Automatic driving interactive test scene generation device and method, equipment and medium

Abstract

The invention relates to the technical field of automatic driving, in particular to an automatic driving interactive test scene generation device, method and equipment and a medium, and the device comprises an expanding and writing device which is used for decomposing and supplementing a target natural language description to generate a plurality of scene descriptions; the weather simulator is used for performing scene simulation according to the weather information to generate weather parameters; the self-vehicle positioner is used for searching a plurality of self-vehicle generation positions on a preset map according to the self-vehicle position; the countermeasure vehicle positioner is used for searching a plurality of countermeasure vehicle static positions on a preset map according to the countermeasure vehicle relative position and the own vehicle position; the behavior generator is used for constructing a behavior topology network according to the interactive behavior library so as to generate a plurality of interactive dynamic behaviors; and the scene generation module is used for constructing a plurality of comprehensive test scenes according to the obtained related data. Therefore, the problem that in the prior art, the interactivity of scene generation and the matching degree with user requirements have obvious defects is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, and particularly relates to an automatic driving interactive test scene generation device and method based on a large language model, equipment and a medium. BACKGROUND

[0002] Currently, the generation of test and training scenes for autonomous vehicles still faces multiple challenges. The scenes collected from the real world are often not dangerous enough, and the proportion of rare dangerous scenes is extremely low, so they cannot provide enough diverse scenes for testing or training. In order to increase the proportion of dangerous scenes in the total test / training scene library, the scenes extracted and collected from the real world have the problem of scarcity of dangerous scenes, and cannot provide enough diverse training or test scenes. In order to increase the frequency of dangerous scenes to increase the robustness of training and testing, existing methods mainly include filtering scenes from real world data and manually predefining adversarial vehicle models in scenes. Although the above two methods can generate certain scenes, they have multiple shortcomings.

[0003] Firstly, the method based on real world data is limited by the availability and diversity of data, especially in terms of obtaining safety critical scenes, which are highly scarce and rarely occur in reality, so it is difficult to fully cover them through real data. In addition, this method requires a large amount of data collection and processing, which has the problem of high cost from both economic and time perspectives.

[0004] Secondly, the method based on adversarial vehicle models in manually predefined scenes can create safety critical scenes and ensure the reasonableness of vehicle behavior, but this method usually relies on manually set rules and scenes, which limits the diversity and authenticity of generated scenes. In addition, these methods are difficult to simulate complex and unpredictable driving behaviors, which is one of the key factors for safety evaluation of autonomous driving systems.

[0005] In recent years, methods based on large language models (LLMs) have been proposed to generate scenarios using natural language descriptions and to generate corresponding scenario codes for playback in simulators using the understanding power of large language models. This approach can improve the flexibility and diversity of scenario generation. However, existing methods of this type still have deficiencies in static scenario diversity, background vehicle behavior interactivity, and integration with existing simulation platforms. For example, while 2D simulators such as SUMO can quickly generate scenarios, they lack the realism of 3D environments, and 3D simulators such as CARLA require complex vehicle control logic. Furthermore, existing LLM methods often struggle to accurately control the behavior of background vehicles when processing natural language descriptions, and they are difficult to integrate with modern vehicle autonomous driving algorithms, i.e., the behavior of background vehicles often relies on predefined behavior patterns rather than being intelligent and purpose-driven as real vehicles. In addition, current large language model-based scenario generation tools generally lack interactivity and dynamism, with background vehicle behavior not interacting with its surrounding environment, resulting in scenarios that are not realistic or challenging enough for simulation testing. Therefore, existing technologies have significant deficiencies in the interactivity of scenario generation and the matching degree with user needs. These shortcomings limit the training efficiency of autonomous driving systems and reduce the robustness of autonomous driving systems to extreme driving environments. Therefore, there is an urgent need in the industry for a new solution that can generate realistic, diverse, and interactive scenarios according to user needs. SUMMARY

[0006] The present application provides an automatic driving interactive test scenario generation device, method, equipment and medium to solve the problem of obvious defects in the interactivity of scenario generation and the matching degree with user needs of the prior art.

[0007] The first aspect of the present application provides an automatic driving interactive test scenario generation device, comprising: an expander for decomposing and supplementing a target natural language description to generate a plurality of scenario descriptions, wherein the plurality of scenario descriptions include weather information, ego vehicle position, adversarial vehicle relative position, and an interactive behavior library; a weather simulator for simulating a scene according to the weather information to generate weather parameters that can be used for simulation environment rendering; an ego vehicle positioner for searching a plurality of ego vehicle generation positions on a preset autonomous driving simulator map according to the ego vehicle position; an adversarial vehicle positioner for searching a plurality of adversarial vehicle static positions on the preset autonomous driving simulator map according to the adversarial vehicle relative position and the ego vehicle position; a behavior generator for constructing a behavior topology network according to the interactive behavior library and generating a plurality of interactive dynamic behaviors using the behavior topology network; and a scenario generation module for constructing a plurality of comprehensive test scenarios according to the weather parameters, the plurality of ego vehicle generation positions, the plurality of adversarial vehicle static positions, and the plurality of interactive dynamic behaviors.

[0008] Optionally, the weather parameter comprises at least one of a solar elevation angle, a wind force, a precipitation amount, a fog concentration, and a dust storm intensity.

[0009] Optionally, the ego vehicle positioner comprises:

[0010] a first constructing unit configured to construct a position feasibility judgment function according to the ego vehicle position; and a first searching unit configured to search, based on the preset automatic driving simulator map, a plurality of ego vehicle generated positions satisfying a preset position requirement by using the position feasibility judgment function.

[0011] Optionally, the adversarial vehicle positioner comprises:

[0012] a second constructing unit configured to construct a surrounding vehicle position information function of the ego vehicle position according to the adversarial vehicle relative position; and a second searching unit configured to input the ego vehicle position into the surrounding vehicle position information function to search the plurality of adversarial vehicle static positions in the preset automatic driving simulator map.

[0013] Optionally, the behavior generator comprises:

[0014] a third constructing unit configured to select a plurality of discrete behaviors in the interactive behavior library and construct the behavior topology network according to the plurality of discrete behaviors; and a generating unit configured to generate an interactive dynamic behavior by using the behavior topology network.

[0015] Optionally, the third constructing unit comprises a screening subunit, a vehicle model selection subunit, a vehicle model configuration subunit, a behavior success / failure condition subunit, and a constructing subunit, wherein:

[0016] the screening subunit is configured to screen a plurality of discrete behaviors conforming to the plurality of scene descriptions in the interactive behavior library; the vehicle model selection subunit is configured to select a corresponding automatic driving model for a vehicle performing any discrete behavior; the vehicle model configuration subunit is configured to interactively adjust a behavior parameter of the corresponding automatic driving model according to each discrete behavior; the behavior success / failure condition subunit is configured to define a motion state of the each discrete behavior; and the constructing subunit is configured to sequentially connect or logically connect in parallel the each discrete behavior according to the motion state of the each discrete behavior to form the behavior topology network.

[0017] The second aspect embodiment of the present application provides an automatic driving interactive test scene generation method, comprising the following steps: decomposing and supplementing the target natural language to generate a plurality of scene descriptions, wherein the plurality of scene descriptions comprise weather information, ego vehicle position, opponent vehicle relative position and interaction behavior library; simulating a scene according to the weather information to generate weather parameters that can be used for simulation environment rendering; searching a plurality of ego vehicle generation positions on a preset automatic driving simulator map according to the ego vehicle position; searching a plurality of opponent vehicle static positions on the preset automatic driving simulator map according to the opponent vehicle relative position and the ego vehicle position; constructing a behavior topology network according to the interaction behavior library, and generating a plurality of interactive dynamic behaviors by using the behavior topology network; and constructing a plurality of comprehensive test scenes according to the weather parameters, the plurality of ego vehicle generation positions, the plurality of opponent vehicle static positions and the plurality of interactive dynamic behaviors.

[0018] The third aspect embodiment of the present application provides an electronic device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the automatic driving interactive test scene generation method as described in the above embodiments.

[0019] The fourth aspect embodiment of the present application provides a computer program product, wherein the computer program / instruction is executed by the processor to implement the automatic driving interactive test scene generation method as described above.

[0020] The fifth aspect embodiment of the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the program is executed by the processor to implement the automatic driving interactive test scene generation method as described above.

[0021] The automatic driving interactive test scene generation device, method, equipment and medium provided by the embodiment of the application can generate a real and diverse 3D scene according to the natural language description of a user, including a road map, weather conditions, background vehicle behavior and the like; through modular design and collaborative work of multiple LLM agents, the efficiency and controllability of scene generation are ensured, and significant progress is made in scene authenticity and consistency between background vehicle behavior and semantic description; various non-learning or learning vehicle control algorithms can be integrated to ensure the rationality of vehicle behavior, so that the behavior of surrounding vehicles is not only constrained by dynamics, but also constrained by a control model, the generated scene can not only simulate a complex traffic environment, but also ensure that the behavior of the background vehicle is physically feasible and conforms to the driving rules in the real world, and then the user can generate a highly real and reliable test scene, thereby improving the training efficiency and safety of the automatic driving system; real-time interaction between surrounding vehicles and the ego vehicle, and between surrounding vehicles, can be realized to obtain a dynamic scene, and corresponding countermeasures can be taken under different ego vehicle behaviors, which makes the generated scene not only highly real, but also dynamically adjusts the vehicle behavior in the scene according to the user's needs, and also enables the automatic driving system to face more diverse and challenging scenes during the training process, thereby further improving the robustness and safety of the system; the user can quickly generate a test scene that meets the actual needs, reduce the dependence on real data, reduce the test cost, and improve the training efficiency and safety of the automatic driving system; through the unique modular design, integrated vehicle control algorithm and real-time interaction capability, multiple key problems in the prior art are solved, and a more effective and flexible solution is provided for the testing and training of an autonomous vehicle.

[0022] Additional aspects and advantages of the application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following and / or can be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0023] The above and / or additional aspects and advantages of the application will become apparent and be readily appreciated from the following description, taken in conjunction with the accompanying drawings, in which:

[0024] Figure 1 A block schematic diagram of an automatic driving interactive test scene generation device provided by an embodiment of the application;

[0025] Figure 2 A specific flowchart of an automatic driving interactive test scene generation device provided by an embodiment of the application;

[0026] Figure 3 A schematic diagram of the working principle of a behavior generator provided by an embodiment of the application;

[0027] Figure 4A structural schematic diagram of a behavior topology network provided by an embodiment of the present application.

[0028] Figure 5 A flowchart of an automatic driving interactive test scene generation method provided by an embodiment of the present application.

[0029] Figure 6 A structural schematic diagram of an electronic device provided by an embodiment of the present application.

[0030] Legend of reference signs:

[0031] 10 - automatic driving interactive test scene generation device, 101 - expander, 102 - weather simulator, 103 - ego vehicle positioner, 104 - opponent vehicle positioner, 105 - behavior generator, 106 - scene generation module, 60 - electronic device, 601 - memory, 602 - processor and 603 - communication interface. DETAILED DESCRIPTION

[0032] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings, in which the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as limiting the present application.

[0033] The automatic driving interactive test scene generation device, method, equipment and medium of the embodiments of the present application are described below with reference to the accompanying drawings.

[0034] Figure 1 A block schematic diagram of an automatic driving interactive test scene generation device provided by an embodiment of the present application.

[0035] As shown in Figure 1 , the automatic driving interactive test scene generation device 10 includes an expander 101, a weather simulator 102, an ego vehicle positioner 103, an opponent vehicle positioner 104, a behavior generator 105 and a scene generation module 106. The device proposed by the embodiments of the present application is that the natural language description input of the user is input into the expander, the expander refines and cuts the information contained therein into descriptions for different scene contents, these descriptions will be respectively transmitted to all the remaining large language model agents, the outputs of each agent will be summarized and stored in a scene parameter file, and the scene playback and evaluation are performed by a scene player.

[0036] The interpreter 101 is used to decompose and supplement the target natural language description to generate multiple scene descriptions, including weather information, ego vehicle position, opponent vehicle relative position, and interaction behavior library. The weather simulator 102 is used to simulate a scene according to the weather information to generate weather parameters that can be used for simulation environment rendering. The ego locator 103 is used to search for multiple ego vehicle generated positions on a preset automatic driving simulator map according to the ego vehicle position. The opponent locator 104 is used to search for multiple opponent vehicle static positions on the preset automatic driving simulator map according to the opponent vehicle relative position and the ego vehicle position. The behavior generator 105 is used to construct a behavior topology network according to the interaction behavior library, and generate multiple interactive dynamic behaviors using the behavior topology network. The scene generation module 106 is used to construct multiple comprehensive test scenes according to the weather parameters, the multiple ego vehicle generated positions, the multiple opponent vehicle static positions, and the multiple interactive dynamic behaviors.

[0037] In some embodiments, as shown in FIG. 1, the interpreter 101 is responsible for interpreting the user's natural language description and decomposing it into multiple aspects such as weather conditions, ego vehicle position, relative position of the opponent vehicle, and its interaction behavior, etc. If the user does not specify some information, the interpreter will supplement it according to its understanding of the scene. Figure 2

[0038] In some embodiments, the weather simulator 102 is responsible for simulating weather conditions during simulation, which is crucial for training or testing involving perception. The weather simulator takes the weather description described in the interpreter as input and generates weather parameters that can be used for simulation environment rendering, including sun elevation angle, wind force, precipitation amount, fog density, dust storm intensity, etc. Weather simulation will have a significant impact on vehicle perception in the simulation environment, and can enhance the realism and diversity of the generated scene.

[0039] In some embodiments, the ego locator 103 includes:

[0040] The first construction unit is used to construct a position feasibility judgment function according to the ego vehicle position.

[0041] The first search unit is used to search for multiple ego vehicle generated positions that meet the preset position requirements based on the preset automatic driving simulator map using the position feasibility judgment function.

[0042] ​Specifically, the ego vehicle locator searches for the required ego vehicle generation points in the CARLA autonomous driving simulator map, such as "in the intersection", "in the roundabout", "in a long straight road", etc. The ego vehicle locator takes the ego vehicle position description in the expander as input, and outputs the ego vehicle generation position feasibility judgment function. The possible ego vehicle position information is input into the function, and if the position meets the input semantic description, it returns true, otherwise it returns false.

[0043] In some embodiments, the adversarial vehicle locator 104 (Adv Locator) comprises:

[0044] A second construction unit for constructing a surrounding vehicle position information function of the ego vehicle position according to the adversarial vehicle relative position;

[0045] A second search unit for inputting the ego vehicle position into the surrounding vehicle position information function to search for a plurality of adversarial vehicle static positions in the preset autonomous driving simulator map.

[0046] Specifically, the adversarial vehicle locator is responsible for generating the static positions of the adversarial vehicles, ensuring that these vehicles are placed on the road in compliance with the ego vehicle. The adversarial vehicle locator takes the adversarial vehicle relative position in the expander output result as input, and outputs a function that searches for surrounding vehicle position information according to the ego vehicle position. If the position exists, is located on the driving lane, and has no other obstacles, the function returns the adversarial vehicle position information, otherwise returns false. This function is combined with the ego vehicle generation position feasibility judgment function generated by the ego vehicle locator, i.e. searching for all possible ego vehicle and surrounding vehicle position information in the map and storing it to the corresponding scene parameter file.

[0047] In some embodiments, the behavior generator 105 (Action Generator) comprises:

[0048] A third construction unit for selecting a plurality of discrete behaviors in the interactive behavior library, and constructing a behavior topology network according to the plurality of discrete behaviors;

[0049] A generation unit for generating an interactive dynamic behavior using the behavior topology network.

[0050] Specifically, as shown in Figure 3 The behavior generator is the key component of the traffic scene dynamic interaction, responsible for generating the dynamic behavior of the adversarial vehicle. This module selects appropriate discrete behaviors, configures these behaviors to meet actual needs, and connects them using parallel or sequential logic to construct an interactive behavior topology network. The behavior generator takes the adversarial vehicle behavior description in the expander output as input, and outputs the behavior topology network in the form of code.

[0051] In some embodiments, the third construction unit comprises a screening subunit, a vehicle model selection subunit, a vehicle model configuration subunit, a behavior success / failure condition subunit, and a construction subunit, wherein,

[0052] The screening subunit is configured to screen a plurality of discrete behaviors that meet a plurality of scenario descriptions in the interaction behavior library.

[0053] The vehicle model selection subunit is configured to select a corresponding autonomous driving model for a vehicle performing any discrete behavior.

[0054] The vehicle model configuration subunit is configured to interactively adjust the behavior parameters of the corresponding autonomous driving model according to each discrete behavior.

[0055] The behavior success / failure condition subunit is configured to define the motion state of each discrete behavior.

[0056] The construction subunit is configured to sequentially connect or logically connect each discrete behavior in parallel according to the motion state of each discrete behavior to form a behavior topology network.

[0057] Specifically, the behavior generator 105 core lies in constructing a behavior topology web composed of discrete behaviors. The first step of the workflow of the behavior generator is the construction of discrete behaviors. As the smallest behavior unit that can be executed by a vehicle, the discrete behavior includes a vehicle model selection subunit, a vehicle model configuration subunit, and a behavior success / failure condition subunit. The vehicle model selection subunit refers to selecting a suitable autonomous driving model for a vehicle performing a specific discrete behavior. In the embodiments of the present application, multiple models can be compatible, such as the built-in autonomous driving model of CARLA, the ACC following vehicle model, and the PlanT intelligent agent based on imitation learning. The vehicle model configuration subunit is to interactively adjust the behavior parameters (such as speed, acceleration, and heading angle) of the LLM model input behavior on the basis of the selected vehicle model. For example, when configuring the “follow vehicle” behavior, the parameters such as the target to follow, the target speed, and the following distance can be adjusted to achieve precise control of the vehicle behavior. The optional behavior success / failure condition subunit is used to define the running state of the discrete behavior, for example, the success condition of the “stop vehicle” behavior can be set as the vehicle speed dropping to zero. Through the above steps, the present application can construct a variety of discrete behaviors with high controllability, which lays a foundation for the generation of complex scenarios.

[0058] The behavior generator proposed in the embodiments of the present application can dynamically combine discrete behaviors to construct an interactive behavior topology network, thereby giving the scene dynamic interactivity. In actual application, the behavior generator first selects behavior units related to the scene from a discrete behavior library based on the scene description. Then, according to the specific requirements of the scene, the selected behavior units are configured with parameters, for example, when the scene description contains "vehicle A quickly cuts into the lane of vehicle B", the "cutting into the lane" behavior unit is selected, and the target vehicle is set to vehicle B, and the cutting speed is set to a high value. In order to simulate more complex vehicle behaviors, the embodiments of the present application also support connecting multiple discrete behaviors using sequential or parallel logic to form behavior sequences or behavior combinations. For example, behaviors can be executed in the order of "accelerate-cut into the lane", or the "follow the vehicle" and "keep in the lane" behaviors can be executed simultaneously. Figure 4 The modules in the network represent discrete behaviors, and the arrows represent the relationship between behaviors, and the network structure can be updated in real time according to the environment and vehicle state. Taking the behavior topology network in Figure 4 For example, the opponent vehicle 1 drives straight to the intersection and turns right, and does not interact with any other vehicle as specified, and if no dangerous situation needs to be responded to, the vehicle will complete the predetermined behavior completely independently; the opponent vehicle 2 and the opponent vehicle 3 interact with the ego vehicle respectively, where the opponent vehicle 2 will read the position and speed information of the ego vehicle in real time, increase its own speed to overtake the ego vehicle, and perform a lane change operation when the ego vehicle has not been completely overtaken, and the opponent 3 will also read the position and speed information of the opponent vehicle in real time, and plan a following strategy according to the ego vehicle's Auto Cruise Control algorithm. The opponent vehicle 4 only interacts with the opponent vehicle 3, reads its speed and position information in real time to complete the following. This dynamic adjustment mechanism gives the background vehicle the ability to respond to changes in the ego vehicle's behavior, enabling dynamic interactive behavior between vehicles.

[0059] The design of the behavior topology network enables background vehicles and ego vehicles to refer to and interact with each other. This interactivity makes the scene more dynamic and realistic, enabling dynamic adjustment of the behavior of the surrounding vehicles according to the behavior of the ego vehicle, thereby improving the safety and challenge of the scene. In this way, users can easily generate test scenarios that meet actual needs, reduce dependence on real data, reduce testing costs, and improve the training efficiency and safety of the autonomous driving system. This unique design enables the embodiments of the present application to significantly improve the training efficiency and safety of the autonomous driving system, reduce testing costs, and provide a new solution for the development of autonomous driving technology.

[0060] In addition, as Figure 2As shown, the embodiment of the present application can also include a random traffic generator that randomly generates traffic flow around the ego vehicle according to the ego vehicle position and the opponent vehicle position to simulate real traffic scenarios, and then inputs into the scene generation module 106 to construct a plurality of comprehensive test scenarios together with the weather parameters, a plurality of ego vehicle generation positions, a plurality of opponent vehicle static positions, and a plurality of interactive dynamic behaviors.

[0061] In summary, the automatic driving interactive test scene generation device according to the embodiment of the present application has the following beneficial effects:

[0062] (1) By integrating a variety of learning or non-learning vehicle control algorithms, it is ensured that the behavior of the background vehicle is not only constrained by dynamics, but also constrained by the control model, thereby improving the authenticity and reliability of scene generation. This design allows users to generate highly realistic and reliable test scenes, reduces dependence on real data, reduces testing costs, and improves the training efficiency and safety of the automatic driving system;

[0063] (2) Real-time interaction between the ego vehicle and the surrounding vehicles, as well as between the surrounding vehicles, can be achieved, resulting in dynamic scenes; and corresponding opponent behaviors can be made in different ego vehicle behaviors, which not only makes the generated scenes highly realistic, but also dynamically adjusts the vehicle behavior in the scene according to the user's needs, and also enables the automatic driving system to face more diverse and challenging scenes during the training process, thereby further improving the robustness and safety of the system;

[0064] (3) Through its unique modular design, integrated vehicle control algorithms, and real-time interaction capabilities, it provides a more effective and flexible solution for testing and training of autonomous vehicles. This innovative solution not only improves the efficiency and authenticity of scene generation, but also provides a new solution for the development of autonomous driving technology.

[0065] Next, the automatic driving interactive test scene generation method according to the embodiment of the present application is described with reference to the accompanying drawings.

[0066] Figure 5 A flowchart of an automatic driving interactive test scene generation method provided by the embodiment of the present application.

[0067] As Figure 5 shown, the automatic driving interactive test scene generation method includes the following steps:

[0068] In step S501, the target natural language is decomposed and supplemented to generate a plurality of scene descriptions, wherein the plurality of scene descriptions include weather information, ego vehicle position, opponent vehicle relative position, and interactive behavior library.

[0069] In step S502, scene simulation is performed according to the weather information to generate weather parameters that can be used for rendering of the simulation environment.

[0070] In some embodiments, the weather parameters include at least one of a solar elevation angle, wind power, precipitation, fog concentration, and dust storm intensity.

[0071] In step S503, a plurality of self-vehicle generation positions are searched on a preset automatic driving simulator map according to the self-vehicle position.

[0072] In some embodiments, searching the plurality of self-vehicle generation positions on the preset automatic driving simulator map according to the self-vehicle position includes:

[0073] constructing a position feasibility judgment function according to the self-vehicle position;

[0074] searching the plurality of self-vehicle generation positions that meet the preset position requirements based on the preset automatic driving simulator map and using the position feasibility judgment function.

[0075] In step S504, a plurality of static positions of the opposing vehicle are searched on the preset automatic driving simulator map according to the relative position of the opposing vehicle and the self-vehicle position.

[0076] In some embodiments, searching the plurality of static positions of the opposing vehicle on the preset automatic driving simulator map according to the relative position of the opposing vehicle and the self-vehicle position includes:

[0077] constructing a surrounding vehicle position information function of the self-vehicle position according to the relative position of the opposing vehicle;

[0078] inputting the self-vehicle position into the surrounding vehicle position information function to search the plurality of static positions of the opposing vehicle in the preset automatic driving simulator map.

[0079] In step S505, a behavior topology network is constructed according to the interaction behavior library, and a plurality of interactive dynamic behaviors are generated using the behavior topology network.

[0080] In some embodiments, constructing the behavior topology network according to the interaction behavior library and generating the plurality of interactive dynamic behaviors using the behavior topology network includes:

[0081] selecting a plurality of discrete behaviors in the interaction behavior library and constructing the behavior topology network according to the plurality of discrete behaviors;

[0082] generating the interactive dynamic behaviors using the behavior topology network.

[0083] In some embodiments, selecting the plurality of discrete behaviors in the interaction behavior library and constructing the behavior topology network according to the plurality of discrete behaviors includes:

[0084] Screening a plurality of discrete behaviors in the interactive behavior library that meet a plurality of scene descriptions;

[0085] Selecting a corresponding automatic driving model for a vehicle performing any discrete behavior;

[0086] Interactively adjusting the behavior parameters of the corresponding automatic driving model according to each discrete behavior;

[0087] Defining the motion state of each discrete behavior;

[0088] According to the motion state of each discrete behavior, sequentially connecting or logically connecting each discrete behavior in parallel to form a behavior topology network.

[0089] In step S506, a plurality of comprehensive test scenes are constructed according to weather parameters, a plurality of ego vehicle generated positions, a plurality of opponent vehicle static positions, and a plurality of interactive dynamic behaviors.

[0090] It should be noted that the foregoing explanation and description of the embodiment of the automatic driving interactive test scene generation device are also applicable to the automatic driving interactive test scene generation method of the embodiment, which will not be described here.

[0091] The automatic driving interactive test scene generation method according to the embodiment of the present application has the following beneficial effects:

[0092] (1) By integrating various learning or non-learning vehicle control algorithms, the behavior of the background vehicle is not only constrained by dynamics but also by control models, thereby improving the authenticity and reliability of scene generation. This design allows users to generate highly realistic and reliable test scenes, reduces dependence on real data, reduces testing costs, and improves the training efficiency and safety of the automatic driving system;

[0093] (2) Real-time interaction between the ego vehicle and the surrounding vehicles, and between the surrounding vehicles, can be achieved to obtain dynamic scenes. In the case of different ego vehicle behaviors, corresponding countermeasures can be taken, which not only makes the generated scenes highly realistic, but also dynamically adjusts the vehicle behavior in the scene according to user needs, and also enables the automatic driving system to face more diverse and challenging scenes during the training process, thereby further improving the robustness and safety of the system;

[0094] (3) Through its unique modular design, integrated vehicle control algorithms, and real-time interaction capabilities, the automatic driving vehicle testing and training provides a more effective and flexible solution. This innovative solution not only improves the efficiency and authenticity of scene generation, but also provides a new solution for the development of autonomous driving technology.

[0095] Figure 6A structural schematic diagram of an electronic device provided by an embodiment of the present application.

[0096] The electronic device can include a memory 601, a processor 602, and a computer program stored in the memory 601 and executable on the processor 602.

[0097] The processor 602 implements the automatic driving interactive test scene generation method provided in the above embodiments when executing the program.

[0098] Further, the electronic device further includes:

[0099] A communication interface 603 for communication between the memory 601 and the processor 602.

[0100] The memory 601 is used to store the computer program executable on the processor 602.

[0101] The memory 601 can include a high-speed RAM memory, and can also include a non-volatile memory, such as at least one disk memory.

[0102] If the memory 601, the processor 602 and the communication interface 603 are independently implemented, the communication interface 603, the memory 601 and the processor 602 can be connected to each other through a bus and complete communication between each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 In the figure, only one thick line is used to represent, but it does not mean that there is only one bus or one type of bus.

[0103] Optionally, in specific implementation, if the memory 601, the processor 602 and the communication interface 603 are integrated on a chip, the memory 601, the processor 602 and the communication interface 603 can complete communication between each other through an internal interface.

[0104] The processor 602 can be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0105] The embodiment of the present application also provides a computer program product, wherein the computer program / instruction is executed by a processor to realize the automatic driving interactive test scene generation method.

[0106] The embodiment of the present application also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to realize the automatic driving interactive test scene generation method.

[0107] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present application and the features of the different embodiments or examples without contradiction.

[0108] In addition, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one feature. In the description of the present application, the meaning of "N" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0109] Any process or method descriptions in flow charts or otherwise described herein represent embodiments of examples that can be implemented as executable instructions stored in a computer readable storage medium (or multiple computer readable storage media) for execution by a processor that can be implemented as hardware, software, firmware, or any combination thereof. In this description and in the claims, the terms "processor" and "computer" are understood to encompass both hardware and software. The various embodiments of the application have been described in the specification and illustrated in the drawings by way of structural and functional equivalents of the examples described in this specification and illustrated in the drawings. Therefore, it is to be understood that the application is not limited to the examples described in this specification and illustrated in the drawings, but includes all structural and functional equivalents of the examples described in this specification and illustrated in the drawings.

[0110] The logic and / or steps represented in the flowcharts and / or described herein, for example, can be considered as a sequence of executable instructions stored in a computer readable medium, which can be executed by an instruction execution system, apparatus or device, such as a computer-based system, a processor-based system, or other system that can fetch the instructions from the instruction execution system, apparatus or device and execute the instructions, or a combination of the above. For the purposes of this specification, a "computer readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus or device. The computer readable medium can be a computer readable storage medium or a computer readable signal medium. The computer readable storage medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a propagation medium. The computer readable signal medium can include, but is not limited to, a computer readable medium that facilitates transfer of the program from one place to another. A specific example of a computer readable medium is a non-transitory computer-readable storage medium. A specific example of a computer readable signal medium is a source or destination of the computer readable medium. Another specific example of a computer readable signal medium is a computer readable signal travelling through space. Thus, a computer readable medium can take many forms of hardware to carry out the program for use by or in connection with the instruction execution system, apparatus or device.

[0111] It should be understood that aspects of the application can be implemented in hardware, software, firmware or a combination thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented in hardware and in another embodiment, the hardware can be implemented with any or a combination of the following technologies, which are all well known in the art: a discrete logic circuit(s) having logic gates for implementing logic functions upon an application of data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array(s) (PGA), a field programmable gate array (FPGA), etc.

[0112] Those of skill in the art would understand that the steps carried out in the above-mentioned embodiment methods can be carried out by program instructions to relevant hardware, and the program can be stored in a computer readable storage medium. When the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0113] In addition, each function unit in each embodiment of the present application can be integrated in one processing module, or each unit can be physically present separately, or two or more units can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of a software function module. When the integrated module is realized in the form of a software function module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0114] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. An interactive test scenario generation device for autonomous driving, characterized in that, The method comprises the following steps: The method comprises the following steps: The weather parameters comprise at least one of the following: solar elevation angle, wind force, precipitation, fog concentration and dust storm intensity. The self-vehicle locator comprises: The first constructing unit is configured to construct a position feasibility judgment function according to the self-vehicle position. The first searching unit is configured to search for a plurality of self-vehicle generated positions satisfying preset position requirements based on the preset automatic driving simulator map and by using the position feasibility judgment function. The anti-vehicle locator comprises:

2. The apparatus for automatically generating an interactive test scenario for autonomous driving according to claim 1, wherein, The second constructing unit is configured to construct a surrounding vehicle position information function of the self-vehicle position according to the anti-vehicle relative position.

3. The apparatus for automatically generating an interactive test scenario for autonomous driving according to claim 1, wherein, The second searching unit is configured to input the self-vehicle position into the surrounding vehicle position information function to search for the plurality of anti-vehicle static positions in the preset automatic driving simulator map. The behavior generator comprises: The third constructing unit is configured to select a plurality of discrete behaviors in the interaction behavior library and construct the behavior topology network according to the plurality of discrete behaviors.

4. The apparatus for automatically generating an interactive test scenario for autonomous driving according to claim 1, wherein, The generating unit is configured to generate the interactive dynamic behavior by using the behavior topology network. The third constructing unit comprises a screening subunit, a vehicle model selection subunit, a vehicle model configuration subunit, a behavior success / failure condition subunit and a constructing subunit, wherein The screening subunit is configured to screen a plurality of discrete behaviors conforming to the plurality of scene descriptions in the interaction behavior library.

5. The apparatus for automatically generating an interactive test scenario for autonomous driving according to claim 1, wherein, The vehicle model selection subunit is configured to select a corresponding automatic driving model for a vehicle performing any discrete behavior. The vehicle model configuration subunit is configured to interactively adjust the behavior parameters of the corresponding automatic driving model according to each discrete behavior. The behavior success / failure condition subunit is configured to define the motion state of each discrete behavior.

6. The apparatus for automatically generating an interactive test scenario for autonomous driving according to claim 5, wherein, The constructing subunit is configured to sequentially connect or parallelly logically connect each discrete behavior according to the motion state of each discrete behavior to form the behavior topology network. The method comprises the following steps: The method comprises the following steps: ​ ​ ​ 7. An automatic interactive test scenario generation method for autonomous vehicles, characterized in that, ​ decomposing and supplementing the target natural language to generate a plurality of scene descriptions, wherein the plurality of scene descriptions comprises weather information, ego vehicle position, adversarial vehicle relative position, and interaction behavior library; performing scene simulation according to the weather information to generate weather parameters that can be used for simulation environment rendering; searching a plurality of ego vehicle generated positions on a preset automatic driving simulator map according to the ego vehicle position; searching a plurality of adversarial vehicle static positions on the preset automatic driving simulator map according to the adversarial vehicle relative position and the ego vehicle position; constructing a behavior topology network according to the interaction behavior library, and generating a plurality of interactable dynamic behaviors using the behavior topology network; constructing a plurality of comprehensive test scenes according to the weather parameters, the plurality of ego vehicle generated positions, the plurality of adversarial vehicle static positions, and the plurality of interactable dynamic behaviors.

8. An electronic device, comprising: comprise: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the automatic driving interactable test scene generation method of claim 7.

9. A computer program product, characterised in that, The computer program / instructions, when executed by the processor, implement the automatic driving interactable test scene generation method of claim 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the automatic driving interactable test scene generation method of claim 7.

Images