test case selection method of the present invention are embodied in:

[0030] A novel method for selecting AEB field test cases is based on a test case set, data expansion, indicator expansion, case empowerment, and case division. A test scenario set is constructed through scenario elements, and accident information, user concerns, and other content are supplemented to the scenario set to form a test case library. Test cases are selected based on user scenarios and accident scenarios, and more practical use cases are provided. The relationship between test cases and product requirements is defined, and a test case table can be customized to avoid disconnection between testing and development, improve test case coverage, and make test cases fit users' actual driving scenarios. It has multiple use case selection dimensions and is convenient for field testing, process management, and product development. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a flow chart of a new method for selecting AEB field test cases. DETAILED DESCRIPTION

[0032] The following is combined with Figure 1 The present invention will be further described in detail with reference to the specific implementation methods. To make the purpose, technical solutions and advantages of the implementation methods clearer, the technical solutions in the implementation methods will be clearly and completely described in combination with the embodiments of the present invention. The specific implementation methods described herein are only used to explain the present invention and are not used to limit the scope of the present invention.

[0033] A new method for selecting AEB field test cases The selected AEB test case set is shown in Table 1.

[0034] Table 1

[0035]

[0036] As attached Figure 1 As shown in the figure, a new AEB site test case selection method is used to construct a test scenario set through scenario elements; the scenario set is supplemented with accident information, user concerns and other contents to form a test case library, and the case selection method is determined through demand input analysis. It includes the following steps:

[0037] 1) Build a test scenario set:

[0038] According to the logical scenarios listed by the scenario elements, the test scenario collection is completed, as shown in Table 2; the collected test scenario elements are exhaustively arranged to list the test scenario elements completely; the evaluation indicators of the test scenario are formulated, and the test scenario indicators include: whether there is a collision, motion relationship, distance, triggering TTC, collision relative speed, and the scenario elements include: self-driving behavior, target number, characteristics, motion relationship, environmental factors, and road facilities;

[0039] Table 2

[0040]

[0041] 2) Build a test case library:

[0042] On the basis of the test scenario set, user data analysis results, accident data analysis results, regulatory item tags, use case indicator calculation, and use case comprehensive scoring logic are added to form a test case library, and feature columns are added to the test case set, and the feature columns include user tags, accident tags, regulatory tags, indicator tags, and use case scoring tags; user data analysis results, accident data analysis results, regulatory item tags, use case indicator calculation, and use case comprehensive scoring logic are determined as follows:

[0043] 2.1) User data analysis: By installing data acquisition equipment on the road test vehicle simulating the user, data containing information on road scene elements, including video data, bus data, and Ethernet data, are collected, and a unified time setting is performed, and the data is segmented and packaged according to time. The packaged data is used as the scene input, and the use cases are counted according to the characteristics of the test case set to obtain the probability of occurrence of the test case scene; 2.2) Accident data analysis: The accident list input is checked according to the key fields. The key fields must include: degree of casualties, property losses, and on-site scene descriptions. According to the on-site scene descriptions, the degree of casualties is counted according to the characteristics of the test case set. For example, if a test case in the test case set has major casualties and property losses in the accident scene set, and these occur frequently, then the accident severity of the test case should be high. 2.3) Regulatory item marking: Mark the test case as subordinate to the regulations.

[0044] 2.4) Use case indicator calculation: By merging the historical test results of each use case, the result set of each indicator under each test case is obtained, as shown in Table 3.

[0045] Table 3

[0046]

[0047] 2.5) Combined with the scenario occurrence probability and accident severity, the test case is given a comprehensive score. The comprehensive score of a single test case = user scenario occurrence probability × user weight + accident severity × accident weight. User weight and accident weight are defined by the product input and sum to 1, as shown in Table 4.

[0048] Table 4

[0049]

[0050] 2.6) Prioritize based on own driving behavior and motion relationship, specifically: the priority of the vehicle with low speed is higher than that of the vehicle with high speed, the priority of the target with small deceleration is higher because the target deceleration is large, and the priority of the target vehicle being stationary is higher because the target vehicle is moving.

[0051] 3) Determine the test case selection logic:

[0052] For the millimeter wave radar solution, the environmental factors and features in the scene elements are filtered, and the database tool is used to remove the duplication and then select to form a test case table. The selection requirements are logically classified as follows:

[0053] 3.1) Verification: No explicit input, meeting regulatory requirements, better than the benchmark vehicle, better than the original vehicle; the selection logic of the verification category is as follows:

[0054] a) No explicit input: After removing duplicates from the self-vehicle driving behavior and motion relationship in the scenario elements, a list of scenarios that are unrelated to the self-vehicle speed and the speed of the preceding vehicle is obtained. A round of testing is performed according to the scenario list. For the passed items, the test cases are supplemented according to the self-vehicle driving behavior and motion relationship in the test case library to form a final test case table;

[0055] b) Meet regulatory requirements: In the Regulatory tab, select the corresponding regulations, such as C-NCAP (2024) and C-IASI, to directly form the final test case table;

[0056] c) Better than the benchmark vehicle: Only benchmark vehicles with the same hardware solutions are discussed, otherwise, the selection is based on step a);

[0057] d) Better than the original model: Only the optimization of the hardware solution is discussed. Otherwise, follow the selection in step a) and test the failed cases in the test case table of the original model. If the failed case passes, select the case with lower priority of the self-driving behavior and motion relationship in the scene elements; if the failed case fails, select all the cases in the scene where it is located. The rules for selecting test cases that are better than the original model are shown in Table 5.

[0058] Table 5

[0059]

[0060] 3.2) Development category: similar use case search, boundary case testing. The development category selection logic is as follows:

[0061] e) Search for similar use cases:

[0062] Same scenario search: compare the labels of test cases one by one, and require that all scenario elements are consistent;

[0063] Same type search: The scene elements only need to meet one or more of the same elements to be considered as similar use cases;

[0064] f) Boundary case testing: Use cases that are in the top 10% of accident severity and the bottom 10% of user scenario probability are defined as boundary cases and are screened according to the corresponding labels.

[0065] The above is only a preferred embodiment of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A new method for selecting AEB field test cases, characterized in that: It includes the following steps: 1) Build a test scenario set: According to the logical scenarios listed by the scenario elements, complete the test scenario collection; exhaustively arrange the collected test scenario elements and list the test scenario elements completely; formulate evaluation indicators for the test scenario, and the test scenario indicators include: whether there is a collision, motion relationship, distance, triggering TTC, and collision relative speed; 2) Build a test case library: On the basis of the test scenario set, user data analysis results, accident data analysis results, regulatory item tags, use case indicator calculations, and use case comprehensive scoring logic are added to form a test case library, and feature columns are added to the test case set, and the feature columns include user tags, accident tags, regulatory tags, indicator tags, and use case scoring tags; 3) Determine the test case selection logic: For the millimeter wave radar solution, the environmental factors and features in the scene elements are filtered, and the database tool is used to remove the duplication and then select to form a test case table. The selection requirements are logically classified as follows: 3.1) Verification: no explicit input, meeting regulatory requirements, better than the benchmark vehicle, better than the original model; 3.2) Development: similar use case search and boundary case testing.

2. According to the novel AEB site test case selection method of claim 1, it is characterized in that: In step 1), the scene elements include: driving behavior of the vehicle, number, characteristics, motion relationship of target objects, environmental factors, and road facilities.

3. According to the novel AEB site test case selection method of claim 1, it is characterized in that: In step 2), the user data analysis result is obtained by installing a data acquisition device on a road test vehicle simulating a user, collecting data containing road scene element information, including video data, bus data, and Ethernet data, performing unified time settings, and segmenting and packaging the data according to time. The packaged data is used as a scene input, and the test case count is performed according to the features in the test case set to obtain the probability of occurrence of the test case scenario.

4. According to the novel AEB site test case selection method of claim 1, it is characterized in that: In step 2), the accident data analysis is to check the accident list input according to key fields, and the key fields include: casualty level, property loss, and on-site scene description. According to the on-site scene description, the casualty level statistics are performed corresponding to the features in the test case set. If a test case in the test case set has heavy casualties and property losses in the accident scene set, and these occur frequently, then the accident severity of the test case should be high.

5. The novel AEB site test case selection method according to claim 1 is characterized in that: In step 2), the use case indicator is calculated by merging the historical test results of each use case to obtain a result set of each indicator under each test case.

6. The novel AEB site test case selection method according to claim 1 is characterized in that: In step 3.1), the verification class selection logic is as follows: a) No explicit input: After removing duplicates from the self-vehicle driving behavior and motion relationship in the scenario elements, a list of scenarios that are unrelated to the self-vehicle speed and the speed of the preceding vehicle is obtained. A round of testing is performed according to the scenario list. For the passed items, the test cases are supplemented according to the self-vehicle driving behavior and motion relationship in the test case library to form a final test case table; b) Meet regulatory requirements: In the Regulatory tab, select the corresponding regulations, such as C-NCAP (2024) and C-IASI, to directly form the final test case table; c) Better than the benchmark vehicle: Only benchmark vehicles with the same hardware solutions are discussed, otherwise, the selection is based on step a); d) Better than the original model: Only the optimization of the hardware solution is discussed. Otherwise, follow the steps in a) to select the failed test cases in the test case table of the original model. If the failed test case passes, select the test case with lower priority in the driving behavior and motion relationship of the vehicle in the scenario elements. If the invalidation test case fails, all test cases in the scenario are selected.

7. The novel AEB field test case selection method according to claim 1 is characterized in that: In step 3.2), the development category selection logic is as follows: e) Similar use case search: Same scenario search: compare the labels of test cases one by one, and require that all scenario elements are consistent; Same type search: The scene elements only need to meet one or more of the same elements to be considered as similar use cases; f) Boundary case testing: Use cases that are in the top 10% of accident severity and the bottom 10% of user scenario probability are defined as boundary cases and are screened according to the corresponding labels.

Images