A method and system for testing the data recording function of DSSAD

Abstract

An embodiment of the present invention discloses a method and system for testing the data recording function of DSSAD, which relates to the technical field of autonomous driving testing. The method includes: collecting real vehicle data and / or simulated driving data of an autonomous vehicle through a data acquisition tool, wherein the collected data includes vehicle state data, driver operation data, autonomous driving system data, and environmental data; using a data playback tool to verify and annotate the collected data, wherein the annotation content includes collision events and events with collision risks; injecting the collected data into the DSSAD to be tested so that the DSSAD records relevant data according to the standard requirements; comparing the data recorded by the DSSAD with the annotated data to verify the data recording function of the DSSAD. It provides a complete test and verification for the data recording function of DSSAD and improves the flexibility of the test.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of autonomous driving testing, and in particular, to a method and system for testing the data recording function of DSSAD. Background Art

[0002] DSSAD (Data Storage System for Automated Driving) has functions such as data circular recording, collision risk event triggering judgment and recording, collision event triggering judgment and recording, and other functions specified by some standards.

[0003] There is no complete solution for how to test and verify the data recording function of DSSAD. Summary of the Invention

[0004] The embodiments of the present invention provide a method and system for testing the data recording function of DSSAD, providing a complete test and verification for the data recording function of DSSAD and improving the flexibility of testing.

[0005] In a first aspect, the embodiments of the present invention provide a method for testing the data recording function of DSSAD, including:

[0006] Collecting real vehicle data and / or simulated driving data of an autonomous driving vehicle through a data collection tool, where the collected data includes vehicle state data, driver operation data, autonomous driving system data, and environmental data;

[0007] Using a data playback tool to check and label the collected data, where the labeled content includes collision events and collision risk events;

[0008] Injecting the collected data into the DSSAD to be tested so that the DSSAD records relevant data according to standard requirements;

[0009] Comparing the data recorded by the DSSAD with the labeled data to verify the data recording function of the DSSAD.

[0010] In a second aspect, the embodiments of the present invention provide a system for testing the data recording function of DSSAD, including:

[0011] A data collection tool for collecting real vehicle data and / or simulated driving data of an autonomous driving vehicle, where the collected data includes vehicle state data, driver operation data, autonomous driving system data, and environmental data;

[0012] A data playback tool for verifying and annotating the collected data, where the annotation content includes collision events and collision-risk events;

[0013] A data injection tool for injecting the collected data into the DSSAD to be tested, so that the DSSAD records relevant data according to the standard requirements;

[0014] A verification and comparison tool for comparing the data recorded by the DSSAD with the annotated data to verify the data recording function of the DSSAD.

[0015] In summary, the embodiments of the present invention provide a method and system for testing the data recording function of a DSSAD, establishing a complete testing process from test data collection, test data playback, test data injection to comparison and verification of recording results. Each stage is independently completed by a separate tool or device, without the need to bind the DSSAD to a real vehicle or a driving simulation system in the traditional way. Instead, it is possible to test the data recording situation of any vehicle at any time period at any time, improving the flexibility of testing the data recording function of the DSSAD. In addition, through the data collection tool and the simulated driving system, it is possible to achieve full coverage of collision events, collision-risk events, and normal data; through data playback, preliminary verification and annotation of the collected data can be carried out to improve the quality of test data; through various comparison and verification mechanisms, it is possible to gradually determine the data recording situation of the DSSAD for key data, obtain accurate test results, and provide accurate data entry points (such as inconsistent events) for the improvement of the DSSAD, thereby continuously optimizing the data recording function of the DSSAD. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 is a schematic structural diagram of a system for testing the data recording function of a DSSAD provided by an embodiment of the present invention;

[0018] Figure 2 is a flowchart of a method for testing the data recording function of a DSSAD provided by an embodiment of the present invention;

[0019] Figure 3 is a schematic diagram of a simulated driving system provided by an embodiment of the present invention;

[0020] Figure 4It is a schematic diagram of the acquisition path of data in different scenarios provided by an embodiment of the present invention;

[0021] Figure 5 It is a flowchart of data acquisition provided by an embodiment of the present invention;

[0022] Figure 6 It is a flowchart of data playback provided by an embodiment of the present invention;

[0023] Figure 7 It is a schematic diagram of data injection provided by an embodiment of the present invention;

[0024] Figure 8 It is a flowchart of data injection provided by an embodiment of the present invention;

[0025] Figure 9 It is a flowchart of data comparison and verification provided by an embodiment of the present invention;

[0026] Figure 10 It is a schematic diagram of the structure of a driver engagement prediction model provided by an embodiment of the present invention. Detailed implementation manners

[0027] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope protected by the present invention.

[0028] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0029] In the description of the present invention, it should also be noted that unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0030] An embodiment of the present invention provides a method for testing the data recording function of DSSAD. To illustrate this method, a data recording function test system of DSSAD that supports the implementation of this method is introduced first. Figure 1 is a schematic structural diagram of a data recording function test system of DSSAD provided by an embodiment of the present invention. As Figure 1 shown, the system includes a data acquisition tool, a data playback tool, a data injection tool, and a verification and comparison tool.

[0031] Among them, the data acquisition tool is used to acquire real vehicle data or simulated driving data of an autonomous vehicle. The real vehicle data comes from the vehicle itself, and the simulated driving data comes from a simulated driving system. The acquired data is input into the data playback tool for verification and annotation, and the played-back data can be viewed on the upper computer. The annotated data is injected into the DSSAD to be tested through the data injection tool, and the DSSAD records it. The verification and comparison tool is used to compare the data recorded by the DSSAD with the annotated data to verify whether the data recording function of the DSSAD meets the requirements.

[0032] It should be noted that the DSSAD in this embodiment can be a DSSAD integrated inside an autonomous vehicle, or a DSSAD tool or box independent of the autonomous vehicle. This embodiment does not make specific limitations.

[0033] Based on the above system, Figure 2 is a flowchart of a method for testing the data recording function of DSSAD provided by an embodiment of the present invention. This method can be executed in cooperation with each part of the above system, or can be executed by other electronic devices. As Figure 2 shown, this method specifically includes:

[0034] S110. Acquire real vehicle data and / or simulated driving data of an autonomous vehicle through the data acquisition tool.

[0035] This step performs data acquisition, including real vehicle data and simulated driving data. Among them, when acquiring real vehicle data, the data acquisition tool can be deployed inside the vehicle, and environmental data and vehicle data are acquired during the actual use of the vehicle, and the synchronization and integrity of the data are ensured. Among them, the environmental data includes video data collected by on-vehicle cameras, and the vehicle data includes vehicle state data (speed, acceleration, speed change, acceleration change, etc.), driver operation data (steering wheel angle, pedal angle, etc.), and autonomous driving system data (computing data and decision-making data of the autonomous driving system, etc.).

[0036] When collecting simulated driving data, in the simulated driving system, a vehicle model and various road scenarios can be established through third-party software, and the road scenarios can be displayed to the driver through a display device. The driver drives the car driving simulator in this road scenario, and the driving operation signals in the car driving simulator will be transmitted to the third-party software to control the movement of the vehicle model. At the same time, DSSAD can store relevant environmental data and vehicle data, such as Figure 3 as shown

[0037] The collected data is divided into three categories, namely normal driving data, collision risk data, and collision data. The first two categories of data can be collected from real vehicles by themselves or through the simulated driving system; the data of the third category of collision scenarios can be collected from real vehicles by cooperating with a third-party professional institution (such as conducting a collision experiment), but most of them are collected through the simulated driving system, as Figure 4 shown

[0038] In a specific embodiment, in combination with Figure 5 the specific steps of data collection include: in the data collection tool, perform the following operations periodically at a frequency higher than that of the CAN signal and the video signal (that is, the frequency at which the data collection tool periodically executes the following steps is higher than the frequency of the CAN signal and the video signal):

[0039] S1-1. Determine whether a CAN signal is received; if a CAN signal is received, store the current system time (i.e., timestamp), data type, CAN data length, and CAN data, and enter S1-2; if a CAN signal is not received, directly enter S1-2.

[0040] S1-2. Determine whether a video signal is received; if a video signal is received, store the current system time (i.e., timestamp), data type, video data length, and video data, and enter the loop of the next cycle; if a video signal is not received, directly enter the loop of the next cycle; until the data collection ends.

[0041] S120. Use the data playback tool to verify and annotate the collected data, where the annotation content includes collision events and collision risk events.

[0042] Among them, collision events and collision risk events are clearly defined in relevant standards and are also important data recorded by DSSAD. In this step, after the data collection is completed, the data playback tool is used to synchronously playback the video data and vehicle data, check the integrity and relevance of the video data and vehicle data before the next data injection, and record the triggering time of collision risk events (also called pre-collision events) and collision events, providing the true value for subsequent comparison and verification.

[0043] In a specific embodiment, in combination withFigure 6 , the specific steps of data playback include: within the data playback tool, periodically perform the following operations at a frequency higher than that of the CAN signal and the video signal (the same frequency as in the above data acquisition tool):

[0044] S2-1. Determine whether the current cycle meets the playback cycle of the CAN signal;

[0045] If the current cycle meets the playback cycle of the CAN signal (i.e., the time interval meets the CAN playback), send the CAN data of the current frame to the playback interface, and read the timestamp and data of the next frame of the CAN signal; continue to determine whether the current CAN data meets the trigger conditions for recording collision events or events with the risk of collision; if the trigger conditions are met, label the collision events or events with the risk of collision, and enter S2-2; if the trigger conditions are not met, enter S2-2;

[0046] If the current cycle does not meet the playback cycle of the CAN signal, directly enter S2-2.

[0047] S2-2. Determine whether the current cycle meets the playback cycle of the video signal;

[0048] If the current cycle meets the playback cycle of the video signal (i.e., the time interval meets the video playback), send the video data of the current frame to the playback interface, compare the current video image with the CAN signal to verify whether the scene is consistent; if it is consistent, read the timestamp and data of the next frame of the video signal and enter the loop of the next cycle; if it is inconsistent, record the inconsistent event for manual analysis;

[0049] If the current cycle does not meet the playback cycle of the video signal, directly enter the loop of the next cycle.

[0050] S130. Inject the collected data into the DSSAD to be tested so that the DSSAD records relevant data according to the standard requirements.

[0051] In this step, data injection is performed. The data recording program is started in advance in the DSSAD to be tested, and then the verified and labeled data is injected into the DSSAD through the data injection tool, and the test of the data recording program can be carried out, as Figure 7 shown.

[0052] In a specific embodiment, in combination with Figure 8, the steps of data injection include: the data injection tool determines the data parsing protocol corresponding to the vehicle model according to the collected data corresponding to the vehicle model; according to the data parsing protocol, the collected data is parsed into CAN data and video data; according to the CAN frequency, the CAN data is sent to the DSSAD through the CAN; according to the video data frequency, the video data is sent to the DSSAD through the Ethernet; the data recording program running inside the DSSAD will record the necessary data.

[0053] S140. Compare the data recorded by the DSSAD with the labeled data to verify the data recording function of the DSSAD.

[0054] This step performs data comparison and verification, comparing the data recorded by the DSSAD with the true value data labeled in S120 to verify whether the data recording function of the DSSAD meets the requirements.

[0055] In a specific embodiment, in combination with Figure 9 , the specific steps of data comparison and verification include:

[0056] S4-1. Perform data classification. Classify the data recorded in the DSSAD into trigger events and continuous records, where trigger events include collision events and events with a risk of collision.

[0057] S4-2. Compare whether the number of trigger events in the recorded data is the same as that in the labeled data; if they are the same, proceed to S4-3; if they are not the same, record the name of the inconsistent event and proceed to S4-3.

[0058] S4-3. Compare in detail whether the data inside each trigger event in the recorded data is the same as that in the labeled data; if they are the same, proceed to S4-4; if they are not the same, record the name of the inconsistent event and proceed to S4-4.

[0059] S4-4. Compare whether the quantity, duration, and file size of the continuous records (such as data during normal driving) in the recorded data are the same as those in the labeled data; if they are the same, proceed to S4-5; if they are not the same, record the name of the inconsistent event and proceed to S4-5.

[0060] S4-5. Randomly select a continuous record file (such as randomly select a video file) from the recorded data and the labeled data, and compare whether the data inside the file is the same. If they are the same, randomly select the next record file until the verification is completed; if they are not the same, record the name of the inconsistent event and randomly select the next record file until the verification is completed.

[0061] If the data in the above comparison and verification process are all consistent, it indicates that the data recording function of DSSAD meets the requirements. If there is any inconsistent event, manual analysis of this event is required.

[0062] Further, as Figure 4 shown, in the data collection stage, the data of collision events comes from third-party professional institutions or simulation driving systems. In practical applications, it is more common to use a simulation driving system to generate test cases for collision events. In order to make the simulated collision events closer to real collision events, this embodiment provides a method to simulate collision events through the real vehicle data of collision risk events, so as to improve the quality of the simulated data and the test effect of the whole method. In a specific embodiment, this process may include the following steps:

[0063] Step 1: Use a data collection tool to collect collision risk events in the real vehicle data.

[0064] Step 2: Generate vehicle state data and environmental data in the simulation driving system according to the specific real vehicle data within a set time period before the collision risk event, so that the driver can control the autonomous driving vehicle model through the automotive driving simulator and simulate the collision events under the specific real vehicle data multiple times.

[0065] Exemplarily, in the relevant standard, it is required that DSSAD records the data 15 s before the starting point of the collision risk event. Here, the set time period can be set to be greater than 15 s, such as 30 s, and the real vehicle data 30 s before the starting point of the collision risk event is recorded in advance. For the convenience of distinction and description, this 30-s real vehicle data is called specific real vehicle data.

[0066] Then, use the environmental data in this specific real vehicle data as the road scene for the vehicle to drive in the simulation driving system, use the vehicle state data at the beginning of a period of time in this specific real vehicle data as the initial vehicle state in the simulation driving system, and use some parameters of the autonomous driving system data (such as system setting parameters with relatively weak real-time variability, etc.) in the beginning period of time in this specific real vehicle data as some initial parameters of the autonomous driving system in the simulation driving system to generate a virtual driving scene. The driver can perform simulated driving through the automotive driving simulator, control the movement of the vehicle model, and simulate multiple collision events in this scene. Since the environmental data, vehicle state data, and some parameters of the autonomous driving system of these collision events are all generated according to the collision risk events in the real vehicle data, it can basically ensure the similarity between the simulated scene and the real vehicle collision scene, making the overall simulated collision events as close as possible to real collision events.

[0067] Step 3: Calculate the driver engagement in the specific real vehicle data and the driver engagement in the early stage of each simulation data; compare the driver engagement in the early stage of each simulation data with the driver engagement in the specific real vehicle data, and select the simulation data with the most similar driver engagement as the final simulated collision event.

[0068] As described above, among the multiple collision events simulated in Step 2, the vehicle state data, environmental data, and some data of the autonomous vehicle are all very close to the real scenario. The only factor with strong uncertainty is the driver's reaction. Different drivers, or the same driver in different simulations, have driving operation deviations, and the intentional collisions of the driver in the simulated collision events and the driver's attempt to avoid collisions in the real scenario will also lead to driving operation deviations. To minimize this deviation, in this embodiment, the driver engagement in the specific real vehicle data (representing the real engagement of people) and the driver engagement in the simulation data within the same time period are calculated, and the two are compared. The simulation data with the most similar driver engagement to the real vehicle data is selected as the simulated collision event closest to the actual situation.

[0069] Specifically, the driver engagement in this embodiment characterizes the impact of the driver's operation on the vehicle's driving, which is related to both the control that the driver subjectively expects to exert on the vehicle and the interaction between the driving operation and the autonomous driving system. Therefore, in the calculation of the driver engagement in this embodiment, the impact of the autonomous driving system on the vehicle's driving is also considered.

[0070] In a specific embodiment, a driver engagement prediction model based on a neural network can be pre-constructed, such as Figure 10As shown, the model includes two feature transformers. One is used to convert the driver operation data vector into an embedding vector with the same dimension as the vehicle state data vector, and the other is used to convert the autonomous driving system data vector into an embedding vector with the same dimension as the vehicle state data vector, and these two embedding vectors are linearly independent. Optionally, each feature transformer can adopt a fully connected layer or a convolutional network, etc.; the two feature transformers are independent of each other, and there is no necessary connection between the structures and parameters. Among them, the driver operation data vector can be generated in the following way: divide the specific real vehicle data into M time slices (M is a natural number), arrange N types of driver operation data in the same time slice as a column vector, and splice the column vectors of all time slices into a driver operation data vector (M×N dimension); or, perform principal component analysis on the column vectors of all time slices to obtain at least one N-dimensional principal component vector, and then perform weighted summation on these principal component vectors according to the corresponding eigenvalues to obtain an N-dimensional driver operation data vector. The autonomous driving system data vector and the vehicle state data vector are similar, as long as the generation methods of the three data vectors are the same. Usually, the dimensions of the three data vectors are different, and their respective dimensions are related to the number of parameters they collect.

[0071] Based on the above model structure, in this embodiment, the vehicle state data, driver operation data, and autonomous driving system data belonging to the same time period (the time period length is the set duration) are used as a sample, and the driver participation degree and the intelligent driving system participation degree in each sample are labeled. Among them, the intelligent driving system participation degree is used to characterize the impact of the intelligent driving system on vehicle driving. Optionally, the driver participation degree can be jointly labeled by drivers and autonomous driving system experts, taking into account both the control that the driver subjectively expects to exert on the vehicle and the objective impact on the vehicle state. The intelligent driving system participation degree is labeled by autonomous driving experts, and both participation degrees are values with an absolute value less than 1.

[0072] After the sample labeling is completed, the sample set is used to train the model, so that after the driver operation data vector and the autonomous driving system data vector belonging to the same sample are respectively input into the two feature transformers, and the two embedding vectors are weighted and summed using the labeled driver participation degree and autonomous driving system participation degree, the vehicle state data vector of the same sample can be obtained.

[0073] Specifically, the following loss function can be constructed L :

[0074]

[0075] Among them, represents the embedding vector obtained after the driver operation data vector passes through one of the feature transformers, Denotes the embedding vector obtained after the data vector of the autonomous driving system passes through another feature converter. Denotes the vehicle state data vector; 、 And All come from the same sample; Denotes that after arranging And As column vectors, the rank of the matrix formed by arranging the two column vectors, Denotes the absolute value; Respectively denote the labeled driver engagement and the engagement of the autonomous driving system; Are the preset weight coefficients respectively.

[0076] During the training process, the model parameters are updated by minimizing L . When L Is minimized, Approaches 0, , used to ensure that And Are non-linearly correlated. And , And Have the same dimension, so in this embodiment, Is expressed as , Of linear weighting to reflect the combined effect of the driver and the autonomous driving system on the vehicle state; L When minimized, Approaches 0, , which can exactly reflect the driver engagement and the engagement of the autonomous driving system in a linearly correlated manner in the vehicle state.

[0077] Furthermore, a loss function L 1 can also be constructed as follows:

[0078]

[0079] Wherein, , And Have the same meanings as in formula (1) and all come from the same sample; Respectively denote the engagement of the autonomous driving system labeled in other samples different from the same sample and the embedding vector obtained after the data vector of the autonomous driving system passes through the feature converter, Denotes the index of other samples, Denotes the sum over all other samples different from the same sample; Are the preset weight coefficients respectively. During the training process, the model parameters are updated by minimizing . When LWhen it is at its minimum, is very large, which is used to constrain that the three data vectors of different samples cannot satisfy the above linear relationship. The addition of the loss term can accelerate the convergence speed of the model and improve the training effect.

[0080] After training is completed, during the model usage stage, the driver operation data vector and the autonomous driving system data vector in the specific real vehicle data are respectively input into two trained feature transformers to obtain two new embedding vectors . Then, the vehicle data vector in the specific real vehicle data is expressed as a weighted sum form of the two new embedding vectors, where the weight of the new embedding vector corresponding to the driver operation data vector is the driver engagement in the specific real vehicle data.

[0081] Similarly, for any collision event simulated in step two, the driver operation data, autonomous driving system data, and vehicle state data within the same time period as the specific real vehicle data are extracted from the full-process simulation data, and the corresponding driver operation data vector , autonomous driving system data feature vector , and vehicle state data vector are respectively generated. and are respectively input into two trained feature transformers to obtain two new embedding vectors . Then, is expressed as in a weighted sum form , where the weight of is the driver engagement in this simulated collision event.

[0082] The same operation is performed on all simulated collision events to obtain the driver engagement of each simulated collision event respectively. Select a simulated collision event that is closest to the driver engagement in the specific real vehicle data as the final simulated collision event. This collision event is generated based on the collision risk event in the real vehicle data, follows the driving scenario (including environmental data, initial vehicle state, and some parameters of the autonomous driving system, etc.) in this actual event, and maintains the driver engagement most similar to this actual event, restoring the actual situation to the greatest extent in all aspects, making the subsequent simulated collision events closer to real collision events, improving the objective authenticity of the test data, and thus improving the accuracy and objectivity of the entire test method.

[0083] The generation method of the above-mentioned simulated collision events can be executed by a processing chip deployed inside the data acquisition tool, or by an electronic device outside the data acquisition tool, or by the cooperation of the data acquisition tool and an independent electronic device. This embodiment does not make specific restrictions.

[0084] In summary, this embodiment provides a method for testing the data recording function of DSSAD, establishing a complete testing process from test data acquisition, test data playback, test data injection to comparison and verification of recording results. Each stage is independently completed by a separate tool or device, without the need to bind DSSAD to a real vehicle or a driving simulation system in the traditional way. Instead, it can test the data recording situation of any time period and any vehicle at any time and place, improving the flexibility of the DSSAD data recording function test. In addition, through the data acquisition tool and the simulated driving system, it is possible to achieve full coverage of collision events, events with collision risks, and normal data; through data playback, preliminary verification and annotation of the acquired data can be performed to improve the quality of the test data; through multiple comparison and verification mechanisms, the recording situation of DSSAD for key data can be gradually determined, obtaining accurate test results, and providing accurate data entry points (such as inconsistent events) for the improvement of DSSAD, thereby continuously optimizing the data recording function of DSSAD.

[0085] As Figure 1 shown, the embodiment of the present invention also provides a system for testing the data recording function of DSSAD. Among them,

[0086] The data acquisition tool is used to acquire real vehicle data and / or simulated driving data of an autonomous vehicle, where the acquired data includes vehicle status data, driver operation data, autonomous driving system data, and environmental data.

[0087] The data playback tool is used to verify and annotate the acquired data, where the annotation content includes collision events and events with collision risks.

[0088] The data injection tool is used to inject the acquired data into the DSSAD to be tested, so that the DSSAD records relevant data according to the standard requirements.

[0089] The verification and comparison tool is used to compare the data recorded by the DSSAD with the annotated data to verify the data recording function of the DSSAD.

[0090] It should be noted that this embodiment and the above method embodiment are based on the same inventive concept, and the limitations in any of the above method embodiments are applicable to this embodiment and can achieve the same beneficial effects.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for testing the data recording function of DSSAD, characterized in that Including: Collecting real vehicle data and / or simulated driving data of an autonomous vehicle through a data collection tool, where the collected data includes vehicle status data, driver operation data, autonomous driving system data, and environmental data; Using a data playback tool to verify and annotate the collected data, where the annotation content includes collision events and collision-risk events; Injecting the collected data into the DSSAD to be tested so that the DSSAD records relevant data according to standard requirements; Comparing the data recorded by the DSSAD with the annotated data to verify the data recording function of the DSSAD; specifically, dividing the data recorded in the DSSAD into trigger events and continuous recordings, where the trigger events include collision events and collision-risk events; comparing whether the number of trigger events in the recorded data is the same as that in the annotated data, and recording the names of inconsistent events; comparing whether the data inside each trigger event in the recorded data is the same as that in the annotated data, and recording the names of inconsistent events; comparing whether the number, duration, and file size of the continuous recordings in the recorded data are the same as those in the annotated data, and recording the names of inconsistent events; randomly extracting continuous recording files from the recorded data and the annotated data, and comparing whether the data inside the files is the same, and recording the names of inconsistent events; where the names of each inconsistent event are used for final manual analysis.

2. The method according to claim 1, characterized in that, The collecting real vehicle data and / or simulated driving data of an autonomous vehicle through a data collection tool includes: Periodically performing the following operations in the data collection tool at a frequency higher than that of the CAN signal and the video signal: S1-1: In response to receiving a CAN signal, storing the timestamp, data type, CAN data length, and CAN data; S1-2: In response to receiving a video signal, storing the timestamp, data type, video data length, and video data.

3. The method according to claim 1, wherein The using a data playback tool to verify and annotate the collected data includes: Periodically performing the following operations in the data playback tool at a frequency higher than that of the CAN signal and the video signal: S2-1: In response to the current cycle satisfying the playback cycle of the CAN signal, sending the current CAN data and reading the timestamp and data of the next frame of the CAN signal; in the case where the CAN data meets the trigger conditions of a collision event or a collision-risk event, annotating the collision event or the collision-risk event; S2-2: In response to the current cycle satisfying the playback cycle of the video signal, sending the current video data and reading the timestamp and data of the next frame of the video signal; comparing the current video image with the CAN signal to verify whether the driving scenarios are consistent.

4. The method according to claim 1, wherein The injecting the collected data into the DSSAD to be tested so that the DSSAD records relevant data according to standard requirements includes: Determining a data parsing protocol according to the vehicle model from which the collected data is sourced; Parsing the collected data into CAN signals and environmental data according to the data parsing protocol; According to the frequencies of the CAN signal and environmental data, send the CAN signal and environmental data to the DSSAD to be tested, so that the DSSAD records relevant data according to the standard requirements.

5. The method according to claim 1, wherein Collect the in-vehicle data and / or simulated driving data of the autonomous vehicle through a data acquisition tool, including: Use the data acquisition tool deployed on the in-vehicle to collect the collision risk events in the in-vehicle data; Generate the vehicle state data and environmental data in the simulated driving system according to the specific in-vehicle data within a set time period before the collision risk event, so that the driver can control the autonomous vehicle model through the automotive driving simulator and simulate the collision events under the specific in-vehicle data multiple times; Compare the driver engagement in the early stage of each simulation data with the driver engagement in the specific in-vehicle data; Select the simulation data with the most similar driver engagement as the final simulated collision event.

6. The method according to claim 5, wherein The comparison of the driver engagement in the early stage of each simulation data with the driver engagement in the specific in-vehicle data includes: Construct a driver engagement prediction model, which includes two feature transformers. The two feature transformers are respectively used to transform the driver operation data vector and the autonomous driving system data vector into two embedding vectors with the same dimension as the vehicle state data vector, and the two embedding vectors are linearly independent; Construct samples using the vehicle state data, driver operation data, and autonomous driving system data belonging to the same time period, and train the model so that after the driver operation data vector and the autonomous driving system data vector belonging to the same sample are respectively input into the two feature transformers, the two embedding vectors are weighted and summed using the labeled driver engagement, and the vehicle state data vector of the same sample can be obtained; Respectively input the driver operation data vector and the autonomous driving system data vector in the specific in-vehicle data into the two trained feature transformers to obtain two new embedding vectors respectively; represent the vehicle state data vector in the specific in-vehicle data as the weighted sum of the two new embedding vectors, and use the weight of the new embedding vector corresponding to the driver operation data vector as the driver engagement.

7. The method according to claim 6, characterized in that, The construction of samples using the vehicle state data, driver operation data, and autonomous driving system data belonging to the same time period, and the training of the model so that after the driver operation data vector and the autonomous driving system data vector belonging to the same sample are respectively input into the two feature transformers, the two embedding vectors are weighted and summed using the labeled driver engagement, and the vehicle state data vector of the same sample can be obtained, includes: Construct the following loss function L : Among them, and respectively represent the embedding vectors obtained after the driver operation data vector and the autonomous driving system data vector pass through the feature converter, represents the vehicle state data vector, represents the rank of the matrix formed by arranging the column vectors and ; represents the absolute value; and respectively represent the marked driver engagement and autonomous driving system engagement; and are respectively preset weight coefficients; By L minimizing, the model parameters are updated.

8. The method according to claim 6, characterized in that, The construction of samples using the vehicle state data, driver operation data, and autonomous driving system data belonging to the same time period, and the training of the model so that after the driver operation data vector and the autonomous driving system data vector belonging to the same sample are respectively input into the two feature transformers, the two embedding vectors are weighted and summed using the labeled driver engagement, and the vehicle state data vector of the same sample can be obtained, includes: Construct the following loss function L : Among them, and respectively represent the embedding vectors obtained after the driver operation data vector and the autonomous driving system data vector from the same sample pass through the feature converter, represents the vehicle state data vector of the same sample, represents the rank of the matrix formed by arranging the column vectors and ; represents the absolute value; and respectively represent the labeled driver engagement and the autonomous driving system engagement; and respectively represent the labeled autonomous driving system engagement in other samples different from the same sample, and the embedding vector obtained after the autonomous driving system data vector passes through the feature converter; represents the index of other samples; , and are preset weight coefficients respectively; By L minimizing, the model parameters are updated.

9. The data recording function test system of DSSAD, characterized in that, Including: A data acquisition tool for collecting real vehicle data and / or simulated driving data of an autonomous vehicle, wherein the collected data includes vehicle status data, driver operation data, autonomous driving system data, and environmental data; A data playback tool for verifying and annotating the collected data, wherein the annotation content includes collision events and collision risk events; A data injection tool for injecting the collected data into the DSSAD to be tested, so that the DSSAD records relevant data according to the standard requirements; A verification and comparison tool for comparing the data recorded by the DSSAD with the annotated data to verify the data recording function of the DSSAD; specifically, dividing the data recorded in the DSSAD into trigger events and continuous records, wherein the trigger events include collision events and collision risk events; comparing whether the number of trigger events in the recorded data is consistent with the annotated data, and recording the names of inconsistent events; comparing whether the data inside each trigger event in the recorded data is consistent with the annotated data, and recording the names of inconsistent events; comparing whether the quantity, duration, and file size of the continuous records in the recorded data are consistent with the annotated data, and recording the names of inconsistent events; randomly extracting continuous record files from the recorded data and the annotated data, and comparing whether the data inside the files is consistent, and recording the names of inconsistent events; wherein the names of each inconsistent event are used for final manual analysis.

Images