Self-driving automobile test scene generation method and device and storage medium

Abstract

The invention discloses an automatic driving automobile test scene generation method and device and a storage medium, and the method comprises the steps: reconstructing accident scenes, structuring accident texts and video information, and carrying out the classification extraction of static elements and dynamic elements in each traffic accident scene; the static elements and the dynamic elements are sampled according to a preset step length, a scene is extracted into a plurality of scene sections, and the scene sections comprise vehicle track data under corresponding timestamps; and inputting the scene section into a trained generative adversarial network model to obtain a test scene with a time axis label, so that a brand new test scene with authenticity and diversity is generated through the generative adversarial network. And then risk evaluation is carried out on the generated scene through a method of combining an importance sampling method and a sensitivity analysis algorithm, feedback is carried out on the generated model, and a scene generation technology is optimized.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving, and specifically to a method, device and storage medium for generating a test scenario for an autonomous driving vehicle. Background Art

[0002] Autonomous vehicles are expected to revolutionize the future of transportation by significantly improving traffic safety. However, for autonomous vehicles to be widely adopted, they must be thoroughly tested to verify their safety. Previous studies have highlighted that more than 10 billion miles of public road testing are required to verify the safety performance of autonomous vehicles, which is a major challenge for manufacturers within the industry. In addition, public road scenarios often lack the necessary complexity and risk levels, and it is critical to evaluate the capabilities of autonomous vehicles in high-risk situations. Summary of the invention

[0003] In response to the above technical problems, the present application provides a method, device and storage medium for generating test scenarios for autonomous driving vehicles to solve the problem of sparse dangerous scenes in the test scenario library, thereby improving test efficiency and enhancing test effectiveness.

[0004] The present application provides a method for generating a test scenario for an autonomous driving vehicle, the method comprising:

[0005] Step S1, reconstructing the accident scene, structuring the accident text and video information, and classifying and extracting the static elements and dynamic elements in each traffic accident scene;

[0006] Step S2, sampling the static elements and the dynamic elements according to a preset step length, extracting the scene into a plurality of scene sections, wherein the scene sections include vehicle trajectory data at a corresponding timestamp;

[0007] Step S3: input the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label.

[0008] Preferably, the static elements include: roadside facilities, road type, vehicle type, vehicle speed, relative distance, braking acceleration, and weather conditions.

[0009] Preferably, the dynamic elements are the state of the vehicle at each time step, including: the longitudinal coordinate, the transverse coordinate, the speed, acceleration, direction, steering, and relative position difference at each coordinate within each step.

[0010] Preferably, the static elements and the dynamic elements are sampled according to a preset step length, and the scene is extracted into a plurality of scene sections, wherein the scene sections include vehicle trajectory data at a corresponding timestamp, including:

[0011] The static elements and the dynamic elements are sampled at a preset time sampling interval to generate a number of independent scene sections, wherein the data of each scene section needs to include static information, dynamic information, and the relative position and behavior characteristics of each participating vehicle.

[0012] Preferably, the generative adversarial network includes a generative network and an adversarial network; the step of inputting the scene section into a trained generative adversarial network model to obtain a test scene with a timeline label includes:

[0013] Inputting the scene section into a trained generative adversarial network model, and obtaining vehicle dynamic information based on the vehicle trajectory data through the generative network; wherein the vehicle dynamic information includes the vehicle acceleration change rate and the steering angle change rate;

[0014] The vehicle dynamic information and the vehicle trajectory data are input into an adversarial network to determine a discrimination result, and a test scene with a timeline label is determined based on the discrimination result.

[0015] Preferably, before inputting the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label, the method includes:

[0016] Inputting the scene section into an initial generative adversarial network model, and obtaining a corresponding training scene section based on the vehicle trajectory data at a corresponding timestamp through the generative network;

[0017] Inputting the training scene section and the scene section into the adversarial network to determine a discrimination result;

[0018] Based on the discrimination result, the generative adversarial network is individually iterated alternately until the set loss function meets the convergence condition, thereby obtaining the trained generative adversarial network model.

[0019] Preferably, the step of inputting the scene section into a trained generative adversarial network model to obtain a test scene with a timeline label comprises:

[0020] The scene section is input into the trained generative adversarial network model to obtain vehicle trajectory data with time axis labels, and the test scene is obtained based on the vehicle trajectory data at the corresponding time.

[0021] The embodiment of the present invention also provides a device for generating a test scenario for an autonomous driving vehicle, comprising:

[0022] The acquisition module is used to reconstruct the accident scene, structure the accident text and video information, and classify and extract the static and dynamic elements in each traffic accident scene;

[0023] A sampling module, used for sampling the static elements and the dynamic elements according to a preset step length, and extracting the scene into a plurality of scene sections, wherein the scene sections include vehicle trajectory data at a corresponding timestamp;

[0024] The output module is used to input the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label.

[0025] The present application also provides a computer device, comprising: a memory and a processor, wherein the memory stores computer program instructions for execution on the processor, and when the processor executes the computer program instructions, the method for generating a test scenario for an autonomous driving vehicle as described above is implemented.

[0026] The present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the method for generating a test scenario for an autonomous driving vehicle are implemented as described above.

[0027] As described above, the accident scene is reconstructed, and the accident text and video information are structured, and the static elements and dynamic elements in each traffic accident scene are classified and extracted; the static elements and the dynamic elements are sampled according to a preset step size, and the scene is extracted into a number of scene sections, and the scene section includes vehicle trajectory data at a corresponding timestamp; the scene section is input into a trained generative adversarial network model to obtain a test scene with a timeline label. In this way, a new test scene with both authenticity and diversity is generated through a generative adversarial network, and then the generated scene is evaluated for danger by combining an importance sampling method with a sensitivity analysis algorithm, and the generation model is fed back and the scene generation technology is optimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application. In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor.

[0029] Figure 1 A schematic diagram of a flow chart of a method for generating a test scenario for an autonomous driving vehicle provided in one embodiment of the present invention;

[0030] Figure 2 A schematic diagram of the structure of an autonomous driving vehicle test scenario generating device provided in one embodiment of the present invention.

[0031] The realization of the purpose, functional features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. The above-mentioned drawings have shown clear embodiments of this application, which will be described in more detail later. These drawings and textual descriptions are not intended to limit the scope of the concept of this application in any way, but to illustrate the concept of this application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0032] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0033] Before further describing the present invention in detail, the nouns and terms involved in the embodiments of the present invention are explained. The nouns and terms involved in the embodiments of the present invention are subject to the following explanations.

[0034] See also Figure 1 This embodiment provides a method for generating a test scenario for an autonomous driving vehicle, the method comprising:

[0035] Step S1, reconstructing the accident scene, structuring the accident text and video information, and classifying and extracting the static elements and dynamic elements in each traffic accident scene;

[0036] Here, by deconstructing the accident and extracting information, we can accurately obtain the key elements required for testing, including static environmental information such as road layout and traffic signals, as well as dynamic parameter distributions such as speed, acceleration, and position of various traffic participants. This refined extraction method makes the test scenario construction more flexible and can quickly adjust the scenario elements according to specific test requirements, thereby improving the pertinence and efficiency of the test.

[0037] In step S1, the static environmental information in the accident includes: roadside facilities, road type, vehicle type, vehicle speed, relative distance, braking acceleration, weather conditions, etc. Dynamic parameters include the status data of traffic participants within every 0.1s, such as longitudinal and lateral position, speed, acceleration, driving direction, steering angle, relative position difference and other information.

[0038] Step S2, sampling the static elements and the dynamic elements according to a preset step length, extracting the scene into a plurality of scene sections, wherein the scene sections include vehicle trajectory data at a corresponding timestamp;

[0039] Here, by sampling the scene information, the complex scene can be decomposed into multiple easy-to-process sections, each of which can be used as an independent test unit. This segmented processing method can reduce the complexity of the entire scene, making subsequent generation and calculation more efficient, while ensuring the full expression of scene diversity and improving the flexibility of scene generation.

[0040] The scene was sampled at a sampling interval of 0.1 seconds to generate scene sections. We obtained parameters such as the vehicle's speed, acceleration, and steering angle within 5 seconds before the collision. Throughout the vehicle's motion, we collected detailed data and segmented the data into different scenes at intervals of 0.1 seconds. Each scene records the position coordinates, speed, acceleration, direction, and other relevant details of the collision parties at a specific timestamp. Through analysis, we can fully understand the vehicle's motion state and provide basic data for subsequent research and analysis. When extracting vehicle dynamic parameters, speed and other related information usually come from videos that capture the entire process before the collision. The time difference method is usually used to calculate the vehicle speed in each collision. This method is a video processing algorithm that detects moving objects by analyzing the differences in pixel values ​​of each video frame and calculates the dynamic characteristics of these objects based on the deviation. The specific calculation formulas are shown in Equations (1) and (2).

[0041]

[0042]

[0043] Among them, Δl represents the distance between two reference points in the video, n represents the number of frames that the vehicle passes between the two reference points, t is the time interval between each frame in the monitoring video, and fps is the frame rate of the monitoring video. We first locate the front and rear positions of the vehicle, then judge the distance based on auxiliary information such as road width and vehicle damage information, and combine the video frame rate to obtain the vehicle's speed information.

[0044] Step S3, inputting the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label;

[0045] The provided accident scenarios are learned through generative adversarial networks, and a large number of test scenarios are generated. Specifically: This method builds the generator and discriminator models of the generative adversarial network based on the GAN network, and uses LeakyReLU and Dropout in the network layer to prevent the gradient from disappearing. In the process of mutual game between the generator and the discriminator, the dynamic equilibrium of the network is achieved.

[0046] The Wasserstein distance is introduced into the network to replace the loss function of the original GAN ​​network. The EM distance is continuously estimated by training the discriminator to solve the problem of model training difficulties caused by the original GAN ​​model's sensitivity to hyperparameters. The focus is on minimizing the Wasserstein distance rather than optimizing the Jensen-Shannon divergence. The optimization objective of the discriminator can be expressed as formula (3):

[0047]

[0048] The goal of formula (3) is to minimize the distance between the distribution of the generator G and the true distribution. In particular, when training WGAN, there is no need to carefully balance the training of the discriminator and the generator, nor is there a need to complexly design the network architecture. Compared with other GAN models, the enhanced WGAN converges faster, has a more robust training process, and generates higher quality samples. An important practical advantage of WGAN is the ability to continuously estimate the EM distance by training the discriminator. In this study, the extracted scenes are used as input to WGAN. Each scene includes a time node, a collision type, and parameters such as the x-coordinate, y-coordinate, speed, direction, and acceleration of the two vehicles. After 600 epochs of training, new slices with timeline labels are generated.

[0049] Based on this model, the present invention utilizes the vehicle trajectory data of accident scenes in the existing accident database, learns some intuitive parameters, generates new trajectory data, and then processes the generated trajectory according to the vehicle dynamics model. After Kalman filtering optimization, more realistic trajectory data is generated.

[0050] The above-mentioned embodiment of the present invention generates a new test scenario that is both realistic and diverse through a generative adversarial network, and then evaluates the risk of the generated scenario through a method that combines an importance sampling method with a sensitivity analysis algorithm, provides feedback to the generation model, and optimizes the scenario generation technology.

[0051] In one implementation, the static elements include: roadside facilities, road type, vehicle type, vehicle speed, relative distance, braking acceleration, and weather conditions.

[0052] In one embodiment, the dynamic elements are the state of the vehicle at each time step, including: the longitudinal coordinate, the lateral coordinate, the speed, acceleration, direction, steering, and relative position difference at each coordinate within each step.

[0053] In one embodiment, the static elements and the dynamic elements are sampled according to a preset step length, and the scene is extracted into a plurality of scene sections, wherein the scene sections include vehicle trajectory data at a corresponding timestamp, including:

[0054] The static elements and the dynamic elements are sampled at a preset time sampling interval to generate a number of independent scene sections, wherein the data of each scene section needs to include static information, dynamic information, and the relative position and behavior characteristics of each participating vehicle.

[0055] In one embodiment, the generative adversarial network includes a generative network and an adversarial network; the step of inputting the scene section into a trained generative adversarial network model to obtain a test scene with a timeline label includes:

[0056] Inputting the scene section into a trained generative adversarial network model, and obtaining vehicle dynamic information based on the vehicle trajectory data through the generative network; wherein the vehicle dynamic information includes the vehicle acceleration change rate and the steering angle change rate;

[0057] The vehicle dynamic information and the vehicle trajectory data are input into an adversarial network to determine a discrimination result, and a test scene with a timeline label is determined based on the discrimination result.

[0058] In one embodiment, before inputting the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label, the process includes:

[0059] Inputting the scene section into an initial generative adversarial network model, and obtaining a corresponding training scene section based on the vehicle trajectory data at a corresponding timestamp through the generative network;

[0060] Inputting the training scene section and the scene section into the adversarial network to determine a discrimination result;

[0061] Based on the discrimination result, the generative adversarial network is individually iterated alternately until the set loss function meets the convergence condition, thereby obtaining the trained generative adversarial network model.

[0062] Here, based on this model, the present invention uses the vehicle trajectory data of the accident scene in the existing accident database to learn some intuitive parameters, generate new trajectory data, and then process the generated trajectory according to the vehicle dynamics model. After Kalman filter optimization, more realistic trajectory data is generated. The specific generation process is as follows: (It should be noted that the generator below is the generation network, and the discriminator is the adversarial network)

[0063] Step 1: Build the generator and discriminator networks

[0064] class Discriminator(nn.Module):

[0065] def __init__(self):

[0066] super().__init__()

[0067] self.model=nn.Sequential(

[0068] nn.Linear(12,512),

[0069] nn.LeakyReLU(0.2),

[0070] nn.Dropout(0.2),

[0071] nn.Linear(512,256),

[0072] nn.LeakyReLU(0.2),

[0073] nn.Dropout(0.2),

[0074] nn.Linear(256,1), )

[0076] class Generator(nn.Module):

[0077] def__init__(self):

[0078] super().__init__()

[0079] self.model=nn.Sequential(

[0080] nn.Linear(12,128),

[0081] nn.LeakyReLU(0.2),

[0082] nn.Linear(128,256),

[0083] nn.LeakyReLU(0.2),

[0084] nn.Linear(256,512),

[0085] nn.LeakyReLU(0.2),

[0086] nn.Linear(512,1024),

[0087] nn.LeakyReLU(0.2),

[0088] nn.Linear(1024,12), )

[0090] Step 2: Edit the Optimizer

[0091] optimizer_G=torch.optim.RMSprop(generator.parameters(),lr=0.0001)

[0092] optimizer_D=torch.optim.RMSprop(discriminator.parameters(),lr=0.0001)

[0093] Step 3: Train the Discriminator

[0094] optimizer_D.zero_grad()

[0095] loss_D=-torch.mean(discriminator(real_samples))+torch.mean(discriminator(fak e_samples))

[0096] loss_D.backward()

[0097] optimizer_D.step()

[0098] Step 4: Train the Generator

[0099] optimizer_G.zero_grad()

[0100] generated_samples=generator(latent_space_samples)

[0101] loss_G=-torch.mean(discriminator(generated_samples))

[0102] loss_G.backward()

[0103] optimizer_G.step()

[0104] Step 5: Get the generated scene. When the Loss value converges, extract the scene data generated by the generator.

[0105]

[0106] After using WGAN to generate a new scene, two parameters are needed to calculate the vehicle dynamic information of the complete time series, namely the acceleration rate of change and the steering angle rate of change. The initial scene consists of the initial states of static elements and dynamic elements. Static elements include weather, lighting and roads, and dynamic elements include the initial speed of the vehicle. Such a scene constitutes the initial state of the entire scene. After that, we will provide some variable values ​​of dynamic elements and calculate the update state at a node every 0.1s, so that the scene can continuously calculate the time series within 5s. Continue down and finally get a complete scene.

[0107] To ensure the accuracy of the calculated trajectory, we sampled two basic change rates from the data set obtained from the collision reconstruction. Sampling based on the cumulative distribution function (CDF) is a method for generating random samples according to a specified probability distribution. Through sampling, we can generate random variables that conform to a specific distribution. CDF represents the integral of the probability density function and fully describes the probability distribution of the real random variable X. The cumulative distribution function is defined as formula (5):

[0108] F X (x) = P (X ≤ x) (5)

[0109] CDF has the following characteristics: (1) CDF is a monotonically increasing function along the X-axis. (2) The image generated by CDF is smoother and has less noise. (3) CDF does not involve external concepts such as bandwidth, ensuring that no data information is lost. (4) For a given data set, CDF is unique.

[0110] In one embodiment, the step of inputting the scene section into a trained generative adversarial network model to obtain a test scene with a timeline label includes:

[0111] The scene section is input into the trained generative adversarial network model to obtain vehicle trajectory data with time axis labels, and the test scene is obtained based on the vehicle trajectory data at the corresponding time.

[0112] In one embodiment, the obtained scenario is input into the virtual simulation test software to test the autonomous driving system under test. We use the scenario editing simulator and the autonomous driving system joint simulation platform to evaluate the authenticity and effectiveness of the generated scenario. The detailed steps are as follows:

[0113] 1. In the simulation platform, the simulation test is carried out by importing the scene map created by the 3D modeling software Unity 3D and using the vehicle model provided by the SVL simulator.

[0114] 2. Select a test scenario at an intersection and program the vehicle’s initial position on a straight road 50 meters from the center of the intersection. Use time series data from WGAN, including the vehicle’s initial position and speed.

[0115] 3. Then, the initial position, movement direction and speed of the autonomous driving vehicle are configured to 0.

[0116] 4. Start the experiment and record the driving behavior and driving conditions of the vehicle controlled by the autonomous driving system in the current scenario.

[0117] In this simulation framework, the autonomous vehicle acts as the host vehicle, while the target vehicle follows a pre-set trajectory. The virtual vehicle motion in a specific scenario is simulated using the integrated collision avoidance and path planning strategies within the platform. Various safety indicators are applied to create hazardous conditions and capture the trajectory and status of all participating vehicles. By performing virtual simulations, the driving behavior of the autonomous vehicle in a specific scenario can be evaluated.

[0118] See also Figure 2 , an embodiment of the present invention further provides a device for generating a test scenario for an autonomous driving vehicle, comprising:

[0119] The acquisition module 21 is used to reconstruct the accident scene, structure the accident text and video information, and classify and extract the static elements and dynamic elements in each traffic accident scene;

[0120] A sampling module 22, for sampling the static elements and the dynamic elements according to a preset step length, and extracting the scene into a plurality of scene sections, wherein the scene sections include vehicle trajectory data at a corresponding timestamp;

[0121] The output module 23 is used to input the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label.

[0122] The present application also provides a computer device, comprising: a memory and a processor, wherein the memory stores computer program instructions for execution on the processor, and when the processor executes the computer program instructions, the method for generating a test scenario for an autonomous driving vehicle as described above is implemented.

[0123] The present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the method for generating a test scenario for an autonomous driving vehicle are implemented as described above.

[0124] By integrating the functions of these modules, the constructed device can not only achieve efficient and accurate generation of autonomous vehicle test scenarios, but also improve the accuracy of robustness assessment of autonomous vehicles, which helps to improve the safety and reliability of autonomous vehicles and promote the development of their practical applications.

[0125] The above examples are only reference examples. In order to avoid redundancy, they are not listed one by one here. In actual development or application, they can be flexibly combined according to actual needs, but any combination belongs to the technical solution of this application and is covered within the scope of protection of this application.

[0126] An embodiment of the present application also provides a computer device, including a processor and a memory, wherein the memory stores computer program instructions for execution on the processor, and when the processor executes the computer program instructions, the method for generating a test scenario for an autonomous driving vehicle as described above is implemented.

[0127] An embodiment of the present application also provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the method for generating a test scenario for an autonomous driving vehicle as described above.

[0128] It is understood that the above scenarios are only examples and do not constitute a limitation on the application scenarios of the technical solutions provided in the embodiments of the present application. The technical solutions of the present application can also be applied to other scenarios. For example, it is known to those skilled in the art that with the evolution of the system architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0129] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0130] The steps in the method of the embodiment of the present application can be adjusted in order, combined and deleted according to actual needs.

[0131] The units in the device of the embodiment of the present application can be merged, divided and deleted according to actual needs.

[0132] In the present application, the same or similar terminology concepts, technical solutions and / or application scenario descriptions are generally described in detail only the first time they appear. When they appear again later, they are generally not repeated for the sake of brevity. When understanding the technical solutions and other contents of the present application, for the same or similar terminology concepts, technical solutions and / or application scenario descriptions that are not described in detail later, reference can be made to the previous related detailed descriptions.

[0133] In the present application, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0134] The various technical features of the technical solution of the present application can be arbitrarily combined. In order to make the description concise, not all possible combinations of the various technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of the present application.

[0135] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, a controlled terminal, or a network device, etc.) to execute the method of each embodiment of the present application.

[0136] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loading and executing computer program instructions on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. Computer instructions can be stored in a storage medium, or transmitted from one storage medium to another storage medium. For example, computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) mode. The storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integration. Available media can be magnetic media, (e.g., floppy disk, storage disk, tape), optical media (e.g., DVD), or semiconductor media (e.g., solid-state storage disk Solid State Disk (SSD)), etc.

[0137] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for generating a test scenario for an autonomous driving vehicle, characterized in that: The method comprises: Step S1, reconstructing the accident scene, structuring the accident text and video information, and classifying and extracting the static elements and dynamic elements in each traffic accident scene; Step S2, sampling the static elements and the dynamic elements according to a preset step length, extracting the scene into a plurality of scene sections, wherein the scene sections include vehicle trajectory data at a corresponding timestamp; Step S3: input the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label.

2. The method for generating a test scenario for an autonomous driving vehicle according to claim 1, wherein: The static elements include: roadside facilities, road type, vehicle type, vehicle speed, relative distance, braking acceleration, and weather conditions.

3. The method for generating a test scenario for an autonomous driving vehicle according to claim 1, wherein: The dynamic elements are the state of the vehicle at each time step, including: the longitudinal coordinate, the lateral coordinate, the speed, acceleration, direction, steering, and relative position difference at each coordinate within each step.

4. The method for generating a test scenario for an autonomous driving vehicle according to claim 1, wherein: The static elements and the dynamic elements are sampled according to a preset step length, and the scene is extracted into a plurality of scene sections, wherein the scene sections include vehicle trajectory data at a corresponding timestamp, including: The static elements and the dynamic elements are sampled at a preset time sampling interval to generate a number of independent scene sections, wherein the data of each scene section needs to include static information, dynamic information, and the relative position and behavior characteristics of each participating vehicle.

5. The method for generating a test scenario for an autonomous driving vehicle according to claim 1, wherein: The generative adversarial network includes a generative network and an adversarial network; the step of inputting the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label includes: Inputting the scene section into a trained generative adversarial network model, and obtaining vehicle dynamic information based on the vehicle trajectory data through the generative network; wherein the vehicle dynamic information includes the vehicle acceleration change rate and the steering angle change rate; The vehicle dynamic information and the vehicle trajectory data are input into an adversarial network to determine a discrimination result, and a test scene with a timeline label is determined based on the discrimination result.

6. The method for generating a test scenario for an autonomous driving vehicle according to claim 1, wherein: Before inputting the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label, the method includes: Inputting the scene section into an initial generative adversarial network model, and obtaining a corresponding training scene section based on the vehicle trajectory data at a corresponding timestamp through the generative network; Inputting the training scene section and the scene section into the adversarial network to determine a discrimination result; Based on the discrimination result, the generative adversarial network is individually iterated alternately until the set loss function meets the convergence condition, thereby obtaining the trained generative adversarial network model.

7. The method for generating a test scenario for an autonomous driving vehicle according to claim 1, wherein: The step of inputting the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label includes: The scene section is input into the trained generative adversarial network model to obtain vehicle trajectory data with time axis labels, and the test scene is obtained based on the vehicle trajectory data at the corresponding time.

8. An automatic driving vehicle test scenario generating device, characterized in that: include: The acquisition module is used to reconstruct the accident scene, structure the accident text and video information, and classify and extract the static and dynamic elements in each traffic accident scene; A sampling module, used for sampling the static elements and the dynamic elements according to a preset step length, and extracting the scene into a plurality of scene sections, wherein the scene sections include vehicle trajectory data at a corresponding timestamp; The output module is used to input the scene section into the trained generative adversarial network model to obtain a test scene with a timeline label.

9. A computer device, characterized in that: It includes a processor and a memory, wherein the memory stores computer program instructions for execution on the processor, and when the processor executes the computer program instructions, the method for generating a test scenario for an autonomous driving vehicle as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: It stores computer instructions, which, when executed by a processor, implement the method for generating a test scenario for an autonomous driving vehicle as described in any one of claims 1 to 7.

Images