Vehicle testing method based on scene feature reconstruction and related device

Abstract

The invention discloses a vehicle testing method based on scene feature reconstruction and a related device. The method comprises the following steps: constructing a scene parameter space; sampling sample points in the constructed scene parameter space, and obtaining a plurality of test scenes; performing a simulation test on the plurality of test scenes, and screening out a dangerous scene set exceeding a risk index threshold in simulation results; reconstructing scene features in the dangerous scene set by using a principal component analysis method to obtain a plurality of feature vectors; and taking the plurality of feature vectors as search directions, sequentially carrying out dangerous scene search, and testing the vehicle by utilizing the searched dangerous scene. By constructing the scene parameter space and sampling the sample points, a plurality of representative test scenes can be quickly generated, blind test is avoided, and the test is more targeted and efficient; in addition, the scene features in the dangerous scene set are reconstructed by using the principal component analysis method, and the feature vector is used as the search direction, so that the dangerous scene search efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving, and particularly relates to a vehicle testing method and related device based on scenario feature reconstruction. Background Art

[0002] To ensure the safe and reliable operation of autonomous vehicles in various complex and changeable traffic environments, strict testing and verification procedures are of crucial importance. Although traditional on-site testing methods are feasible, due to their high costs, long time consumption, and limited scenario coverage, they have gradually become difficult to meet the growing testing requirements. Therefore, scenario-based testing methods, with their significant advantages of high efficiency and low cost, are gradually becoming the mainstream choice in the industry.

[0003] Scenario-based testing methods can simulate more diverse traffic scenarios, improving the comprehensiveness and accuracy of testing. However, how to efficiently and accurately search for potential dangerous scenarios in the high-dimensional scenario space remains a key problem that needs to be solved urgently. Traditional methods mainly rely on empirical judgment and only extract a small number of typical scenarios for testing. This approach is highly subjective and prone to missing those crucial dangerous scenarios.

[0004] To overcome this limitation, intelligent optimization algorithms have been introduced to accelerate the search process. These algorithms can not only efficiently search in the parameter space but also usually combine risk assessment indicators to quantitatively evaluate the risk of scenarios. For example, by calculating the magnitude of the time to collision, the search range and direction in the parameter space of the optimization algorithm are determined. However, the calculation of risk indicators often needs to be carried out in real time. Before the start of the simulation test, we cannot predict the specific risk magnitudes of each scenario in the parameter space, so it is difficult to determine the optimal search direction in advance.

[0005] In addition, the search method based on the surrogate model provides a new idea for solving this problem. This method can give the search direction based on theory, but the prerequisite is that the dangerous scenarios need to satisfy a certain distribution law to construct an accurate fitting function. Therefore, the application scope of this method is relatively limited and still requires further research and improvement. Summary of the Invention

[0006] The purpose of the present invention is to provide a vehicle testing method and related device based on scenario feature reconstruction to solve the technical defect that the prior art cannot effectively search for and identify potential dangerous scenarios.

[0007] To achieve the above purpose, the present invention adopts the following technical solutions: In the first aspect, a vehicle testing method based on scenario feature reconstruction is provided, including: Construct a scenario parameter space; In the constructed scenario parameter space, sample sample points to obtain multiple test scenarios; Conduct simulation tests on multiple test scenarios, and screen out a set of dangerous scenarios in the simulation results that exceed the risk index threshold; Use the principal component analysis method to reconstruct the scenario features in the set of dangerous scenarios to obtain multiple feature vectors; Using the multiple feature vectors as search directions, sequentially search for dangerous scenarios, and use the searched dangerous scenarios to test the vehicle.

[0008] Further, the construction of the scenario parameter space specifically includes: Based on the driving functions of the vehicle to be tested, design functional scenarios corresponding to the driving functions; Determine the dimension and value range of the scenario variables in the functional scenario, and combine the dimension and value range of the scenario variables to obtain a scenario parameter space.

[0009] Further, the driving functions of the vehicle include the vehicle emergency braking function and the vehicle cut-in and cut-out functions.

[0010] Further, in the constructed scenario parameter space, sample sample points to obtain multiple test scenarios, specifically including: In the constructed scenario parameter space, use a random algorithm to perform initial sample point sampling to obtain multiple sample points; Among them, each sample point corresponds to a test scenario.

[0011] Further, conducting simulation tests on multiple test scenarios and screening out a set of dangerous scenarios in the simulation results that exceed the risk index threshold specifically includes: Conduct simulation tests on multiple test scenarios to check whether the vehicle driving functions corresponding to the test scenarios are normal; Based on the risk index, screen out a set of dangerous scenarios in the simulation results that exceed the risk index threshold.

[0012] Further, the risk index includes a value that can reflect the real-time risk of the test scenario, a value that can consider the lateral risk and longitudinal risk of the vehicle, and the value of the risk index is positively correlated with the driving risk; Taking the maximum real-time risk as the scenario risk value, when the scenario risk value exceeds the threshold, add the test scenario to the set of dangerous scenarios.

[0013] Further, using the principal component analysis method to reconstruct the scenario features in the set of dangerous scenarios to obtain multiple feature vectors specifically includes: Using the principal component analysis method, the scene features in the dangerous scene set are reconstructed to obtain multiple principal components, and the principal components correspond to the eigenvectors.

[0014] Further, taking the multiple eigenvectors as search directions, dangerous scene searches are performed in sequence, and the vehicles are tested using the searched dangerous scenes, specifically including: Taking the multiple eigenvectors as search directions, dangerous scene searches are performed in sequence; among them, when no dangerous scene is found in a certain search direction, the dangerous scenes found in this search direction are added to the dangerous scene set, and the search continues through other search directions until all search directions are completed, and the vehicles are tested using the searched dangerous scenes.

[0015] In a second aspect, a vehicle test system based on scene feature reconstruction is provided, including: A construction module for constructing a scene parameter space; A sampling module for sampling sample points in the constructed scene parameter space; A simulation module for performing simulation tests on multiple test scenes; A reconstruction module for reconstructing the scene features in the dangerous scene set; A search module for searching for dangerous scenes; A test module for testing vehicles.

[0016] In a third aspect, a mobile terminal is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the vehicle test method based on scene feature reconstruction described above are implemented.

[0017] In a fourth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the vehicle test method based on scene feature reconstruction described above are implemented.

[0018] In a fifth aspect, a computer program product is provided, including computer instructions, and the computer instructions direct a computing device to perform operations corresponding to the vehicle test method based on scene feature reconstruction described above.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. By constructing a scenario parameter space and sampling sample points, multiple representative test scenarios can be quickly generated, avoiding blind testing and making the testing more targeted and efficient. Additionally, the principal component analysis method is used to reconstruct the scenario features in the dangerous scenario set, and the obtained feature vectors are used as the search directions, further improving the efficiency of dangerous scenario search. This targeted search method can locate potential dangerous scenarios faster, thus accelerating the testing process and solving the technical defect that the prior art cannot effectively search for and identify potential dangerous scenarios.

[0020] 2. By designing functional scenarios based on the driving functions of the vehicle to be tested, it is ensured that the test scenarios are highly relevant to the actual usage scenarios of the vehicle. This targeted design helps to more accurately evaluate the performance of the vehicle under specific driving functions. Determine the dimensionality and value range of the scenario variables in the functional scenario and combine them into a scenario parameter space, enabling the testing to cover a wider range of driving conditions and environmental changes. This comprehensive testing helps to discover potential problems of the vehicle in different scenarios and improve the coverage of the testing.

[0021] 3. Incorporating the vehicle's emergency braking function and cut-in and cut-out functions into the construction of the scenario parameter space has significant beneficial effects in aspects such as improving the accuracy of vehicle safety performance evaluation, enhancing the pertinence and practicality of testing, promoting the development of autonomous driving technology, and improving road traffic safety levels.

[0022] 4. The random algorithm can ensure the uniform and random selection of sample points within the scenario parameter space, which helps to generate a variety of different test scenarios covering a wide range of driving conditions and environmental changes. Through diverse test scenarios, the performance of the vehicle in different situations can be more comprehensively evaluated, and potential defects or deficiencies can be discovered.

[0023] 5. By conducting simulation tests on multiple test scenarios, it can systematically verify whether the driving functions of the vehicle are normal under various driving conditions. Based on the preset risk indicators, screening the simulation results can accurately identify those dangerous scenarios that may cause abnormal vehicle driving or accidents. By optimizing these scenarios, the safety and reliability of the vehicle can be significantly improved.

[0024] 6. The risk indicators cover real-time risk, lateral risk, and longitudinal risk, and can comprehensively evaluate the potential risks in the test scenarios. The real-time risk reflects the current danger level of the scenario, while the lateral risk and longitudinal risk respectively consider the possibility of the vehicle deviating from the lane and rear-ending accidents. This comprehensive risk assessment helps to more accurately identify dangerous scenarios and ensure the diversity and comprehensiveness of the test scenarios.

[0025] 7. Convert the original high-dimensional scene feature data into low-dimensional principal component data. These principal components retain most of the important information in the original data while removing redundant information, thus simplifying the data structure. The data after dimensionality reduction is easier to process and analyze, reducing the computational complexity and improving the efficiency of data processing.

[0026] 8. By using multiple feature vectors as search directions, it is possible to ensure searching in a wide range of scene feature spaces, thus covering potential dangerous scenes more comprehensively. This helps to discover the potential risks of the vehicle under different feature combinations and provides comprehensive data support for vehicle performance evaluation and improvement. Brief Description of the Drawings

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0028] Figure 1 It is a flowchart of vehicle testing based on scene feature reconstruction provided by the present invention; Figure 2 It is a schematic diagram of a vehicle testing system based on scene feature reconstruction provided by the present invention. Detailed Embodiments

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and shown in the accompanying drawings here can be arranged and designed in various different configurations.

[0030] Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0031] It should be noted that similar reference numerals and letters indicate similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0032] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper", "lower", "horizontal", "inner", etc. are used to indicate the orientation or positional relationship, it is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed during use. This is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation on the present invention. In addition, terms such as "first", "second", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0033] In addition, if the term "horizontal" appears, it does not mean that the component is required to be absolutely horizontal, but it can be slightly inclined. For example, "horizontal" only means that its direction is more horizontal relative to "vertical", and it does not mean that the structure must be completely horizontal, but it can be slightly inclined.

[0034] In the description of the embodiments of the present invention, it should also be noted that unless otherwise clearly specified and limited, if terms such as "set", "installed", "connected", "linked" are used, they should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0035] To ensure the safe and reliable operation of autonomous vehicles in various complex and changeable traffic environments, strict testing and verification processes are crucial. Although traditional on-site testing methods are feasible, due to their high costs, long time consumption, and limited scenario coverage, they are gradually unable to meet the growing testing needs. Therefore, scenario-based testing methods, with their significant advantages of high efficiency and low cost, are gradually becoming the mainstream choice in the industry.

[0036] Scenario-based testing methods can simulate more diverse traffic scenarios, improving the comprehensiveness and accuracy of testing. However, how to efficiently and accurately search for potential dangerous scenarios in the high-dimensional scenario space remains a key problem that needs to be solved urgently. Traditional methods mainly rely on empirical judgment and only extract a small number of typical scenarios for testing. This method is highly subjective and prone to missing those crucial dangerous scenarios.

[0037] To overcome this limitation, intelligent optimization algorithms are introduced to accelerate the search process. Such algorithms can not only efficiently search in the parameter space but also usually combine risk assessment metrics to quantitatively evaluate the risks of scenarios. For example, by calculating the magnitude of the time to collision, the search range and direction of the optimization algorithm in the parameter space are determined. However, the calculation of risk metrics often needs to be carried out in real time. Before the start of the simulation test, we cannot predict the specific risk magnitudes of each scenario in the parameter space. Therefore, it is difficult to determine the optimal search direction in advance.

[0038] In addition, the search method based on the surrogate model provides a new idea for solving this problem. This method can give the search direction based on theory, but the prerequisite is that the dangerous scenarios need to satisfy certain distribution rules to construct an accurate fitting function. Therefore, the application scope of this method is relatively limited and still needs further research and improvement.

[0039] To solve the above technical deficiencies, the inventors provide a vehicle testing method and related device based on scenario feature reconstruction.

[0040] The following further describes the present invention in detail with reference to the accompanying drawings: In the first aspect, an embodiment of the present invention provides a vehicle testing method based on scenario feature reconstruction, as Figure 1 shown, including: S101. Construct a scenario parameter space. Exemplarily, first, based on the driving functions of the vehicle to be tested, design function scenarios corresponding to the driving functions. Among them, the driving functions of the vehicle include the vehicle emergency braking function and the vehicle cut-in and cut-out functions. Then, determine the dimension and value range of the scenario variables in the function scenario, and combine the dimension of the scenario variables and the value range to obtain the scenario parameter space. In the above steps, by designing function scenarios based on the driving functions of the vehicle to be tested, it is ensured that the test scenarios are highly relevant to the actual use scenarios of the vehicle, and the performance of the vehicle under specific driving functions can be evaluated more accurately. Moreover, by determining the dimension and value range of the scenario variables in the function scenario and combining them into the scenario parameter space, the test can cover a wider range of driving conditions and environmental changes, which helps to discover potential problems of the vehicle in different scenarios and improve the coverage of the test. In addition, the construction of the scenario parameter space provides a clear variable range and dimension for the test, making the test process more controllable. Testers can adjust the values of the scenario variables as needed to simulate different driving scenarios. At the same time, this controllability also enhances the repeatability of the test. The same scenario parameter settings can ensure consistent test results in different times or different test environments, which helps to verify and improve the performance of the vehicle. Further, in terms of function scenario design, emergency braking is an important part of the vehicle safety performance. By considering emergency braking situations, the emergency braking effects of the vehicle at different speeds and on different road conditions can be simulated, and the braking distance, braking stability, and working performance of the anti-lock braking system (ABS) of the vehicle can be evaluated more accurately, thereby improving the safety performance of the vehicle in emergency situations. The cut-in and cut-out functions involve the performance of the vehicle in driving behaviors such as lane changing and overtaking. Incorporating these functions into the scenario parameter space can simulate the cut-in and cut-out behaviors of the vehicle at different speeds and different vehicle distances, and evaluate the handling, stability, and obstacle avoidance capabilities of the vehicle, which is of great significance for reducing the risk of traffic accidents and improving road traffic efficiency.

[0041] S102. In the constructed scenario parameter space, perform sample point sampling and obtain multiple test scenarios. Exemplarily, in the constructed scenario parameter space, use a random algorithm to perform initial sample point sampling and obtain multiple sample points. Among them, each sample point corresponds to a test scenario. The stop condition for the initial sampling is:

[0042] In the formula is a constant, and the set range is 5 to 10, and the value increases as increases.

[0043] Random algorithms can ensure the uniform and random selection of sample points within the scenario parameter space and generate a variety of different test scenarios, covering a wide range of driving conditions and environmental changes. Through diverse test scenarios, the performance of the vehicle in different situations can be evaluated more comprehensively, and potential defects or deficiencies can be discovered. At the same time, random sampling helps to avoid the concentration or bias of test scenarios, ensuring that the test can cover all regions within the scenario parameter space, revealing the behavioral differences of the vehicle under different driving scenarios, and providing comprehensive data support for the performance improvement of the vehicle. Moreover, through efficient sample point sampling, the test cycle can be shortened, the test cost can be reduced, and the test efficiency can be improved; the test scenarios generated by random sampling can be used as the initial data set for subsequent optimization. Based on these data, optimization algorithms or machine learning techniques can be further used to search for better test scenarios or identify key performance indicators.

[0044] S103. Conduct simulation tests on multiple said test scenarios, and screen out a set of dangerous scenarios whose simulation results exceed the risk index threshold; Exemplarily, conduct simulation tests on multiple test scenarios to check whether the vehicle driving functions corresponding to the test scenarios are normal, and then based on the risk index, screen out a set of dangerous scenarios whose simulation results exceed the risk index threshold. The set of dangerous scenarios is represented by C. In this process, by conducting simulation tests on multiple test scenarios, the driving functions of the vehicle under various driving conditions can be systematically checked, including verifying the basic functions of the vehicle such as acceleration, braking, steering, lane changing in and out, as well as more complex autonomous driving or assisted driving functions, which helps to ensure that the vehicle can drive safely and reliably in all situations and improve the overall performance of the vehicle. Based on the preset risk index, screening the simulation results can accurately identify those dangerous scenarios that may cause abnormal vehicle driving or accidents. Through simulation tests, the driving performance of the vehicle can be evaluated quickly and efficiently in a virtual environment without actually driving the vehicle for testing, thus saving a large amount of time and cost; at the same time, the screening process based on the risk index is also automated, further improving the test efficiency. The screened set of dangerous scenarios provides valuable data support for the performance improvement of the vehicle. By analyzing the vehicle behavior, environmental conditions, and risk factors in these scenarios, the deficiencies or defects in the vehicle performance can be found. Based on these data, vehicle manufacturers can make targeted improvements and optimizations to improve the safety performance and driving experience of the vehicle.

[0045] Among the risk indicators, the risk indicators include values that can reflect the real-time risk of the test scenario, values that can consider the lateral risk and longitudinal risk of the vehicle, and the values of the risk indicators are positively correlated with the driving risk; taking the maximum real-time risk as the scenario risk value, when the scenario risk value exceeds the threshold, add the test scenario to the dangerous scenario set. In this step, the risk indicators cover real-time risk, lateral risk, and longitudinal risk, and can comprehensively evaluate the potential risks in the test scenario; the real-time risk reflects the current danger level of the scenario, while the lateral risk and longitudinal risk respectively consider the possibility of the vehicle deviating from the lane and rear-end accidents. This comprehensive risk assessment helps to more accurately identify dangerous scenarios and ensure the diversity and comprehensiveness of the test scenarios. Secondly, the values of the risk indicators are positively correlated with the driving risk, which means that the larger the risk indicator, the higher the danger level of the test scenario. This positive correlation makes risk identification more intuitive and accurate. By setting a reasonable risk indicator threshold, dangerous scenarios exceeding the threshold can be accurately screened out, avoiding misjudging low-risk scenarios as dangerous scenarios, thereby improving the test efficiency. Taking the maximum real-time risk as the scenario risk value can ensure the selection of the most representative dangerous scenarios among multiple test scenarios, and these scenarios have higher value for optimizing and improving vehicle performance.

[0046] By screening out the dangerous scenario set through simulation testing, the performance of the vehicle can be efficiently evaluated in a virtual environment without actually driving the vehicle for testing, which not only saves time and cost but also improves the safety of the testing; at the same time, the results of the simulation testing can provide valuable data support for vehicle manufacturers to guide the design and improvement of vehicles.

[0047] S104. Use the principal component analysis method to reconstruct the scenario features in the dangerous scenario set to obtain multiple feature vectors; exemplarily, use the principal component analysis method to reconstruct the scenario features in the dangerous scenario set to obtain multiple principal components, and the principal components correspond to the feature vectors; among them, when performing principal component analysis, the sample is C, the index is the value of the scenario variable in C, and k principal components are selected for analysis. The k principal components need to meet the following conditions:

[0048] In the formula is the contribution value of the i-th principal component.

[0049] S105. Using the multiple feature vectors as search directions, perform dangerous scenario searches in sequence, and use the searched dangerous scenarios to test the vehicle; exemplarily, the feature vector corresponding to the k-th principal component is ; use the feature vector as the search direction; when the feature vector When there is no dangerous scenario in the direction, the search is paused, and the dangerous scenarios obtained from the search are added to C; continue to use the eigenvector of the second principal component as the search direction for the search, and add the dangerous scenarios obtained from the search to C, and so on until the search is completed in the direction of the eigenvector Finally, use the original optimization algorithm strategy to search until the convergence condition is reached.

[0050] It should be noted that the direction of the eigenvector is used to constrain the search direction of the optimization algorithm, and at the same time, the position corresponding to the scenarios in C in the parameter space is used as the reference point for the search; the method for determining the reference point is to use any scenario point in C, or the midpoint of the line connecting any two scenario points as the reference point, and search along the direction of the eigenvector; The condition for stopping the search is that no new dangerous scenarios appear after searches in the direction of this eigenvector; after stopping the search in the direction of the k-th eigenvector, the search strategy changes to the original search strategy of the optimization algorithm until the convergence condition of the original search strategy is reached, and finally the dangerous set C under the scenario is obtained.

[0051] When applying the above method to a vehicle, for example, design the scenario of the leading vehicle cutting in as the functional scenario, which includes four-dimensional scenario variables: the speed of the host vehicle , the speed of the leading vehicle , the distance between the host vehicle and the leading vehicle , the lane-changing time between the host vehicle and the leading vehicle , and set the value range to [30, 40] km / h, [20, 30] km / h, [30, 60] m, [3, 6] s, and 10 steps are evenly discretized in each dimension.

[0052] Initialize random sampling, take , and the stop condition for the initialized sampling is:

[0053] The risk index needs to have the following characteristics: the risk index can reflect the real-time risk of the scenario, can comprehensively consider the lateral and longitudinal risks of the vehicle, and the value of the risk index is positively correlated with the driving risk, and the maximum real-time risk is used as the scenario risk value; In the embodiment, the driving risk index used is the Discretized Normalized Drivable Area (DNDA)[1].

[0054] The larger the DNDA is, it indicates that the passable area of the measured vehicle in the future period is smaller and the driving risk is higher.

[0055] Set the threshold of the scenario risk value to 0.8. When DNDA > 0.8, add the scenario to C; During principal component analysis, the samples are C, and the indicators are the values of the scenario variables in C. Analyze and select k principal components. The k principal components need to meet the following conditions:

[0056] In the formula is the contribution value of the i-th principal component; in the embodiment, the number of selected principal components k = 2, .

[0057] Use the eigenvector to constrain the search direction of the optimization algorithm, and at the same time use the scenario points in C as the reference points for search; The selected optimization algorithm in the embodiment is the particle swarm algorithm, and the eigenvector is [0.45, 0.32, -0.38, 0.53].

[0058] The method for determining the reference point is to use any one scenario point in C, or the midpoint of the line connecting any two scenario points as the reference point. The two methods for confirming the reference point are randomly selected with equal probability, and then search along the direction of the eigenvector; The condition for stopping the search is that no new dangerous points appear after 24 searches in the direction of this eigenvector, and then switch to using the eigenvector to constrain the search direction; in the embodiment [0.45, -0.32, -0.52, 0.27]. After stopping the search in the direction of the second eigenvector, the search strategy changes to the original search strategy of the particle swarm algorithm until the convergence condition of the original search strategy is reached, and finally the dangerous set C in the logical scenario is obtained.

[0059] On the second aspect, a vehicle test system based on scenario feature reconstruction is provided, as Figure 2 shown, including: A construction module for constructing a scenario parameter space; A sampling module for sampling sample points in the constructed scenario parameter space; A simulation module for performing simulation tests on multiple test scenarios; A reconstruction module for reconstructing the scenario features in the dangerous scenario set; A search module for searching for dangerous scenarios; A test module for testing the vehicle.

[0060] By constructing a module to build a scenario parameter space, it can cover various driving scenarios that a vehicle may encounter, ensuring the comprehensiveness of testing. The sampling module samples sample points in the scenario parameter space, which can efficiently select representative test scenarios, reduce unnecessary test times, and improve test efficiency.

[0061] The simulation module conducts simulation tests on multiple test scenarios, which can simulate various situations that a vehicle may encounter during actual driving, making the test results closer to the real situation. Through simulation tests, the performance, safety, and reliability of the vehicle can be comprehensively evaluated in a virtual environment, reducing risks and costs in actual testing.

[0062] The reconstruction module reconstructs the scenario features in the dangerous scenario set, which can highlight the key features of these scenarios, facilitating subsequent analysis and research. By reconstructing dangerous scenarios, the vehicle's performance in extreme or dangerous situations can be tested more targeted, thereby discovering and solving potential safety problems in advance.

[0063] The search module searches for dangerous scenarios, which can quickly locate the test scenarios that need to be focused on, improving the pertinence and efficiency of testing.

[0064] The test module tests the vehicle. Combining the previous sampling, simulation, and reconstruction results, it can form a complete test process, ensuring the comprehensiveness and accuracy of testing.

[0065] This system provides an efficient test platform for the research and development of vehicle technology, which can help R & D personnel discover problems, optimize designs, and improve performance faster. Through continuous testing and improvement, it can promote the continuous progress and innovation of vehicle technology, laying a foundation for the development of future intelligent transportation systems.

[0066] In a third aspect, a mobile terminal is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the vehicle testing method based on scenario feature reconstruction as described above are implemented.

[0067] In a fourth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the vehicle testing method based on scenario feature reconstruction as described above are implemented.

[0068] In a fifth aspect, a computer program product is provided, including computer instructions, and the computer instructions direct a computing device to perform operations corresponding to the vehicle testing method based on scenario feature reconstruction as described above.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than limit the scope of its protection. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that after reading the present invention, various changes, modifications or equivalent replacements can still be made to the specific implementation manners of the invention, but these changes, modifications or equivalent replacements are all within the scope of protection of the pending claims of the invention.

Claims

1. A vehicle testing method based on scene feature reconstruction, characterized in that: include: Construct scene parameter space; In the constructed scene parameter space, sample points are sampled to obtain a plurality of test scenes; Performing simulation tests on a plurality of the test scenarios, and screening out a set of dangerous scenarios that exceed a risk indicator threshold in the simulation results; Using principal component analysis, reconstructing scene features in the dangerous scene set to obtain multiple feature vectors; The plurality of feature vectors are used as search directions to sequentially search for dangerous scenes, and the vehicle is tested using the searched dangerous scenes.

2. The vehicle testing method based on scene feature reconstruction according to claim 1 is characterized in that: The constructing of the scene parameter space specifically includes: Based on the driving function of the vehicle to be tested, designing a functional scenario corresponding to the driving function; The dimension and value range of the scene variables in the functional scene are determined, and the dimension and value range of the scene variables are combined to obtain a scene parameter space.

3. The vehicle testing method based on scene feature reconstruction according to claim 2 is characterized in that: The driving functions of the vehicle include a vehicle emergency braking function, and a vehicle cut-in and cut-out function.

4. The vehicle acceleration test method based on scene feature reconstruction according to claim 1, characterized in that: In the constructed scene parameter space, sample points are sampled and multiple test scenes are obtained, including: In the constructed scene parameter space, a random algorithm is used to initialize sample points and obtain multiple sample points; Among them, each sample point corresponds to a test scene.

5. The vehicle testing method based on scene feature reconstruction according to claim 1 is characterized in that: Perform simulation tests on multiple test scenarios, and screen out dangerous scenario sets that exceed risk index thresholds in the simulation results, specifically including: Conduct simulation tests on multiple test scenarios to verify whether the vehicle driving functions corresponding to the test scenarios are normal; Based on the risk index, the dangerous scenario set that exceeds the risk index threshold in the simulation results is screened out.

6. The vehicle testing method based on scene feature reconstruction according to claim 5 is characterized in that: The risk index includes a value that can reflect the real-time risk of the test scenario, a value that can take into account the lateral risk and longitudinal risk of the vehicle, and a value of the risk index that is positively correlated with the driving risk; The maximum real-time risk is used as the scenario risk value. When the scenario risk value exceeds the threshold, the test scenario is added to the dangerous scenario set.

7. The vehicle testing method based on scene feature reconstruction according to claim 1 is characterized in that: The principal component analysis method is used to reconstruct the scene features in the dangerous scene set to obtain multiple feature vectors, including: The principal component analysis method is used to reconstruct the scene features in the dangerous scene set to obtain multiple principal components, which correspond to the feature vectors.

8. The vehicle testing method based on scene feature reconstruction according to claim 1 is characterized in that: Using the plurality of feature vectors as search directions, searching for dangerous scenes in sequence, and testing the vehicle using the searched dangerous scenes, specifically includes: Using multiple feature vectors as search directions, dangerous scene searches are performed in sequence. When no dangerous scene is found in a certain search direction, the dangerous scene found in that search direction is added to the dangerous scene set, and the search is continued in other search directions until all search directions are searched, and the vehicle is tested using the found dangerous scenes.

9. A vehicle testing system based on scene feature reconstruction, characterized in that: include: Construction module, used to construct the scene parameter space; A sampling module, used to sample sample points in the constructed scene parameter space; A simulation module, used to simulate multiple test scenarios; A reconstruction module, used to reconstruct scene features in the dangerous scene set; A search module, used to search for dangerous scenes; A test module is used to test the vehicle.

10. A mobile terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the vehicle testing method based on scene feature reconstruction as described in any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the vehicle testing method based on scene feature reconstruction as described in any one of claims 1 to 8 are implemented.

12. A computer program product comprising computer instructions, characterized in that: The computer instructions instruct the computing device to execute operations corresponding to the vehicle testing method based on scene feature reconstruction as described in any one of claims 1-8.

Images