CN-121996558-A - Intelligent driving real vehicle test case generation method and electronic equipment

CN121996558ACN 121996558 ACN121996558 ACN 121996558ACN-121996558-A

Abstract

The application relates to the technical field of intelligent driving tests, in particular to a method and a device for generating an intelligent driving real vehicle test case. The method comprises the steps of obtaining multi-source data, identifying target scene fragments from the multi-source data, extracting features of the target scene fragments to obtain feature information, inputting the feature information into a pre-trained scene generation model to obtain a test scene set, wherein the scene generation model is obtained by training on the basis of historical traffic scene data, performing risk assessment on each test scene in the test scene set to determine risk levels of each test scene, and sequencing the priority of each test scene according to the risk levels to generate a test case sequence. The method can effectively cover extreme weather, multi-target game and the like tail scenes, realize reasonable inclination of the test resources to high-risk scenes, and remarkably improve scene coverage capability, test efficiency and safety.

Inventors

ZHANG QIFENG
DING GUOLIANG
YANG NAN
XU RUI
WANG HONGXIN
WANG XING

Assignees

武汉江夏楚能汽车技术研发有限公司

Dates

Publication Date: 20260508
Application Date: 20260126

Claims (10)

1. The intelligent driving real vehicle test case generation method is characterized by comprising the following steps of: acquiring multi-source data and identifying a target scene segment from the multi-source data; Extracting features of the target scene segment to obtain feature information; Inputting the characteristic information into a pre-trained scene generation model to obtain a test scene set, wherein the scene generation model is trained based on historical traffic scene data; Performing risk assessment on each test scene in the test scene set, and determining the risk level of each test scene; And sequencing the priority of each test scene according to the risk level to generate a test case sequence.
2. The method of claim 1, wherein the multi-source data comprises at least two of real vehicle sensor data, high-precision map information, historical test logs, and traffic flow simulation data.
3. The method of claim 1, wherein identifying the target scene segment from the multi-source data comprises clustering the multi-source data using a density clustering algorithm, and using the clustered high-risk scene segment as the target scene segment.
4. The method of claim 1, wherein the characteristic information comprises at least one of environmental characteristic information, traffic participant status information, vehicle self status information, and scene constraint information.
5. The method of claim 1, wherein the scene generation model generates an impedance network or diffusion model, and wherein the set of test scenes comprises edge scenes including environmental parameters, traffic participant behavior parameters, and vehicle status parameters.
6. The method of claim 1, wherein performing risk assessment on each test scenario in the set of test scenarios to determine a risk level for each test scenario comprises: Extracting risk factors from each test scene, and carrying out normalization processing on the risk factors; calculating the contribution degree of each risk factor by adopting a feature importance analysis method; weighting calculation is carried out according to the contribution degree of each risk factor, so as to obtain a risk score; and dividing the risk grade according to the risk score.
7. The method of claim 1, further comprising, after generating the sequence of test cases: virtual simulation verification is carried out on the test case sequence, and test cases which do not meet the feasibility condition are filtered; performing real vehicle testing in stages according to scene complexity; And obtaining a real vehicle test result, comparing the real vehicle test result with expected behaviors, and updating parameters of the scene generation model according to the comparison result.
8. The method of claim 7, wherein the performing real-vehicle testing in stages according to scene complexity includes performing static obstacle testing, dynamic multi-objective interaction testing, and extreme condition testing in sequence.
9. The method of claim 7, wherein the security downgrade process is triggered in response to detecting an abnormal condition during the staged execution of the real vehicle test; the abnormal state includes brake response delay overrun or sensor perception confidence below a preset threshold.
10. An electronic device comprising a processor, a memory and a computer program stored in the memory and executable on the processor, wherein execution of the computer program by the processor causes the electronic device to perform the method of any one of claims 1-9.

Description

Intelligent driving real vehicle test case generation method and electronic equipment Technical Field The application relates to the technical field of intelligent driving tests, in particular to an intelligent driving real vehicle test case generation method and electronic equipment. Background With the rapid development of intelligent driving technology, the importance of real vehicle testing as a key link for verifying the safety and reliability of an automatic driving algorithm is increasingly highlighted. The current intelligent driving real vehicle testing field mainly adopts the following technical routes that firstly, a test case design method driven by manual experience is adopted, a test engineer refines classical scenes such as lane changing, sudden braking and the like based on driving experience and manually designs the test case, secondly, a fixed scene library is built through simulation software to carry out simulation test, and thirdly, a simple clustering or regression model is utilized to extract repeated scenes from historical data to serve as the test case. These approaches meet the basic test requirements to some extent, but as the complexity of intelligent driving systems increases, their limitations develop. In the aspect of scene coverage, the occurrence probability of tail scenes such as extreme weather, multi-target game and the like in an actual road environment is low, the scenes are difficult to completely identify and restore based on an empirical induction mode, and meanwhile, differences exist between a fixed simulation scene and an actual vehicle test environment in the aspects of sensor noise characteristics, road friction coefficients, weather illumination conditions and the like, so that a test result is deviated from an actual working condition. In terms of test efficiency, the artificial design test case has a longer period and needs repeated iterative adjustment, and is difficult to match with the fast updated rhythm of the intelligent driving algorithm version. In the aspect of dynamic adaptability, the traditional testing strategy is solidified, and the adjustment flexibility is insufficient when facing road condition change, vehicle performance difference or new problems found in the testing process. In terms of safety, the direct application of part of dangerous scenes to real vehicle testing has the risk of causing collision accidents. In the aspect of cost control, the rationality of the test scene design directly influences the test times and the resource consumption. In summary, how to improve the scene coverage capability, the test efficiency and the safety of the intelligent driving real vehicle test is a technical problem to be solved in the current field. Disclosure of Invention In view of the above, the embodiment of the application provides a method, a device and electronic equipment for generating an intelligent driving real vehicle test case, which can improve the scene coverage capability, the test efficiency and the safety of the intelligent driving real vehicle test. A first aspect of the embodiment of the application provides a method for generating an intelligent driving real vehicle test case, which comprises the following steps: acquiring multi-source data and identifying a target scene segment from the multi-source data; Extracting features of the target scene segment to obtain feature information; Inputting the characteristic information into a pre-trained scene generation model to obtain a test scene set, wherein the scene generation model is trained based on historical traffic scene data; Performing risk assessment on each test scene in the test scene set, and determining the risk level of each test scene; And sequencing the priority of each test scene according to the risk level to generate a test case sequence. A second aspect of an embodiment of the present application provides an intelligent driving real vehicle test case generating device, including: the data processing module is used for acquiring multi-source data and identifying target scene fragments from the multi-source data; The feature extraction module is used for extracting features of the target scene segment to obtain feature information; the scene generation module is used for inputting the characteristic information into a pre-trained scene generation model to obtain a test scene set, wherein the scene generation model is trained based on historical traffic scene data; the risk evaluation module is used for performing risk evaluation on each test scene in the test scene set and determining the risk level of each test scene; and the priority ranking module is used for ranking the priorities of the test scenes according to the risk level to generate a test case sequence. A third aspect of the embodiment of the present application provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, where