Data storage method, road test system and computer readable storage medium

Abstract

The invention provides a data storage method, a road test system and a computer readable storage medium. The method comprises the steps that an acquisition module acquires test data; the processing module filters the test data according to the data filtering model to obtain first filtering data; the cloud server trains a data filtering model according to the first filtering data and the test data to obtain updated model parameters; the processing module filters the test data according to the updated model parameters and the data filtering model to obtain second filtering data; the storage module stores the second filtering data. According to the method, the data storage amount of the test vehicle in the road test process can be reduced, and the data storage efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of road testing technology, specifically to a data storage method, a road testing system, and a computer-readable storage medium. Background Technology

[0002] In advanced autonomous driving solutions, various data collected by sensors such as LiDAR, cameras, and millimeter-wave radar are the foundation for autonomous driving devices to perceive the external environment.

[0003] Currently, a common method for storing test data involves transmitting all data collected by test vehicles offline and storing it on a cloud server, requiring manual annotation and analysis. This data includes a significant amount of redundant data. This redundancy consumes storage space on both the cloud server and the test vehicle, resulting in low data storage efficiency. Furthermore, it consumes computing resources on both the cloud server and the test vehicle, impacting the accuracy and timeliness of data analysis during processing. Summary of the Invention

[0004] This application provides a data storage method, a road testing system, and a computer-readable storage medium. The method can reduce the amount of data stored during road testing of test vehicles and improve data storage efficiency.

[0005] In a first aspect, embodiments of this application disclose a data storage method applied to a road testing system. The road testing system includes a test vehicle and a cloud server. The test vehicle includes a data acquisition module, a processing module, and a storage module. The data acquisition module includes one or more combinations of LiDAR, a camera, millimeter-wave radar, a positioning module, and an inertial measurement unit, comprising:

[0006] The acquisition module collects test data;

[0007] The processing module filters the test data according to the data filtering model to obtain the first filtered data;

[0008] The cloud server trains the data filtering model based on the first filtered data and the test data to obtain updated model parameters.

[0009] The processing module filters the test data according to the updated model parameters and the data filtering model to obtain the second filtered data;

[0010] The storage module stores the second filtered data.

[0011] In some embodiments, the acquisition module includes a lidar, and the acquisition module acquires test data, including: the cloud server generating multiple preset operating parameters for the lidar, wherein the multiple preset operating parameters include one or more combinations of preset voltage, preset power, preset scan frame rate, and preset point frequency; the lidar acquires data according to the multiple preset operating parameters to obtain a point cloud, wherein the test data is the point cloud. The cloud server can control the lidar on the test vehicle to start scanning, and the initial operating state of the lidar can be controlled based on the multiple preset operating parameters.

[0012] In some embodiments, the processing module filters the test data according to a data filtering model to obtain first filtered data, including: the processing module performing feature statistics on the point cloud according to the data filtering model to obtain multiple point cloud parameters of the point cloud, wherein the multiple point cloud parameters include one or more combinations of point cloud density, noise detection rate, and target object missed detection rate; the processing module filters the point cloud according to the data filtering model and the multiple point cloud parameters to obtain filtered point cloud, wherein the filtered point cloud is the first filtered data. Using multiple point cloud parameters as evaluation benchmarks to perform multi-dimensional evaluation of the point cloud acquired by LiDAR, and then filtering the point cloud based on the multi-dimensional evaluation results, can effectively reduce the storage capacity occupied by redundant data and improve storage efficiency. Furthermore, the types of multiple point cloud parameters statistically analyzed by the data filtering model can be adjusted according to test requirements, improving the flexibility of the data filtering strategy.

[0013] In some embodiments, the processing module filters the test data according to a data filtering model to obtain first filtered data, including: the processing module segments the point cloud according to the data filtering model to obtain multiple sub-point clouds, wherein each sub-point cloud corresponds to a target object, the target object includes static target objects and dynamic target objects, the static target object includes one or more combinations of road signs, buildings and ground, and the dynamic target object includes one or more combinations of vehicles and pedestrians; the processing module performs target identification on each sub-point cloud according to the data filtering model to determine the static sub-point cloud among the multiple sub-point clouds, wherein the static sub-point cloud is the sub-point cloud corresponding to the static target object, and the static sub-point cloud is the first filtered data. Based on the data filtering model, static sub-point clouds in the point cloud can be effectively filtered out. Static target objects are generally static environmental targets within the test area. Obtaining static sub-point clouds through filtering can achieve effective collection of road environment information, avoid storing redundant data (such as sub-point clouds corresponding to dynamic target objects), and improve data storage efficiency.

[0014] In some embodiments, the cloud server trains the data filtering model based on the first filtered data and the test data to obtain updated model parameters, including: the cloud server comparing the test data and the first filtered data to obtain a comparison result; and the cloud server training the data filtering model based on the comparison result to obtain updated model parameters. The cloud server deploys the same data filtering model as the acquisition module. Based on the comparison result, the cloud server can obtain the filtering effect of the data filtering model, and then adjust the various model parameters of the data filtering model to optimize the data filtering model, thereby improving the data filtering efficiency and accuracy of the data filtering model at the vehicle end.

[0015] In some embodiments, the cloud server trains the data filtering model based on the first filtered data and the test data to obtain updated model parameters, including: the cloud server filtering the test data according to the data filtering model to obtain third filtered data; the cloud server comparing the first filtered data and the third filtered data to obtain a comparison result; and the cloud server training the data filtering model based on the comparison result to obtain updated model parameters. The cloud server deploys the same data filtering model as the acquisition module. The computing power of the cloud server is greater than that of the processing module of the test vehicle, and the quality of the third filtered data obtained by the cloud server based on the data filtering model is stronger than that of the first filtered data. Based on the comparison result, the cloud server can adjust various model parameters of the data filtering model to optimize the data filtering model, making it more compatible with the computing power of the processing module of the test vehicle, thereby improving the data filtering efficiency and accuracy of the data filtering model on the vehicle side.

[0016] In some embodiments, the test vehicle further includes a status management module, and the method further includes: the status management module acquiring the operating status data of the acquisition module, wherein the operating status data includes one or more combinations of data timestamps, operating current, and operating voltage; the status management module generating fault warning information based on the operating status data of the acquisition module, and sending the fault warning information to the cloud server. This method enables the cloud server to monitor the operating status of the acquisition module, promptly determine whether the acquisition module is in a fault state, reduce the impact of the acquisition module being in a fault state on the data filtering process, and improve the accuracy of the data filtering process.

[0017] Secondly, embodiments of this application disclose a road testing system, including a test vehicle and a cloud server. The test vehicle includes a data acquisition module, a processing module, and a storage module. The data acquisition module includes one or more of the following: lidar, camera, millimeter-wave radar, positioning module, and inertial measurement unit.

[0018] The acquisition module is used to collect test data;

[0019] The processing module is used to filter the test data according to the data filtering model to obtain the first filtered data;

[0020] The cloud server is used to train the data filtering model based on the first filtered data and the test data to obtain updated model parameters.

[0021] The processing module is further configured to filter the test data according to the updated model parameters and the data filtering model to obtain second filtered data;

[0022] The storage module is used to store the second filtered data.

[0023] Thirdly, embodiments of this application disclose a computer-readable storage medium storing a computer program that, when executed, implements the data storage method as described in any of the first aspects.

[0024] This application discloses a data storage method. The method deploys a data filtering model with different data storage strategies on the vehicle side, and filters test data based on the data filtering model, thereby reducing the storage volume occupied by redundant data and improving data storage efficiency. Furthermore, the data filtering model is trained using a cloud server, further improving the accuracy of data filtering and data storage efficiency. Attached Figure Description

[0025] Figure 1 A flowchart illustrating a data storage method provided in an embodiment of this application;

[0026] Figure 2 A schematic diagram of a road testing system provided in an embodiment of this application;

[0027] Figure 3 A schematic diagram of a data acquisition module provided in an embodiment of this application;

[0028] Figure 4 A schematic diagram of a test vehicle provided for an embodiment of this application;

[0029] Figure 5 This is a schematic diagram of a test scenario provided in an embodiment of this application. Detailed Implementation

[0030] The technical solutions of this application will now be described clearly and in detail with reference to the accompanying drawings. In the description of the embodiments of this application, unless otherwise stated, " / " indicates "or," for example, A / B can mean A or B. "And / or" in the text is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, in the description of the embodiments of this application, "multiple" refers to two or more. Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature.

[0031] Currently, common data storage solutions for test data collected during vehicle operation involve storing the data in the vehicle's local storage module or uploading it to a cloud server's storage module, followed by manual annotation and analysis. On one hand, storing all test data places high demands on the test vehicle's storage capacity; on the other hand, test data often includes redundant data, which consumes the test vehicle's computing resources, thus affecting the accuracy and timeliness of data analysis during processing.

[0032] In one embodiment, this application discloses a road testing system, combined with Figure 1 and Figure 2The road testing system includes a test vehicle and a cloud server. The test vehicle includes a data acquisition module, a processing module, a first storage module, a first transmission module, and a status management module. The cloud server includes a second transmission module, a cloud management module, a test configuration module, a status monitoring module, a model training module, and a second storage module. The cloud management module is used to maintain at least one of the following functions of the cloud server: permission management function, process management function, and operation management function. In one example, the acquisition module collects test data and transmits it to the processing module. The processing module filters the test data according to a data filtering model to obtain first filtered data. The processing module also transmits the first filtered data and the test data to a second transmission module on the cloud server via the first transmission module. The second transmission module transmits the first filtered data and the test data to a model training module, which trains a data filtering model based on the first filtered data and the test data to obtain updated model parameters. The model training module also transmits the updated model parameters back to the first transmission module via the second transmission module. The first transmission module transmits the updated model parameters back to the processing module. The processing module further filters the test data based on the updated model parameters and the data filtering model to obtain second filtered data. A first storage module on the test vehicle stores the second filtered data. The first and second transmission modules interact via HTTPS (Hypertext Transfer Protocol Secure) to achieve bidirectional data transmission between the test vehicle and the cloud server.

[0033] In some embodiments, the processing module of the test vehicle may be a Field-Programmable Gate Array (FPGA), a System-on-Chip (SoC), a Central Processing Unit (CPU), a Network Processor (NP), a Digital Signal Processing Circuit, a Micro Controller Unit (MCU), an Application-Specific Integrated Circuit (ASIC), or any combination thereof, for implementing the relevant functions. In some embodiments, the first storage module is a hard disk drive (HDD) or a solid-state drive (SSD), and the second storage module is also an HDD or an SSD.

[0034] based on Figure 1 and Figure 2 The road testing system shown in the application discloses a data storage method, which is applied to the road testing system described above. The data storage method includes the following scheme:

[0035] S110, the acquisition module collects test data.

[0036] like Figure 3 As shown, the acquisition module includes multiple sensors, including at least one lidar and an auxiliary acquisition module. The auxiliary acquisition module includes a combination of one or more sensors selected from a camera, millimeter-wave radar, a positioning module, and an inertial measurement unit (IMU). Different types of sensors are used to acquire different types of test data. The lidar is used to acquire point clouds, the millimeter-wave radar is used to acquire millimeter-wave data, and the camera is used to acquire images. The IMU includes an accelerometer and a gyroscope, and is used for pre-integration to obtain the vehicle's pose change within each time step. The positioning module is a GPS-based positioning sensor used to acquire GPS timestamps and vehicle positioning data.

[0037] In one embodiment, the acquisition module collects test data including: a test configuration module on a cloud server generates test configuration instructions and transmits the test configuration instructions to a second transmission module; the second transmission module transmits the test configuration instructions to a first transmission module; and the first transmission module transmits the test configuration instructions to the acquisition module. In one example, such as... Figure 4 As shown, the test vehicle is equipped with a first millimeter-wave radar 101, a second millimeter-wave radar 102, a first lidar 201, a second lidar 202, a first camera 301, a second camera 302, a positioning module (not shown in the figure), and an inertial measurement unit (not shown in the figure). The data acquisition module controls one or more sensors from the lidar, camera, millimeter-wave radar, positioning module, and inertial measurement unit to acquire data according to test configuration instructions, obtaining test data and transmitting the test data to the processing module. Based on the test configuration instructions, different types of sensors can be controlled to acquire data, thereby controlling the type of test data acquired by the test vehicle, reducing the generation of redundant data, and meeting the test requirements of different test scenarios. Alternatively, based on the test configuration instructions, the vehicle's trajectory can be controlled simultaneously with the acquisition module acquiring test data.

[0038] In one embodiment, taking a lidar as an example, the control of the data acquisition process by the test configuration module over the acquisition module is illustrated. The test configuration instructions include multiple preset operating parameters for the lidar, which include one or more combinations of preset scan time, preset voltage, preset power, preset scan frame rate, preset vertical field of view, preset horizontal field of view, and preset point frequency. At least one lidar acquires data according to these preset operating parameters to obtain a point cloud, wherein the test data is the point cloud. In one example, combined with... Figure 4 and Figure 5 ,like Figure 5 The test scenario shown includes test vehicle 1, test vehicle 2, static target 401, and static target 402. The test configuration module of the cloud server generates a first test configuration instruction and a second test configuration instruction, each including multiple preset operating parameters for the LiDAR. The first test configuration instruction is transmitted to the first transmission module of test vehicle 1 via the second transmission module of the cloud server, and the second test configuration instruction is transmitted to the first transmission module of test vehicle 2 via the second transmission module of the cloud server. The first transmission module of test vehicle 1 transmits the first test configuration instruction to the second LiDAR 202 in the acquisition module of test vehicle 1. The second LiDAR 202 on test vehicle 1 scans the static target 402 according to the first test configuration instruction. The first transmission module of test vehicle 2 transmits the second test configuration instruction to the first LiDAR 201 and the second LiDAR 202 in the acquisition module of test vehicle 2. The first LiDAR 201 on test vehicle 2 scans the static target 401 according to the second test configuration instruction, and the second LiDAR 202 on test vehicle 2 scans the static target 402 according to the second test configuration instruction. Alternatively, the first transmission module of test vehicle 2 transmits the second test configuration command to the first lidar 201 in the acquisition module of test vehicle 2. The first lidar 201 on test vehicle 2 scans the static target 401 according to the second test configuration command. Based on the cloud server, multiple test vehicles can be controlled to scan simultaneously, and different numbers of lidars on each test vehicle can be controlled to scan, or lidars at different installation positions can be controlled to scan. In addition, based on multiple preset operating parameters, the initial operating state of each lidar can be controlled, and the amount of data generated by each lidar during one frame scan can be adjusted.

[0039] In one embodiment, before transmitting the test data to the processing module, the data storage method further includes: clock synchronizing one or more timestamps among the lidar data timestamp, camera data timestamp, millimeter-wave radar data timestamp, and inertial measurement unit data timestamp based on the GPS timestamp of the positioning module, in order to improve the accuracy of the test data.

[0040] The same data filtering model is deployed in the processing module of the test vehicle and the model training module of the cloud server. Before the processing module filters the test data based on the data filtering model to obtain the first filtered data, the data storage method includes: the test configuration module generates a preset storage data instruction and transmits the preset storage data instruction to the second transmission module; the second transmission module transmits the preset storage data instruction to the first transmission module; the first transmission module transmits the preset storage data instruction to the processing module; and the data filtering model configures its parameters according to the preset storage data instruction.

[0041] S120: The processing module filters the test data based on the data filtering model to obtain the first filtered data.

[0042] In one embodiment, the test data includes point clouds acquired by LiDAR, images acquired by a camera, millimeter-wave data acquired by millimeter-wave radar, vehicle positioning data acquired by a positioning module, and vehicle pose data acquired by an inertial measurement unit. The processing module configures a data filtering model based on received preset stored data instructions. The data filtering model filters the data according to its type, and the first filtered data includes data acquired by one or more sensors. In one example, the processing module filters the test data based on the data filtering model to obtain the first filtered data, which includes: the processing module filtering the test data based on the data filtering model to obtain the first filtered data, wherein the first filtered data includes point clouds acquired by LiDAR and images acquired by a camera; or the first filtered data includes point clouds acquired by LiDAR, vehicle positioning data acquired by the positioning module, and vehicle pose data acquired by the inertial measurement unit.

[0043] In one embodiment, the preset storage data instruction is used to enable the data filtering model to perform multi-dimensional evaluation on the test data, and the data filtering model filters the test data according to the multi-dimensional evaluation results. Taking the point cloud collected by the lidar as an example, the processing module filters the test data according to the data filtering model to obtain the first filtered data, including: the processing module performs feature statistics on the point cloud based on the data filtering model to obtain multiple point cloud parameters of the point cloud, where the multiple point cloud parameters include one or a combination of one or more common point cloud parameter types such as point cloud density A1, noise detection rate A2, target object missed detection rate A3, detection distance A4, and point cloud frame rate A5; the processing module filters the point cloud according to the data filtering model and the multiple point cloud parameters to obtain the filtered point cloud, where the filtered point cloud is the first filtered data. In one example, the processing module filters the point cloud according to the data filtering model and the multiple point cloud parameters to obtain the filtered point cloud, including: inputting point cloud parameters such as the noise detection rate A2 and the target object missed detection rate A3, which are used to characterize the overall quality of the point cloud, into the data filtering model, and the data filtering model performs multi-dimensional evaluation on the point cloud according to the noise detection rate A2 and the target object missed detection rate A3 to obtain the first dimension evaluation result and the second dimension evaluation result. For example, when 0% < A2 ≤ 50%, the first dimension evaluation result is marked as 0 points; when 50% < A2 ≤ 75%, the first dimension evaluation result is marked as 10 points; when 75% < A2 ≤ 85%, the first dimension evaluation result is marked as 20 points; when 85% < A2 ≤ 100%, the first dimension evaluation result is marked as 30 points; when 0% < A3 ≤ 5%, the second dimension evaluation result is marked as 30 points; when 5% < A2 ≤ 10%, the second dimension evaluation result is marked as 20 points; when 10% < A2 ≤ 20%, the second dimension evaluation result is marked as 10 points; when 20% < A2, the second dimension evaluation result is marked as 0 points. The above-mentioned thresholds corresponding to the scores can be adjusted according to actual test requirements. When the sum of the first dimension evaluation result and the second dimension evaluation result is greater than or equal to 50 points, the point cloud is retained as the filtered point cloud; or when the sum of the first dimension evaluation result and the second dimension evaluation result is equal to 60 points, the point cloud is retained as the filtered point cloud. In another example, the processing module filters the test data according to the data filtering model to obtain the first filtered data, including: the processing module divides the point cloud into multiple sub-point clouds according to the field of view range (horizontal field of view range or vertical field of view range); then performs multi-dimensional evaluation on each sub-point cloud based on the data filtering model to obtain the first dimension evaluation result of each sub-point cloud and the second dimension evaluation result of each sub-point cloud; when the sum of the first dimension evaluation result of each sub-point cloud and the second dimension evaluation result of each sub-point cloud is greater than or equal to 50 points, the sub-point cloud is retained as the filtered point cloud; or when the sum of the first dimension evaluation result of each sub-point cloud and the second dimension evaluation result of each sub-point cloud is equal to 60 points, the sub-point cloud is retained as the filtered point cloud.Using multiple point cloud parameters acquired by LiDAR as evaluation benchmarks, a multi-dimensional evaluation of the point cloud is performed. The point cloud is then filtered based on the multi-dimensional evaluation results, effectively reducing the amount of redundant data stored and improving storage efficiency. Furthermore, adjusting the types of point cloud parameters statistically analyzed by the data filtering model according to testing requirements enhances the flexibility of the preset data storage strategy.

[0044] In one embodiment, based on actual testing requirements, the target objects in the test scenario include one or more combinations of road signs, vehicles, buildings, pedestrians, or ground. Preset storage data instructions are used to enable the data filtering model to retain test data corresponding to different types of target objects. In one example, the processing module filters the test data according to the data filtering model to obtain first filtered data, including: the processing module segments the point cloud according to the data filtering model to obtain multiple sub-point clouds, where each sub-point cloud corresponds to a target object. Target objects include static and dynamic targets. Static targets include one or more combinations of road signs, buildings, and ground (such as curves, lane lines, ramps, road edges, or other types of roads). Dynamic targets include one or more combinations of vehicles (other than the test vehicle body) and pedestrians. The processing module performs target identification on each sub-point cloud according to the data filtering model to determine the static sub-point cloud among the multiple sub-point clouds. The static sub-point cloud is the sub-point cloud corresponding to the static target object, and the static sub-point cloud is the first filtered data. Based on the data filtering model, the static sub-point clouds in the point cloud can be effectively filtered out. Static targets are generally static environmental targets within the test area. Therefore, obtaining static sub-point clouds through filtering can effectively collect road environment information, avoid storing redundant data (such as sub-point clouds corresponding to dynamic targets), and improve data storage efficiency.

[0045] Prior to step S130, the data storage method further includes: the processing module transmitting the first filtered data and test data to the first transmission module, and the first transmission module transmitting the first filtered data and test data to the second transmission module in the cloud server.

[0046] S130: The cloud server trains a data filtering model based on the first filtered data and the test data, and obtains the updated model parameters.

[0047] In one embodiment, the second transmission module transmits the first filtered data and test data to the model training module, wherein the model training module is equipped with the same data filtering model as the processing module of the test vehicle. The model training module filters the test data according to the data filtering model to obtain third filtered data; the model training module compares the first filtered data and the third filtered data to obtain a comparison result; the model training module trains the data filtering model based on the comparison result to obtain updated model parameters. The process of the model training module filtering the test data according to the data filtering model to obtain the third filtered data is described in step S120. Based on the comparison result, the cloud server can adjust the various model parameters of the data filtering model to optimize the data filtering model, making it more compatible with the computing power of the processing module of the test vehicle, thereby improving the data filtering efficiency and accuracy of the data filtering model on the vehicle side.

[0048] In another embodiment, the second transmission module transmits the first filtered data and test data to the model training module, wherein the model training module is equipped with a data filtering model consistent with the processing module of the test vehicle. The model training module compares the first filtered data and the test data to obtain a comparison result; the model training module trains the data filtering model based on the comparison result to obtain updated model parameters. The cloud server, based on the comparison result, can obtain the filtering effect of the data filtering model, and then adjust the various model parameters of the data filtering model to optimize the data filtering model, thereby improving the data filtering efficiency and accuracy of the data filtering model on the vehicle side.

[0049] S140: The processing module filters the test data based on the updated model parameters and the data filtering model to obtain the second filtered data.

[0050] In one embodiment, the model training module transmits the updated model parameters to the second transmission module, the second transmission module transmits the updated model parameters to the first transmission module, and the first transmission module transmits the updated model parameters to the processing module. The processing module updates the data filtering model based on the updated model parameters to obtain the updated data filtering model; the processing module then filters the test data based on the updated data filtering model to obtain the second filtered data.

[0051] S150: The first storage module stores the second filtered data.

[0052] In one embodiment, the first storage module stores the second filtered data. Compared to directly storing test data, this improves storage efficiency and reduces the storage capacity occupied by redundant data. In one example, when the storage space of the first storage module is insufficient, it can also transmit the second filtered data to a second transmission module via a first transmission module. The second transmission module then transmits the second filtered data to the second storage module, which stores it via a cloud server. This reduces the storage capacity requirements of the first storage module on the test vehicle and prevents test data corruption and loss due to insufficient storage capacity. The second storage module can be used to store test data, updated model parameters, and a combination of one or more of the first and second filtered data for data backup.

[0053] In one embodiment, based on Figure 1 and Figure 2 The road testing system shown in this application discloses a data storage method, including the following scheme:

[0054] S210, The data acquisition module collects test data;

[0055] The specific process of step S210 is the same as that of step S110, and will not be repeated here.

[0056] The S220 and cloud server train a data filtering model based on the calibration data to obtain updated model parameters.

[0057] In one embodiment, the calibration data includes multiple test data sets with annotation information. The annotation information is used to indicate whether each sub-data set within each test data set (such as point clouds, images, vehicle pose data, vehicle positioning data, or millimeter-wave data, or sub-point clouds or sub-images) needs to be filtered or retained. The model training module filters the multiple test data sets based on the data filtering model, obtaining multiple fourth-filtered data sets, where each test data set corresponds to one fourth-filtered data set. The model training module compares the annotation information of each test data set with the corresponding fourth-filtered data set to obtain a comparison result. The model training module trains the data filtering model based on the comparison result to obtain updated model parameters.

[0058] S230: The processing module filters the test data according to the updated model parameters to obtain the fifth filtered data.

[0059] In one embodiment, the model training module transmits the updated model parameters to the test configuration module, the test configuration module transmits the updated model parameters to the first transmission module through the second transmission module, and the first transmission module transmits the updated model parameters to the processing module; the processing module updates the data filtering model according to the updated model parameters to obtain the updated data filtering model; the processing module filters the test data according to the updated data filtering model to obtain the fifth filtered data.

[0060] S240, the first storage module stores the fifth filter data.

[0061] In one embodiment, based on Figure 1 and Figure 2 The data storage method of the road testing system described in any of the above embodiments further includes: a status management module acquiring working status data of the processing module, working status data of the acquisition module, or storage status data of the storage module. The working status data of the processing module includes one or more combinations of operating voltage, processor operating frequency, or real-time computing power; the working status data of the acquisition module includes one or more combinations of data timestamp, operating current, and operating voltage; and the storage status data includes one or more combinations of remaining storage capacity, number of corrupted bits, write rate, and read rate. In one example, the status management module generates fault warning information based on the working status data of the acquisition module or the storage status data of the storage module, and sends the fault warning information to the status monitoring module of the cloud server. The status monitoring module provides a warning to the administrator in the cloud based on the fault warning information. This method enables the cloud server to monitor the working status of the processing module, acquisition module, or storage module of the test vehicle, promptly determine whether it is in a fault state, reduce the impact of a faulty acquisition module or storage module on the data filtering process, and improve the accuracy of the data filtering process.

[0062] In the embodiments described above, both the processing module of the test vehicle and the cloud server are equipped with artificial intelligence models (data filtering models). The cloud server controls the size and type of data collected by the test vehicle by sending test configuration instructions and preset storage data instructions to the test vehicle, thereby enabling data interaction between the test vehicle and the cloud server. This allows users to easily adjust the data collection and filtering strategies of different test vehicles in real time, meeting the needs of different types of road tests and test data, accelerating testing speed, and facilitating data analysis and comparison. Different data storage strategies can effectively prevent redundant data from occupying the storage capacity of the test vehicle's first storage module. Furthermore, the cloud server trains the data filtering model based on the data or calibration data uploaded by the test vehicle and sends the updated model parameters back to the test vehicle, further improving data storage efficiency and data filtering accuracy, and reducing the computing power requirements of the test vehicle's processing module.

[0063] In one embodiment, this application discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the data storage method shown in steps S110 to S150 or S210 to S240. In this embodiment, the processor is a central processing unit (CPU) of a cloud server.

[0064] Through the above description of the embodiments, those skilled in the art will understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0065] In the description of this application, it should be understood that, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. The terms “and / or” and “and / or” as used herein describe the relationship between related objects, indicating that three relationships may exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. The character “ / ” generally indicates that the preceding and following related objects are in an “or” relationship. The singular forms “a” and “an” are also intended to include the plural forms unless the context clearly indicates otherwise. When the terms “comprising” and / or “including” are used in this specification, they indicate the presence of the stated features, elements, and / or components, but do not exclude the presence or addition of one or more other features, elements, components, and / or combinations thereof, i.e., including any and all combinations of one or more of the associated listed items. Ordinal numbers such as “first” and “second” referenced in the embodiments of this application are merely identifiers and do not imply any particular order or relative importance. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances. The phrase "one or more embodiments" as used herein does not refer to the same embodiment, but rather to a combination of specific features, structures, or characteristics based on any suitable manner. The above are merely preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A data storage method, characterized in that, This invention relates to a road testing system, which includes a test vehicle and a cloud server. The test vehicle includes a data acquisition module, a processing module, and a storage module. The data acquisition module includes one or more of the following: LiDAR, camera, millimeter-wave radar, positioning module, and inertial measurement unit. The data storage method includes: The acquisition module collects test data; The processing module filters the test data according to the data filtering model to obtain the first filtered data; The cloud server trains the data filtering model based on the first filtered data and the test data to obtain updated model parameters. The processing module filters the test data according to the updated model parameters and the data filtering model to obtain the second filtered data; The storage module stores the second filtered data.

2. The method according to claim 1, wherein the acquisition module includes a lidar, characterized in that, The acquisition module collects test data, including: The cloud server generates multiple preset operating parameters for the lidar, wherein the multiple preset operating parameters include one or more combinations of preset voltage, preset power, preset scan frame rate and preset spot frequency; The lidar collects data according to multiple preset operating parameters to obtain a point cloud, wherein the test data is the point cloud.

3. The method according to claim 1 or 2, characterized in that, The processing module filters the test data according to the data filtering model to obtain first filtered data, including: The processing module performs feature statistics on the point cloud according to the data filtering model to obtain multiple point cloud parameters, wherein the multiple point cloud parameters include one or more combinations of point cloud density, noise detection rate and target object missed detection rate. The processing module filters the point cloud according to the data filtering model and multiple point cloud parameters to obtain a filtered point cloud, wherein the filtered point cloud is the first filtered data.

4. The method according to claim 1 or 2, characterized in that, The processing module filters the test data according to the data filtering model to obtain first filtered data, including: The processing module performs point cloud segmentation on the point cloud according to the data filtering model to obtain multiple sub-point clouds. Each sub-point cloud corresponds to a target object. The target object includes static target objects and dynamic target objects. The static target object includes one or more combinations of road signs, buildings and ground. The dynamic target object includes one or more combinations of vehicles and pedestrians. The processing module performs target identification on each of the sub-point clouds according to the data filtering model, and determines the static sub-point cloud among the multiple sub-point clouds, wherein the static sub-point cloud is the sub-point cloud corresponding to the static target object, and the static sub-point cloud is the first filtered data.

5. The method according to claim 1, characterized in that, The cloud server trains the data filtering model based on the first filtered data and the test data to obtain updated model parameters, including: The cloud server compares the test data and the first filtered data to obtain a comparison result. The cloud server trains the data filtering model based on the comparison results to obtain updated model parameters.

6. The method according to claim 1, characterized in that, The cloud server trains the data filtering model based on the first filtered data and the test data to obtain updated model parameters, including: The cloud server filters the test data according to the data filtering model to obtain the third filtered data; The cloud server compares the first filtered data and the third filtered data to obtain a comparison result. The cloud server trains the data filtering model based on the comparison results to obtain updated model parameters.

7. The method according to claim 1, wherein the test vehicle further comprises a status management module, characterized in that, The method further includes: The status management module acquires the working status data of the acquisition module, wherein the working status data includes one or more combinations of data timestamp, working power and working voltage; The status management module generates a fault warning signal based on the working status data of the acquisition module, and sends the fault warning signal to the cloud server.

8. A road testing system, characterized in that, It includes a test vehicle and a cloud server. The test vehicle includes a data acquisition module, a processing module, and a storage module. The data acquisition module includes one or more of the following: lidar, camera, millimeter-wave radar, positioning module, and inertial measurement unit. The acquisition module is used to collect test data; The processing module is used to filter the test data according to the data filtering model to obtain the first filtered data; The cloud server is used to train the data filtering model based on the first filtered data and the test data to obtain updated model parameters. The processing module is further configured to filter the test data according to the updated model parameters and the data filtering model to obtain second filtered data; The storage module is used to store the second filtered data.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, is used to implement the method as described in any one of claims 1 to 7.

Images