Intelligent driving test system and intelligent driving test method based on data driving

Abstract

The invention provides an intelligent driving test system and an intelligent driving test method based on data driving. The intelligent driving test system comprises a data acquisition unit and a test unit. The data acquisition unit determines a preset acquisition scene based on roads, traffic infrastructures, temporary operations, traffic participants, environments, and digital information. And acquiring a preset acquisition scene by using a sensor of the vehicle to obtain sensor data. A laser radar and other devices are added to collect a preset collection scene to obtain point cloud data. And processing the point cloud data to obtain true value data. And the test unit tests the ISP algorithm based on the sensor data to obtain image processing data. And testing a sensing algorithm based on the image processing data and the target truth value data to obtain up-to-standard sensing data. And testing the regulation and control algorithm by using the standard perception data. Based on the system, various preset acquisition scenes can be covered for testing, and the testing accuracy is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent driving, in particular to an intelligent driving test system and method based on data driving. BACKGROUND

[0002] With the development of science and technology, intelligent driving technology has become an important direction for future transportation development. Intelligent driving technology enables vehicles to autonomously perceive the environment, plan paths, and perform driving tasks through sensors, algorithms, and control systems. To cope with complex and variable traffic scenarios, a large amount of scenario data is needed to support algorithm development and verification.

[0003] The current intelligent driving system test method mainly starts from system requirement definition, combines relevant test regulations, builds test cases (algorithms), and executes test cases through real vehicle testing or simulation testing to achieve the purpose of algorithm testing. However, the existing test data mainly focuses on regular scenarios, and it is difficult to cover complex and variable traffic scenarios (such as extreme weather, sudden obstacles, etc.). SUMMARY

[0004] Therefore, the present application is committed to providing an intelligent driving test system and method based on data driving, so as to cover a variety of preset collection scenarios, meet a variety of test requirements, and improve the accuracy of intelligent driving testing.

[0005] In a first aspect, the present application provides an intelligent driving test system based on data driving, which comprises a data collection unit and a test unit.

[0006] The data collection unit is configured to determine a preset collection scenario based on one or more scene elements in road, traffic infrastructure, temporary operation, traffic participants, environment, and digital information; collect the preset collection scenario based on the sensors of the vehicle to obtain sensor data; collect the preset collection scenario based on radar to obtain point cloud data, and process the point cloud data to obtain true value data.

[0007] The test unit is configured to test an image signal processing (ISP) algorithm based on the sensor data to obtain image processing data; test a perception algorithm based on the image processing data and the true value data to obtain qualified perception data; and test a regulation and control algorithm based on the qualified perception data.

[0008] In one possible implementation, the sensor data includes Raw data collected by the image sensor of the vehicle.

[0009] In one possible implementation, the testing unit is configured to input the image processing data into the perception algorithm and output a perception result; compare the perception result with the ground truth data; when the perception result matches the ground truth data, determine that the perception result is the qualified perception data; otherwise, determine that the perception result is the unqualified perception data.

[0010] The sensing algorithm is iteratively optimized based on the substandard sensing data.

[0011] In one possible implementation, the test unit is further configured to compare the compliance perception data with the operational design domain to determine the perception data within the domain and the perception data outside the domain.

[0012] The control algorithm was tested based on the sensing data within the domain.

[0013] Based on the external sensing data, a requirements assessment is conducted to determine whether the operational design domain should be expanded.

[0014] In one possible implementation, the data acquisition unit is used to determine the actual planned mileage distribution corresponding to the preset acquisition scenario based on the city level and city mileage distribution corresponding to the preset acquisition scenario.

[0015] In one possible implementation, the data acquisition unit is used to acquire sensor data obtained by the sensors of the mass-produced vehicle in the preset acquisition scenario based on the bypass mode, wherein the mass-produced vehicle refers to the vehicle that the intelligent driving algorithm theoretically controls.

[0016] In one possible implementation, the system further includes a data processing unit for performing compliance processing and / or data slicing on the sensor data and the ground truth data to obtain target sensor data and target ground truth data.

[0017] In one possible implementation, the data processing unit is further configured to perform data generalization and / or data reconstruction based on the target sensor data and / or the target ground truth data to obtain a simulation scenario;

[0018] The testing unit is used to test the ISP algorithm based on the target sensor data to obtain the image processing data; to perform simulation tests on the ISP algorithm, the perception algorithm, and the control algorithm in the simulation scenario; or, to feed back the target sensor data and the target ground truth data into the system of the ISP algorithm, the perception algorithm, and the control algorithm for feedback testing.

[0019] Secondly, this application provides an intelligent driving testing method, the method comprising:

[0020] The preset data collection scenario is determined by combining one or more scenario elements from roads, traffic infrastructure, temporary operations, traffic participants, environment, and digital information.

[0021] The vehicle's sensors collect data from the preset data acquisition scenario to obtain sensor data.

[0022] Based on radar, point cloud data is collected from the preset acquisition scenario, and the point cloud data is processed to obtain ground truth data.

[0023] The ISP algorithm was tested based on the sensor data to obtain image processing data;

[0024] The perception algorithm is tested based on the image processing data and the ground truth data to obtain compliant perception data;

[0025] The regulatory control algorithm was tested based on the compliance perception data.

[0026] Thirdly, this application provides an intelligent driving testing device, the device comprising:

[0027] The scenario determination module is used to determine the preset collection scenario based on a combination of one or more scenario elements, including roads, traffic infrastructure, temporary operations, traffic participants, environment, and digital information.

[0028] The first acquisition module is used to acquire sensor data based on the vehicle's sensors in the preset acquisition scenario.

[0029] The second acquisition module is used to acquire point cloud data based on the preset acquisition scenario using radar.

[0030] The first processing module is used to process the point cloud data to obtain true value data;

[0031] The testing module is used to test the ISP algorithm based on the sensor data to obtain image processing data; to test the perception algorithm based on the image processing data and the ground truth data to obtain compliance perception data; and to test the regulatory control algorithm based on the compliance perception data.

[0032] Fourthly, this application provides an electronic device, the device comprising: a memory and a processor;

[0033] The memory is used to store the relevant program code;

[0034] The processor is used to call the program code to execute the intelligent driving test method described in any of the implementations of the second aspect above.

[0035] Fifthly, this application provides a computer-readable storage medium for storing a computer program for executing the intelligent driving test method described in any implementation of the second aspect above.

[0036] Sixthly, this application provides a computer program product, which includes a computer program / instruction, and when the computer program / instruction is executed by a processor, it implements the intelligent driving test method described in any of the implementations of the second aspect above.

[0037] In the above implementation of this application, the intelligent driving test system includes a data acquisition unit and a test unit. The data acquisition unit can determine multiple preset acquisition scenarios based on a six-layer scenario structure, where the six-layer scenario structure includes roads, traffic infrastructure, temporary operations, traffic participants, environment, and digital information. It uses vehicle sensors to collect data from the preset acquisition scenarios, obtaining sensor data. It can also add equipment such as LiDAR to collect data from the preset acquisition scenarios, obtaining accurate point cloud data. The point cloud data is then processed to obtain ground truth data. The test unit can test the ISP algorithm based on the sensor data, obtaining image processing data. Then, based on the image processing data and ground truth data, it tests the perception algorithm to obtain compliant perception data. The compliant perception data is then used to test the regulatory control algorithm. The system provided by this application can collect data from multiple real-world test scenarios, covering various test scenarios and improving test accuracy. Based on the above system, full-process testing of multiple algorithms can be achieved, more closely resembling the testing process of real-world scenarios and improving testing efficiency. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments provided in this application. For those skilled in the art, other drawings can be obtained based on these drawings.

[0039] Figure 1 This is a schematic diagram of a data-driven intelligent driving test system provided in an embodiment of this application.

[0040] Figure 2 This is a schematic diagram of a six-layer scene structure provided in an embodiment of this application.

[0041] Figure 3 This is a schematic diagram illustrating a method for testing intelligent driving algorithms, as provided in an embodiment of this application.

[0042] Figure 4This is a schematic diagram of another data-driven intelligent driving test system provided in an embodiment of this application.

[0043] Figure 5 This is a flowchart of an intelligent driving test method provided in an embodiment of this application.

[0044] Figure 6 This is a schematic diagram of an intelligent driving test device provided in an embodiment of this application.

[0045] Figure 7 This is a schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0046] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are merely exemplary implementations of this application and not all implementation methods. Those skilled in the art can obtain other embodiments in conjunction with the embodiments of this application without creative effort, and these embodiments are also within the protection scope of this application.

[0047] Intelligent driving technology, through sensors, algorithms, and control systems, enables vehicles to autonomously perceive their environment, plan routes, and perform driving tasks. To cope with complex and ever-changing traffic scenarios, massive amounts of scenario data are needed to support the research and validation of algorithms.

[0048] Current testing methods for intelligent driving systems primarily start with defining system requirements, combining relevant testing regulations, building test cases (algorithms), and executing these test cases through real-vehicle testing or simulation testing to achieve the purpose of algorithm testing. However, existing test datasets mostly focus on conventional scenarios and are difficult to cover complex and ever-changing traffic scenarios (such as extreme weather, sudden obstacles, etc.).

[0049] Based on this, this application provides a data-driven intelligent driving testing system that can cover multiple preset data acquisition scenarios, meet various testing needs, and improve the accuracy of intelligent driving testing. Specifically, the system includes a data acquisition unit and a testing unit. The data acquisition unit can determine multiple preset data acquisition scenarios based on a six-layer scenario structure encompassing roads, traffic infrastructure, temporary operations, traffic participants, the environment, and digital information. It uses vehicle sensors to collect data from these preset scenarios, obtaining sensor data. Radar and other equipment can also be added to collect data from the preset scenarios, obtaining accurate point cloud data. Ground truth data is obtained by processing the point cloud data. The testing unit can test the ISP algorithm based on the sensor data, obtaining image processing data. Then, based on the image processing data and ground truth data, it tests the perception algorithm to obtain compliant perception data. The compliant perception data is then used to test the regulatory control algorithm. Through the system provided in this application, data from multiple scenarios can be collected, covering multiple testing scenarios and improving testing accuracy. Based on the above system, full-process testing of multiple algorithms can be achieved, more closely resembling the testing process of real-world scenarios and improving testing efficiency.

[0050] To facilitate understanding of the technical solutions provided in the embodiments of this application, a detailed description will be given below in conjunction with the accompanying drawings.

[0051] See Figure 1 The diagram shown is a schematic of a data-driven intelligent driving test system provided in an embodiment of this application. This system 100 can be used to test intelligent driving algorithms. Optionally, the system 100 can be configured on an electronic device or in a server.

[0052] The system 100 includes a data acquisition unit 101 and a testing unit 102. The data acquisition unit 101 is primarily used to determine preset acquisition scenarios and collect the necessary data for testing based on these scenarios. Preset acquisition scenarios can be understood as the scenarios in which the intelligent driving algorithm is applied. Since it may not be possible to directly deploy the intelligent driving algorithm in the real-world testing scenarios during actual testing, preset acquisition scenarios can be determined in advance, and data from these scenarios can be collected as input data for the intelligent driving algorithm. This allows for subsequent testing of the intelligent driving algorithm through methods such as backfeed testing and simulation testing.

[0053] In practical implementation, the data acquisition unit 101 can determine a preset acquisition scenario based on at least one scenario element from a six-layer scenario structure: road, traffic infrastructure, temporary operations, traffic participants, environment, and digital information. The six-layer scenario structure can be understood as determining the preset acquisition scenario based on six levels of scenario elements. Through this six-layer scenario structure, a large number of preset acquisition scenarios can be determined, covering complex and boundary scenarios, and large-scale data from various preset scenarios can be collected to support the verification and testing of intelligent driving algorithms. The corresponding real-world scenario is then located based on the preset acquisition scenario, and the vehicle's actual sensors are used to collect data from the real preset acquisition scenario, obtaining sensor data.

[0054] To improve the accuracy of test data, the data acquisition unit 101 can collect sensor data from a preset acquisition scenario based on the vehicle's own sensors. Since the data collected by the vehicle's own sensors may contain errors, additional devices such as LiDAR can be added to collect data from the preset acquisition scenario, resulting in more accurate point cloud data. In one possible implementation, a truth system can be formed by adding LiDAR, blind-spot LiDAR, and a Real-time Kinematic (RTK) inertial navigation system to measure and store truth data. The blind-spot LiDAR can compensate for the blind spots of the vehicle's own sensors in specific scenarios by emitting laser beams and receiving reflected signals to perceive and identify targets such as vehicles, pedestrians, and obstacles in the environment. The RTK inertial navigation system can provide continuous position, velocity, and attitude information in environments with weak or no satellite signals.

[0055] When testing intelligent driving algorithms, ground truth data can be used as a reference to test the accuracy of the algorithms. Therefore, point cloud data collected by radar can be processed to obtain ground truth data. For example, the collected point cloud data may contain erroneous or missing values; filtering and registration can be performed on the point cloud data. For testing perception algorithms, it is necessary to identify targets in images based on the perception algorithm. In this case, joint calibration can be achieved based on sensor data and point cloud data. The sensor data (e.g., captured images) can be manually labeled using point cloud data, and the labeled results can be used as ground truth data.

[0056] Optionally, when the sensor data includes image sensor data, the test unit 102 can use the image sensor data to test the Image Signal Processing (ISP) algorithm, obtaining processed image processing data. The performance of the ISP algorithm can be determined based on the quality of the image processing data. When the image processing data does not meet the requirements, the ISP algorithm can be further optimized based on the image processing data. When the image processing data meets the requirements, the perception algorithm can be tested based on the image processing data and ground truth data to obtain compliant perception data. The control algorithm is then tested based on the compliant perception data.

[0057] In one possible implementation, the sensor data may include raw data acquired by the vehicle's image sensors. Raw data represents the unprocessed or unconverted data acquired by the image sensors, preserving more complete information and facilitating subsequent data processing.

[0058] For example, the ISP algorithm can be tested based on raw data to obtain image processing data. The ISP algorithm can perform noise reduction, white balance, and color correction on the input raw data, outputting image processing data in YUV or RGB format, improving image quality and facilitating subsequent image processing. Since image quality directly affects the results of perception and control algorithms, after testing the ISP algorithm with raw data, it can be further optimized based on the image processing data to further improve the quality of images processed by the ISP.

[0059] When the intelligent driving algorithm is a perception algorithm, it needs to identify targets such as pedestrians, bicycles, or roadblocks in the captured images (image processing data). In this case, the perception results of the perception algorithm need to be compared with the ground truth images (images of pedestrians, bicycles, or roadblocks) to determine the test results of the perception algorithm and obtain the correct perception data that meets the standards. When the intelligent driving algorithm is a planning and control algorithm, it plans the vehicle's driving path and controls the vehicle to drive according to the output control commands. After inputting the compliant perception data into the planning and control algorithm, the test results of the planning and control algorithm can be determined based on whether the control commands output by the algorithm meet the requirements of intelligent driving. For example, the test results of the planning and control algorithm can be determined based on whether the planned path is correct and safe, the time taken for the planned path, the distance of the planned path, and whether the vehicle is far enough away from obstacles.

[0060] Test unit 102 can perform various testing requirements, including backfeed testing and simulation testing, when testing multiple intelligent algorithms such as ISP algorithms, perception algorithms, and traffic control algorithms. Backfeed testing means that sensor data and ground truth data can be fed back into the system applying the intelligent driving algorithm to obtain test results. Simulation testing means that a simulation environment can be created based on sensor data or ground truth data, and the intelligent driving algorithm can be deployed in the simulation environment for testing.

[0061] The core concept of data-driven decision-making is to use data as the central driving force, through data collection, analysis, and modeling, and then make decisions and take actions based on this data. In this embodiment, data from a large number of test scenarios can be collected to cover a variety of test scenarios.

[0062] In one possible implementation, system 100 may further include a data processing unit, primarily used to process sensor data and ground truth data, including compliance processing and data slicing, to obtain target sensor data and target ground truth data. Subsequent testing units can then test the intelligent driving algorithm based on the target sensor data and target ground truth data. For example, compliance processing such as desensitization and declassification can be performed on the sensor data and ground truth data. Furthermore, data slicing of the sensor data and ground truth data can be performed according to different testing requirements, such as time and test type. By processing large-scale data and using the processed data to test the intelligent driving algorithm, various testing needs can be met, and the accuracy of the testing can be improved.

[0063] Optionally, when it is not necessary to acquire raw data to test the ISP algorithm, the data processing unit can use a debug ISP board or a domain controller to output the raw data acquired by the image sensor as YUV or RGB format image data for subsequent image processing. The debug ISP board can be understood as a circuit board simulating ISP functionality. The domain controller integrates ISP image processing capabilities and can output YUV or RGB format image data.

[0064] For details, please refer to Figure 2 As shown, Figure 2 This is a schematic diagram of a six-layer scene structure provided in an embodiment of this application.

[0065] according to Figure 2As can be seen, the first layer of the six-layer scene structure is the road layer, mainly including road geometry and topology, road surface quality, etc. The second layer is traffic infrastructure, mainly including lane lines, traffic lights, traffic signs, etc. The third layer is temporary operations of the first and second layers, mainly including road construction, roadblocks, road closures, etc. The fourth layer is traffic participants, mainly including vehicles, pedestrians, bicycles, etc.; the fifth layer is the environment, mainly including weather, lighting, temperature, etc. The sixth layer is digital information, mainly including V2X information, digital map information, etc. One or more scene elements of each layer can be combined to determine multiple preset collection scenarios, thereby meeting the needs of covering multiple test scenarios and improving the accuracy of testing. In one possible implementation, the test unit can conduct testing in the following way: inputting image processing data into the perception algorithm and outputting perception results. For example, the image processing data can be images captured in the preset collection scenario and processed by the ISP algorithm, while the ground truth data is pre-annotated images, such as pedestrians, vehicles, roadblocks, etc. in the images. Specifically, after inputting the image processing data into the perception algorithm, perception results can be output, such as identifying pedestrians, vehicles, or roadblocks. The perceived data is compared with the true data. If the perceived data matches the true data, the perceived data is determined to be compliant; if the perceived data does not match the true data, the perceived data is determined to be non-compliant.

[0066] After obtaining the compliance perception data, it is equivalent to obtaining accurate recognition results for the preset collection scenario. The compliance perception data can be used to test the control algorithm. When the test results of the control algorithm do not meet the requirements of intelligent driving, the algorithm can be further iterated and optimized by adjusting its parameters.

[0067] Optionally, if the proportion of substandard sensing data is high, it indicates that the test results of the sensing algorithm do not meet the accuracy requirements. In this case, the parameters of the sensing algorithm can be adjusted, and the sensing algorithm can be further iterated and optimized based on the substandard sensing data.

[0068] In intelligent driving, the operational design domain of the intelligent driving system can be pre-defined based on actual needs. This domain defines the effective operating conditions of the intelligent driving system and includes factors such as vehicle status, road type, traffic conditions, and weather conditions. Based on this, after testing the perception algorithm and obtaining compliant perception data, the testing unit can compare this compliant data with the operational design domain to determine the in-domain perception data that meets the domain's requirements and the out-of-domain perception data that does not. Then, the control algorithm is tested based on the in-domain perception data.

[0069] In one possible implementation, the operational design domain can be set according to testing requirements, and the intelligent driving system may also operate effectively outside the operational design domain. Therefore, requirements assessment can be performed based on external perception data to determine whether to expand the operational design domain to meet more testing needs.

[0070] Optionally, the perception algorithm can be a single-sensor perception algorithm, where sensor data collected by a single sensor is input into an ISP algorithm to obtain image processing data. This image processing data is then input into the perception algorithm to output a perception result. The accuracy of the single-sensor perception algorithm is determined based on the perception result and the ground truth data. Alternatively, the perception algorithm can be a perception fusion algorithm. Perception fusion algorithms can collaboratively process data from multiple sensors (such as cameras, LiDAR, etc.) to overcome the limitations of a single sensor and generate more accurate perception results. In this case, sensor data collected by multiple sensors needs to be processed by an ISP algorithm to obtain image processing data, which is then input into the perception algorithm to output a perception result. The accuracy of the perception algorithm is determined based on the perception result and the ground truth data.

[0071] When the data acquisition unit collects data from a preset scenario using vehicle sensors, it needs to control the vehicle to travel a certain distance to collect data from the preset scenario. In one possible implementation, the actual planned mileage distribution corresponding to the preset scenario can be determined based on the city level and city mileage distribution, i.e., the actual mileage collected while controlling the vehicle's movement. That is, a correspondence can be pre-established between different city levels, the city mileage distribution for each city, and the actual planned mileage distribution for each city. Then, when collecting data from the preset scenario, the actual planned mileage distribution corresponding to the preset scenario is determined based on the city level and city mileage distribution, within the pre-established correspondence. The city level can be divided into first-tier, second-tier, third-tier, and fourth / fifth-tier cities, and the city mileage distribution can include total city mileage, total city road mileage, total highway mileage, and total mileage of other roads. Based on this, the determined actual planned mileage distribution for the preset scenario can include: total planned mileage, planned city road mileage, planned highway mileage, and planned mileage of other roads. By determining the actual planned mileage distribution corresponding to the preset scenario, a large amount of test scenario data can be covered, improving the accuracy of the test.

[0072] Table 1 provides the city mileage distribution for testing in several different cities, as well as the actual planned mileage distribution for the predetermined data collection scenarios. The proportion of different planned mileages can also be determined based on the city mileage distribution and the actual planned mileage distribution. The unit of mileage is kilometers (km).

[0073] Table 1

[0074]

[0075] In one possible implementation, the data acquisition unit can collect sensor data from a preset scenario using sensors on a mass-produced vehicle. Here, "mass-produced vehicle" refers to the vehicle theoretically controlled by the intelligent driving algorithm, i.e., the vehicle targeted for application during the development of the algorithm. Typically, there are two approaches to vehicle-based data acquisition: the first is using a modified vehicle, which is not the vehicle used in developing the intelligent driving algorithm. The installation position and viewing angle of the cameras on a modified vehicle differ from those on a mass-produced vehicle, potentially affecting the test results of the intelligent driving algorithm in actual use. The second approach is using a mass-produced vehicle, resulting in more reliable algorithms and test results.

[0076] In one possible implementation, the data acquisition unit can acquire sensor data obtained by the vehicle's sensors in a preset acquisition scenario using a bypass mode. The bypass mode is a data acquisition method that allows the copying of necessary data without interfering with the original data communication. For example, the data acquisition unit can acquire Raw data collected by the image sensors of a mass-produced vehicle using the bypass mode for subsequent testing and optimization of the ISP algorithm.

[0077] After the data processing unit processes the sensor data and ground truth data, it obtains the target sensor data and target ground truth data. The data processing unit can then construct a simulation environment based on this data, allowing the testing unit to test the intelligent driving algorithm within the simulation environment. Optionally, to meet diverse testing needs and expand the diversity of simulation environments, the data processing unit can perform data generalization or data reconstruction on the target sensor data and / or target ground truth data. Based on the generalized or reconstructed data, more simulation environments can be created, improving the coverage of test scenarios. The testing unit can deploy intelligent driving algorithms such as perception algorithms, traffic control algorithms, and ISP algorithms in the simulation environment to perform simulation testing of these algorithms.

[0078] Data generalization can be understood as generalizing from two known data points to all data within the range of those two known data points. For example, taking the distance between a vehicle and an obstacle as an example, if the first data collection shows a distance of 20 meters and the second data collection shows a distance of 5 meters, then the distance between the vehicle and the obstacle can be generalized to any value between 5 meters and 20 meters, thus allowing for the generation of more test scenarios based on the generalized data. Data reconstruction can be understood as reconstructing new data formats by splitting or merging existing data.

[0079] Optionally, the test unit can test the ISP algorithm based on target sensor data to obtain image processing data. The ISP algorithm, perception algorithm, and control algorithm can be simulated and tested in a simulation scenario. Alternatively, the target sensor data and target ground truth data can be fed back into the system of the ISP algorithm, perception algorithm, and control algorithm for backfeed testing.

[0080] Based on this, this application provides a schematic diagram for testing intelligent driving algorithms. See also... Figure 3 As shown, Figure 3 This is a schematic diagram illustrating a method for testing intelligent driving algorithms, as provided in an embodiment of this application.

[0081] This method can be executed by a test unit, and the intelligent driving algorithm can include perception algorithms and control algorithms.

[0082] The testing unit inputs image processing data and ground truth data into the perception algorithm to obtain compliant and non-compliant perception data. The perception algorithm is then further iteratively optimized based on the non-compliant perception data.

[0083] Once the compliance perception data is obtained, it can be compared with the operational design domain to determine the domain-specific perception data that meets the operational design domain requirements and the domain-specific perception data that does not. Based on the domain-specific perception data, a requirements assessment can be conducted to determine whether the operational design domain should be expanded to meet more testing needs. Then, the control algorithm is tested based on the domain-specific perception data.

[0084] When testing the control algorithm, vehicle bus data and in-domain perception data can be used as inputs to enable the algorithm to output control commands for the vehicle. The test results are then determined based on these control commands and the requirements of intelligent driving. Vehicle bus data refers to communication data transmitted through the in-vehicle bus network, enabling information sharing and collaborative control among various systems within the vehicle. If the test results of the control algorithm do not meet the requirements of intelligent driving, the algorithm can be further iterated and optimized by adjusting its parameters.

[0085] In one possible implementation, multiple modules can be set up in the data processing unit and the testing unit, each module implementing a corresponding function to meet the functional requirements of the data processing unit and the testing unit. See Table 2 for details. The data processing unit includes a data collection module, a data management module, a labeling module, etc., while the testing unit may include a training module, a refeeding module, an evaluation module, etc.

[0086] The system provided in this application embodiment can cover various testing scenarios and utilize sensors to collect data from real testing scenarios, thereby improving testing accuracy. It can meet diverse testing needs, such as reflow testing, simulation testing, and real vehicle testing.

[0087] Table 2

[0088]

[0089] Furthermore, embodiments of this application also provide a data-driven intelligent driving testing system. See also Figure 4 The diagram shown is a schematic of another data-driven intelligent driving test system provided in an embodiment of this application.

[0090] The system includes a data acquisition unit 101, a testing unit 102, and a data processing unit 103;

[0091] The data acquisition unit 101 can acquire vehicle bus data; determine a preset acquisition scenario and acquire sensor data from the preset scenario, including Raw data acquired by an image sensor; it can also acquire data from the preset acquisition scenario based on radar and process the acquired point cloud data to obtain ground truth data.

[0092] The data processing unit 103 can perform compliance processing on the sensor data collected by the data acquisition unit 101, such as desensitization and declassification. Data mining is performed based on the processed target sensor data. It can perform operations such as labeling the sensor data with ground truth, data slicing, data reconstruction, and generalization. Post-processing of the ground truth data, including filtering and registration, can be performed to obtain the target ground truth data.

[0093] Test unit 102 can use the raw data in the target sensor data to test the ISP algorithm and obtain image processing data. It can also test perception algorithms, control algorithms, etc., based on the image processing data and the target ground truth data. Test unit 103 can implement various test scenarios such as simulation testing and refeedback testing.

[0094] It should be noted that the specific implementation principles of the data acquisition unit 101, the testing unit 102, and the data processing unit 103 can be found in the above embodiments, and will not be described in detail here.

[0095] Based on the above system embodiments, this application also provides an intelligent driving testing method. See [link to previous document]. Figure 5 The diagram shown is a flowchart of an intelligent driving test method provided in an embodiment of this application.

[0096] Optionally, this method can be performed by an intelligent driving test system. The method may include the following steps:

[0097] S501: Determine the preset data collection scenario by combining one or more scenario elements from roads, traffic infrastructure, temporary operations, traffic participants, environment, and digital information.

[0098] S502: Based on the vehicle's sensors, the preset acquisition scenario is collected to obtain sensor data;

[0099] S503: Based on radar, collect data from a preset scene to obtain point cloud data;

[0100] S504: Process point cloud data to obtain ground truth data;

[0101] S505: Test the ISP algorithm based on sensor data to obtain image processing data;

[0102] S506: Test the perception algorithm based on image processing data and ground truth data to obtain qualified perception data;

[0103] S507: Test the regulatory control algorithm based on compliance perception data.

[0104] In one possible implementation, the sensor data includes raw data acquired by the vehicle's image sensors.

[0105] It should be noted that the beneficial effects of the intelligent driving testing method provided in this application embodiment can be found in the above system embodiment, and will not be repeated here.

[0106] Based on the above system and method embodiments, this application also provides an intelligent driving testing device. See also... Figure 6 The diagram shown is a schematic of an intelligent driving test device provided in an embodiment of this application.

[0107] The device 600 includes:

[0108] The scenario determination module 601 is used to determine the preset collection scenario by combining one or more scenario elements from roads, traffic infrastructure, temporary operations, traffic participants, environment, and digital information.

[0109] The first acquisition module 602 is used to acquire sensor data based on the vehicle's sensors in the preset acquisition scenario.

[0110] The second acquisition module 603 is used to acquire point cloud data based on the preset acquisition scene using radar.

[0111] The first processing module 604 is used to process the point cloud data to obtain true value data;

[0112] The testing module 605 is used to test the ISP algorithm based on the sensor data to obtain image processing data; to test the perception algorithm based on the image processing data and the ground truth data to obtain compliance perception data; and to test the regulatory control algorithm based on the compliance perception data.

[0113] In one possible implementation, the first acquisition module 602 is used to acquire Raw data collected by the image sensor of the vehicle.

[0114] In one possible implementation, the testing module 605 is used to input the image processing data into the perception algorithm and output the perception result; compare the perception result with the ground truth data; when the perception result matches the ground truth data, determine that the perception result is the qualified perception data; otherwise, determine that the perception result is the unqualified perception data; and iteratively optimize the perception algorithm based on the unqualified perception data.

[0115] In one possible implementation, the test module 605 is further configured to compare the compliance perception data with the operational design domain to determine the perception data within the domain and the perception data outside the domain; test the regulatory control algorithm based on the perception data within the domain; and conduct a demand assessment based on the perception data outside the domain to determine whether to expand the operational design domain.

[0116] In one possible implementation, the scenario determination module 601 is used to determine the actual planned mileage distribution corresponding to the preset data collection scenario based on the city level and city mileage distribution corresponding to the preset data collection scenario.

[0117] In one possible implementation, the first acquisition module 602 is used to acquire sensor data obtained by the sensors of the mass-produced vehicle in the preset acquisition scenario based on the bypass mode, wherein the mass-produced vehicle refers to the vehicle that the intelligent driving algorithm theoretically controls.

[0118] In one possible implementation, the system 600 further includes: a data processing unit, used to perform compliance processing and / or data slicing on the sensor data and the ground truth data to obtain target sensor data and target ground truth data.

[0119] In one possible implementation, the data processing unit is further configured to perform data generalization and / or data reconstruction based on the target sensor data and / or the target ground truth data to obtain a simulation scenario;

[0120] The test module 605 is used to test the ISP algorithm based on the target sensor data to obtain the image processing data; to perform simulation tests on the ISP algorithm, the perception algorithm, and the control algorithm in the simulation scenario; or, to feed back the target sensor data and the target ground truth data into the system of the ISP algorithm, the perception algorithm, and the control algorithm for feedback testing.

[0121] Based on the above method and device embodiments, this application also provides an electronic device. The following description will be given in conjunction with the accompanying drawings.

[0122] See Figure 7 , Figure 7 This is a schematic diagram of an electronic device provided in an embodiment of this application.

[0123] The device 700 includes: a memory 701 and a processor 702;

[0124] The memory 701 is used to store relevant program code;

[0125] The processor 702 is used to call the program code and execute the intelligent driving test method described in the above method embodiment.

[0126] Furthermore, this application embodiment also provides a computer-readable storage medium for storing a computer program for executing the intelligent driving test method described in the above method embodiments.

[0127] This application also provides a computer program product, which includes a computer program / instruction. When the computer program / instruction is executed by a processor, it implements the intelligent driving test method described in the above method embodiments.

[0128] It should be noted that the computer-readable medium described above in this application may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0129] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0130] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The device embodiments described above are merely illustrative. The units or modules described as separate components may or may not be physically separate. The components shown as units or modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the units or modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0131] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functions, and operations that may be implemented by methods, apparatuses, and devices according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0132] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0133] It should also be noted that, in this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0134] The steps of the methods or algorithms described in conjunction with the embodiments disclosed in this application can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0135] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data-driven intelligent driving testing system, characterized in that, The system includes: a data acquisition unit and a testing unit; The data acquisition unit is used to determine a preset acquisition scenario by combining one or more scenario elements from roads, traffic infrastructure, temporary operations, traffic participants, environment, and digital information; to acquire sensor data based on vehicle sensors; to acquire point cloud data based on radar data based on the preset acquisition scenario; and to process the point cloud data to obtain ground truth data. The testing unit is used to test the image signal processing (ISP) algorithm based on the sensor data to obtain image processing data; to test the perception algorithm based on the image processing data and the ground truth data to obtain compliant perception data; and to test the regulatory control algorithm based on the compliant perception data.

2. The system according to claim 1, characterized in that, The sensor data includes: Raw data collected by the vehicle's image sensor.

3. The system according to claim 1, characterized in that, The testing unit is used to input the image processing data into the perception algorithm and output the perception result; compare the perception result with the true value data, and when the perception result matches the true value data, determine that the perception result is the qualified perception data; Otherwise, the perception result is determined to be substandard perception data; The sensing algorithm is iteratively optimized based on the substandard sensing data.

4. The system according to claim 3, characterized in that, The testing unit is also used to compare the compliance perception data with the operational design domain to determine the perception data within the domain and the perception data outside the domain. The control algorithm was tested based on the sensing data within the domain. Based on the external sensing data, a requirements assessment is conducted to determine whether the operational design domain should be expanded.

5. The system according to claim 1, characterized in that, The data acquisition unit is used to determine the actual planned mileage distribution corresponding to the preset acquisition scenario based on the city level and city mileage distribution corresponding to the preset acquisition scenario.

6. The system according to claim 1, characterized in that, The data acquisition unit is used to acquire sensor data obtained by the sensors of the mass-produced vehicle in the preset acquisition scenario based on the bypass mode. The mass-produced vehicle refers to the vehicle that the intelligent driving algorithm theoretically controls.

7. The system according to claim 1, characterized in that, The system further includes a data processing unit, used to perform compliance processing and / or data slicing on the sensor data and the ground truth data to obtain target sensor data and target ground truth data.

8. The system according to claim 7, characterized in that, The data processing unit is also used to perform data generalization and / or data reconstruction based on the target sensor data and / or the target true value data to obtain a simulation scene; The testing unit is used to test the ISP algorithm based on the target sensor data to obtain the image processing data; The ISP algorithm, the perception algorithm, and the control algorithm are simulated and tested in the simulation scenario; or, The target sensor data and the target true value data are fed back into the system of the ISP algorithm, the perception algorithm and the planning and control algorithm for feedback testing.

9. A method for testing intelligent driving, characterized in that, The method includes: The preset data collection scenario is determined by combining one or more scenario elements from roads, traffic infrastructure, temporary operations, traffic participants, environment, and digital information. The vehicle's sensors collect data from the preset data acquisition scenario to obtain sensor data. Based on radar, point cloud data is collected from the preset acquisition scenario, and the point cloud data is processed to obtain ground truth data. The ISP algorithm was tested based on the sensor data to obtain image processing data; The perception algorithm is tested based on the image processing data and the ground truth data to obtain compliant perception data; The regulatory control algorithm was tested based on the compliance perception data.

10. The method according to claim 9, characterized in that, The sensor data includes: Raw data collected by the vehicle's image sensor.

Images