Method and device for automatically generating expected function security scene and electronic equipment

Abstract

The invention discloses an expected function safety scene automatic generation method and device and electronic equipment. The method comprises the steps that sensor data, camera data and vehicle driving track data are collected to be used for training a first condition instance generative adversarial network (IC-GAN); generating an expected functional security scene by using the first condition instance generative adversarial network, wherein the expected functional security scene comprises an unknown and unsafe scene; classifying the expected function security scene based on a machine learning network model to obtain a classification result, and optimizing parameters of the first condition instance generative adversarial network according to the classification result to form a second condition instance generative adversarial network, and generating a new expected function security scene by using the second condition instance generative adversarial network. According to the invention, through the condition instance generative adversarial network, the requirement for generating an unknown scene can be met; through scene classification and a feedback optimization mechanism, the test and verification efficiency of the automatic driving system can be improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a method, device and electronic equipment for automatically generating expected functional safety scenarios. Background Art

[0002] With the rapid development of science and technology, autonomous driving technology has gradually become a hot topic of research. In order to ensure the safety of autonomous driving systems in actual operation, a lot of testing and verification is required. However, actual road testing has many limitations, such as high cost and difficult to control the test environment. Therefore, simulation testing has become an important alternative.

[0003] In simulation testing, the generation of expected functional safety scenarios is a key step. These scenarios need to be simulated under various conditions and environments to cover as many possible situations in actual driving as possible, especially unknown and unsafe scenarios in expected functional safety. However, existing simulation methods often have difficulty in generating high-quality, realistic scenario data, resulting in inaccurate test results and even missing some potential safety risks.

[0004] With the rapid development of autonomous driving technology, how to ensure the safety of autonomous driving systems has become an important research issue. As an important aspect of autonomous driving safety, safety of intended functions (SOTIF) aims to identify and manage hazards and risks caused by insufficient functions or reasonably foreseeable human misuse. In order to effectively conduct SOTIF assessments, a large number of meaningful intended functional safety scenarios need to be generated.

[0005] Existing scenario generation methods often fail to fully consider various complex traffic environments and driving conditions, resulting in a lack of diversity and authenticity in the generated scenarios. Especially after the introduction of autonomous driving technology into existing traffic flows, traffic scenarios have become more complex and changeable, and the demand for unknown scenarios of expected functional safety has increased, which has also put higher requirements on scenario generation. Traditional scenario generation methods are often limited by the amount of data and scenario diversity, and it is difficult to cover all possible driving scenarios.

[0006] Traditional simulation methods often fail to truly simulate the complexity and uncertainty of the actual traffic environment when generating expected functional safety scenarios, especially when there are few traffic scenarios involving self-driving cars. This results in low quality of generated scenario data, limited coverage, insufficient consideration of long-tail scenarios, and a large gap with actual scenarios. Summary of the invention

[0007] In response to the above technical problems, the present invention provides a method, device and electronic device for automatically generating expected functional safety scenarios, which can meet the requirements of expected functional safety for generating unknown scenarios.

[0008] According to a first aspect of the present invention, a method for automatically generating expected functional safety scenarios is provided, comprising: collecting sensor data, camera data, and vehicle driving trajectory data for training a first conditional instance generation adversarial network (IC-GAN); using the first conditional instance generation adversarial network to generate expected functional safety scenarios, wherein the expected functional safety scenarios include unknown and unsafe scenarios; classifying the expected functional safety scenarios based on a machine learning network model to obtain classification results, and optimizing the parameters of the first conditional instance generation adversarial network according to the above classification results to form a second conditional instance generation adversarial network, and using the second conditional instance generation adversarial network to generate new expected functional safety scenarios.

[0009] In an optional implementation, the collecting sensor data, camera data, and vehicle driving trajectory data for training the first conditional instance generative adversarial network includes: The first conditional instance generative adversarial network is trained based on random noise and conditional information, wherein the conditional information includes a security risk condition, and the security risk condition is used to guide the first conditional instance generative adversarial network to generate an unknown and unsafe scenario.

[0010] In an optional implementation, the classifying the expected functional safety scenario based on the machine learning network model to obtain a classification result includes: The expected functional safety scenarios are classified based on a deep belief network model to obtain classification results; the classification results include performance limitations, complex environments, and human-computer interactions.

[0011] In an optional implementation, the classifying the expected functional safety scenario based on the machine learning network model to obtain a classification result, and optimizing the parameters of the first conditional instance generation adversarial network according to the classification result to form a second conditional instance generation adversarial network, including: Based on the machine learning network model, the expected functional safety scenarios are classified to obtain scenario types and the number of scenario types. If the data of the scenario types do not meet the preset conditions, the parameters of the first conditional instance generation adversarial network are optimized to form a second conditional instance generation adversarial network.

[0012] In an optional implementation, if the data of the scenario type does not meet a preset condition, optimizing the parameters of the first conditional instance generative adversarial network to form a second conditional instance generative adversarial network includes: If the data of the scenario type does not meet the preset conditions, an early stopping method or a regularization method is used to optimize the parameters of the first conditional instance generation adversarial network to form a second conditional instance generation adversarial network.

[0013] In an optional embodiment, the method for automatically generating expected functional safety scenarios further includes: evaluating the practicality and safety of the expected functional safety scenarios by comparing the types and frequencies of safety risks in real scenarios with those in the expected functional safety scenarios, and adjusting the parameters of the second conditional instance generation adversarial network based on the evaluation results of the practicality and safety.

[0014] According to a second aspect of the present invention, there is provided an apparatus for automatically generating expected functional safety scenarios, comprising: a training module for collecting sensor data, camera data, and vehicle driving trajectory data for training a first conditional instance generative adversarial network (IC-GAN); a first generation module for generating expected functional safety scenarios using the first conditional instance generative adversarial network, wherein the expected functional safety scenarios include unknown and unsafe scenarios; a second generation module for classifying the expected functional safety scenarios based on a machine learning network model to obtain classification results, and optimizing the parameters of the first conditional instance generative adversarial network according to the above classification results to form a second conditional instance generative adversarial network, and generating new expected functional safety scenarios using the second conditional instance generative adversarial network.

[0015] In an optional implementation, the first generation module is further used to classify the expected functional safety scenarios based on a deep belief network model to obtain classification results; the classification results include performance limitations, complex environments, and human-computer interactions.

[0016] According to a third aspect of the present invention, there is provided an electronic device, comprising: At least one processor; and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the method for automatically generating expected functional safety scenarios as described in the first aspect of an embodiment of the present invention.

[0017] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a computer, the method for automatically generating expected functional safety scenarios as described in the first aspect of the embodiment of the present invention is executed.

[0018] The present invention can generate realistic scenes that are combined with objects or scenes that have not been seen, unknown, or sudden in the training data set through conditional instance generative adversarial networks, thereby meeting the requirements of expected functional safety for generating unknown scenes; through scene classification and feedback optimization mechanism, the generated scene data is more targeted and practical, which helps to improve the testing and verification efficiency of the autonomous driving system and ensure its performance and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 The figure is a flow chart of a method for automatically generating an expected functional safety scenario in an embodiment of the present invention.

[0021] Figure 2 Schematic diagram of a module of a device for automatically generating expected functional safety scenarios in an embodiment of the present invention.

[0022] Figure 3 Schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0024] It should be understood that the terms "first", "second", etc. in the claims, specifications and drawings of the present invention are used to distinguish different objects rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections.

[0025] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the claims, the singular forms of "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in the specification of the present invention and the claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0026] The GAN model used in existing technologies has been proven to have superior performance in image generation, but it is limited by the scope of its training data set. It is not widely used in the generation of expected functional safety scenarios for autonomous driving, and lacks targeted optimization and insufficient consideration of long-tail scenarios. The scene data generated by the GAN model lacks an effective classification mechanism, making the testing and verification process difficult. At the same time, there is no method for loop feedback optimization of the generated model, resulting in the quality and safety of scene generation not being able to continue to improve with the increase in data volume.

[0027] In response to the above-mentioned technical defects, the present invention combines the conditional instance generative adversarial network (IC-GAN) model with the deep belief network (DBN) model to more realistically simulate the complexity and uncertainty in the actual traffic environment, and provide a more realistic test scenario for the autonomous driving system, especially considering the traffic scenarios involving autonomous driving cars, which can cope with the increasing number of autonomous driving cars with higher and higher levels of autonomous driving in the future.

[0028] As an improved GAN model, the conditional instance generative adversarial network has stronger migration and combination capabilities by introducing instance representation and neighbor samples. It can generate realistic scenes of combinations of unknown, sudden objects (which can cope with the sudden appearance of traffic participants in traffic scenes, such as electric vehicles, pedestrians, etc.) or scenes that have not been seen in the training dataset, which improves the diversity and authenticity of the generated scenes. It can more accurately simulate and generate expected functional safety scenarios with a wider coverage and high similarity to the real road environment. Therefore, the conditional instance generative adversarial network can meet the requirements of expected functional safety for generating unknown scenes.

[0029] The scenario classification method and feedback optimization mechanism based on the deep belief network (DBN) model make the generated scenario data more targeted and practical, which helps to improve the testing and verification efficiency of the autonomous driving system and ensure its performance and safety. Different from the traditional scenario classification into urban road scenarios, highway scenarios, rural road scenarios, intersection scenarios, parking lot scenarios, etc., the expected functional safety can be analyzed from three perspectives: performance limitations, complex environments, and human-computer interaction to improve the quality and safety of the scenario. The expected functional safety scenario data generated by the deep belief network model is more targeted and practical. Therefore, the scenarios of expected functional safety can also be classified and optimized according to these three categories.

[0030] See also Figure 1 The present invention provides a method for automatically generating an expected functional safety scenario, comprising the following steps.

[0031] S100: Collect sensor data, camera data, and vehicle driving trajectory data for training a first conditional instance-generated adversarial network (IC-GAN).

[0032] The conditional instance generative adversarial network is a deep learning model consisting of a generator and a discriminator. The task of the generator is to generate fake data, while the task of the discriminator is to determine whether the input data comes from the real data set or fake data generated by the generator. Unlike traditional GAN ​​networks, conditional instance generative adversarial networks introduce instance conditions, which are usually specific and clear parameters or features used to guide the model to generate scenarios that meet specific functional safety requirements. For the expected functional safety scenario generation task, the expected functional safety classification instance conditions are divided into three categories of labels: performance limitations, complex environments, and human-computer interaction. The model will generate scenarios of corresponding categories based on these three labels.

[0033] Collect raw data, including sensor data, camera data, and vehicle driving trajectory. Collect raw data from public transportation data platforms, actual road test data, and data generated by simulators. Clean, convert, extract features, and annotate these data. There may be outliers, missing values, or inconsistent formats in the data, which require cleaning and format conversion. Extract features related to the expected functional safety from the data, such as vehicle speed, driving trajectory, traffic signal status, etc.

[0034] Construct the first conditional instance generative adversarial network, which includes a generator and a discriminator. The task of the generator is to generate simulated scene data based on random noise and condition information (such as scene type, weather conditions, safety risks, etc.); the task of the discriminator is to determine whether the input data comes from the real scene or fake data generated by the generator. Use the preprocessed data to train the conditional instance generative adversarial network. The parameters of the generator and the discriminator are continuously adjusted through the back-propagation algorithm to make the generated data as close to the real data distribution as possible.

[0035] S200: Generate an expected functional safety scenario using the first conditional instance generative adversarial network, where the expected functional safety scenario includes an unknown and unsafe scenario.

[0036] The trained first conditional instance generative adversarial network is used to generate scene data with similar characteristics and distribution according to specific condition information (such as scene type, weather conditions, etc.). These scene data can include vehicle trajectories, traffic signal status, road geometry, etc.

[0037] In an embodiment of the present invention, the condition information includes a security risk condition, and the security risk condition is used to guide the first condition instance to generate an unknown and unsafe scenario by the adversarial network.

[0038] Of course, the first conditional instance generative adversarial network in the present invention can also generate urban road scenes, highway scenes, rural road scenes, intersection scenes, and parking lot scenes according to different conditional information. The difference is that the classification of the scenes is different from the classification of the prior art. It should be understood that the above-mentioned scenes are limited to the amount of data describing the scenes.

[0039] S300: Classify the expected functional safety scenario based on the machine learning network model to obtain a classification result, and optimize the parameters of the first conditional instance generation adversarial network according to the above classification result to form a second conditional instance generation adversarial network, and use the second conditional instance generation adversarial network to generate a new expected functional safety scenario.

[0040] The generated scenario data is classified using a deep belief network to distinguish different types of safety risks for better testing and verification. The use of a deep belief network includes steps such as data preprocessing, feature extraction, classifier training, and classification result evaluation.

[0041] In one embodiment, the expected functional safety scenarios are classified based on the deep belief network model to obtain classification results; the classification results include performance limitations, complex environments, and human-computer interactions. Of course, in other embodiments, the classification results can also be increased or reduced to meet different functional requirements.

[0042] The deep belief network classifies the scenario data generated by the first conditional instance generative adversarial network, especially for unknown and unsafe scenarios in expected functional safety, so as to facilitate better testing and verification by developers in the subsequent testing process. The classification results can be used to understand whether the generated scenario data is close to the real data distribution. If the generated scenario data is not close to the real data distribution, the developer can adjust the parameters of the first conditional instance generative adversarial network to form a second conditional instance generative adversarial network, and train the second conditional instance generative adversarial network again. Use the second conditional instance generative adversarial network to generate new expected functional safety scenarios. If the generated scenario data is still not close to the real data distribution, the second conditional instance generative adversarial network can be adjusted according to the classification results of the deep belief network to form a third conditional instance generative adversarial network. And so on, the ideal conditional instance generative adversarial network is finally obtained.

[0043] The present invention can generate realistic scenes that are combined with objects or scenes that have not been seen, unknown, or sudden in the training data set through conditional instance generative adversarial networks, thereby meeting the requirements of expected functional safety for generating unknown scenes; through scene classification and feedback optimization mechanism, the generated scene data is more targeted and practical, which helps to improve the testing and verification efficiency of the autonomous driving system and ensure its performance and safety.

[0044] The present invention utilizes the characteristics of conditional instance generative adversarial networks and deep belief networks to complete the training of generative models and scenario generation for expected functional safety scenarios, thereby solving the defects in the prior art and providing a more advanced and comprehensive solution for the generation, testing and verification of expected functional safety scenarios for autonomous driving systems.

[0045] Furthermore, when training the first conditional instance to generate an adversarial network, the first conditional instance to generate an adversarial network can be trained based on random noise and conditional information, and the conditional information includes security risk conditions, and the security risk conditions are used to guide the first conditional instance to generate an unknown and unsafe scenario.

[0046] In one embodiment, in order to make the generated scene data closer to the real data distribution, multiple loss functions are designed, including reconstruction loss, adversarial loss and conditional loss, and the parameters of the generator and discriminator are iteratively updated using optimization algorithms such as gradient descent to minimize the loss function. For example, the parameters of the generator and discriminator are continuously adjusted through the backpropagation (BP) algorithm to make the generated data as close to the real data distribution as possible.

[0047] The present invention can generate simulation data for unknown and unsafe scenarios in expected functional safety, providing strong support for the testing and verification of autonomous driving systems.

[0048] In one embodiment, the above step S300 also includes the following method for generating feedback to the first conditional instance adversarial network, so as to optimize the first conditional instance adversarial network.

[0049] Based on the machine learning network model, the expected functional safety scenarios are classified to obtain scenario types and the number of scenario types. If the data of the scenario types do not meet the preset conditions, the parameters of the first conditional instance generation adversarial network are optimized to form a second conditional instance generation adversarial network.

[0050] For example, the deep belief network model classifies the expected functional safety scenarios and obtains three classification results: performance limitation, complex environment, and human-computer interaction. If the number of the three classification results does not conform to the distribution ratio of the actual data (the distribution ratio can be preset according to the needs), it means that the first conditional instance generation adversarial network needs to be adjusted, such as the first conditional instance generation adversarial network has an overfitting problem.

[0051] In other embodiments, the deep belief network model classifies the expected functional safety scenarios to obtain four classification results: performance limitations, complex environments, human-computer interactions, and safety risks. The safety risk category is mainly used to identify simulation data of unknown and unsafe scenarios in expected functional safety.

[0052] Furthermore, if the data of the scene type does not meet the preset condition of the distribution ratio, an early stopping method or a regularization method is used to optimize the parameters of the first conditional instance generation adversarial network to form a second conditional instance generation adversarial network.

[0053] Early Stopping is a regularization technique used to train machine learning models to prevent overfitting and improve generalization. It prevents overfitting by monitoring the performance of the model on the validation set during training and stopping training when the model performance no longer improves.

[0054] Regularization is a common technique in machine learning to avoid overfitting and improve the generalization ability of the model.

[0055] For the first condition example, the parameters of the generative adversarial network can be manually debugged by the developer, such as by adjusting each constraint function. Of course, some parameters can also be increased or decreased one by one according to the preset conditions during the training process.

[0056] Furthermore, the method for automatically generating expected functional safety scenarios also includes: evaluating the practicality and safety of the expected functional safety scenarios by comparing the types and frequencies of safety risks in real scenarios with those in the expected functional safety scenarios, and adjusting the parameters of the second conditional instance generation adversarial network based on the evaluation results of the practicality and safety.

[0057] Developers can input the generated scenario data into the autonomous driving system for simulation testing and verification. By comparing the types and frequencies of safety risks in real scenarios and generated scenarios, the practicality and safety of the generated scenario data can be evaluated. If the evaluation results are found to be inconsistent with expectations, the first conditional instance generative adversarial network can be further adjusted to further improve the quality and safety of the generated scenarios.

[0058] See also Figure 2 The present invention also provides an automatic generation device for expected functional safety scenarios, including a training module 21, a first generation module 22, and a second generation module 23.

[0059] The training module 21 is used to collect sensor data, camera data, and vehicle driving trajectory data for training the first conditional instance generative adversarial network (IC-GAN).

[0060] The first generating module 22 is used to generate an expected functional safety scenario by using the first conditional instance to generate an adversarial network, where the expected functional safety scenario includes an unknown and unsafe scenario.

[0061] The second generation module 23 is used to classify the expected functional safety scenarios based on the machine learning network model to obtain classification results, and optimize the parameters of the first conditional instance generation adversarial network according to the above classification results to form a second conditional instance generation adversarial network, and use the second conditional instance generation adversarial network to generate new expected functional safety scenarios.

[0062] Furthermore, the first generating module 22 is also used to classify the expected functional safety scenarios based on the deep belief network model to obtain classification results; the classification results include performance limitations, complex environments, and human-computer interactions.

[0063] For the description of the device for automatically generating expected functional safety scenarios, please refer to the method for automatically generating expected functional safety scenarios, which will not be repeated here.

[0064] like Figure 3 As shown, the present invention also provides an electronic device, including: At least one processor; and at least one memory in communication with the processor, wherein: the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the above-mentioned method for automatically generating expected functional safety scenarios.

[0065] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for automatically generating expected functional safety scenarios is implemented.

[0066] It can be understood that computer-readable storage media may include: any entity or device capable of carrying a computer program, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), and a software distribution medium, etc. A computer program includes a computer program code. The computer program code may be in source code form, object code form, an executable file, or some intermediate form, etc. A computer-readable storage medium may include: any entity or device capable of carrying a computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), and a software distribution medium, etc.

[0067] In certain embodiments of the present invention, the electronic device may include a controller or a processor, and the controller is a single-chip microcomputer chip that integrates a processor, a memory, a communication module, etc. The processor may refer to a processor included in the controller. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.

[0068] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0069] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0070] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for automatically generating an expected functional safety scenario, characterized in that: include: Collect sensor data, camera data, and vehicle trajectory data for training the first conditional instance-generated adversarial network (IC-GAN); Generate an expected functional safety scenario using the first conditional instance generative adversarial network, where the expected functional safety scenario includes an unknown and unsafe scenario; The expected functional safety scenarios are classified based on the machine learning network model to obtain classification results, and the parameters of the first conditional instance generation adversarial network are optimized according to the above classification results to form a second conditional instance generation adversarial network, and the second conditional instance generation adversarial network is used to generate new expected functional safety scenarios.

2. The method for automatically generating expected functional safety scenarios according to claim 1, characterized in that: The collecting sensor data, camera data, and vehicle driving trajectory data for training the first conditional instance generation adversarial network includes: The first conditional instance generative adversarial network is trained based on random noise and conditional information, wherein the conditional information includes a security risk condition, and the security risk condition is used to guide the first conditional instance generative adversarial network to generate an unknown and unsafe scenario.

3. The method for automatically generating expected functional safety scenarios according to claim 1, characterized in that: The classifying the expected functional safety scenario based on the machine learning network model to obtain a classification result includes: The expected functional safety scenarios are classified based on a deep belief network model to obtain classification results; the classification results include performance limitations, complex environments, and human-computer interactions.

4. The method for automatically generating expected functional safety scenarios according to claim 3, characterized in that: The method of classifying the expected functional safety scenario based on the machine learning network model to obtain a classification result, and optimizing the parameters of the first conditional instance generation adversarial network according to the classification result to form a second conditional instance generation adversarial network, includes: Based on the machine learning network model, the expected functional safety scenarios are classified to obtain scenario types and the number of scenario types. If the data of the scenario types do not meet the preset conditions, the parameters of the first conditional instance generation adversarial network are optimized to form a second conditional instance generation adversarial network.

5. The method for automatically generating expected functional safety scenarios according to claim 4, characterized in that: If the data of the scenario type does not meet the preset condition, optimizing the parameters of the first conditional instance generation adversarial network to form a second conditional instance generation adversarial network, including: If the data of the scenario type does not meet the preset conditions, an early stopping method or a regularization method is used to optimize the parameters of the first conditional instance generation adversarial network to form a second conditional instance generation adversarial network.

6. The method for automatically generating expected functional safety scenarios according to claim 1, characterized in that: Also includes: By comparing the types and frequencies of security risks in real scenarios and the expected functional safety scenarios, the practicality and safety of the expected functional safety scenarios are evaluated, and the parameters of the second conditional instance generation adversarial network are adjusted based on the evaluation results of the practicality and safety.

7. An automatic generation device for expected functional safety scenarios, characterized in that: include: A training module, used to collect sensor data, camera data, and vehicle trajectory data for training the first conditional instance generative adversarial network (IC-GAN); A first generating module, configured to generate an expected functional safety scenario using the first conditional instance generative adversarial network, wherein the expected functional safety scenario includes an unknown and unsafe scenario; The second generation module is used to classify the expected functional safety scenarios based on the machine learning network model to obtain classification results, and optimize the parameters of the first conditional instance generation adversarial network according to the above classification results to form a second conditional instance generation adversarial network, and use the second conditional instance generation adversarial network to generate new expected functional safety scenarios.

8. The device for automatically generating expected functional safety scenarios according to claim 7, characterized in that: The first generation module is also used to classify the expected functional safety scenarios based on the deep belief network model to obtain classification results; the classification results include performance limitations, complex environments, and human-computer interactions.

9. An electronic device, characterized in that: include: at least one processor; and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the method for automatically generating expected functional safety scenarios as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a computer, the method for automatically generating an expected functional safety scenario as claimed in any one of claims 1 to 6 is executed.

Images