Method and device for building automatic driving simulation scene, storage medium and equipment

Abstract

The invention discloses a building method and device of an automatic driving simulation scene, a storage medium and equipment, and belongs to the technical field of automatic driving. Modeling real driving data recorded by a real vehicle to obtain metadata of each real object in a real driving environment, the metadata including point cloud data, category information and attitude information of the real object; object information of a virtual object is obtained in the built automatic driving simulation scene, the virtual object is an obstacle object encountered by the to-be-simulated automatic driving vehicle, and the object information comprises category information and attitude information of the virtual object; metadata matched with the object information is searched for; and in the automatic driving simulation scene, replacing the simulation data of the virtual object with the matched point cloud data in the metadata. According to the invention, by introducing real data, the complexity of a real driving environment is better reflected in simulation, especially the real driving environment under an extreme condition is simulated, and the cost and risk in an actual road test are reduced.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and in particular to a method, device, storage medium and equipment for building an autonomous driving simulation scene. Background Art

[0002] In the process of autonomous driving research and development, the accuracy of sensor data and the fidelity of simulation effects are crucial. Existing autonomous driving simulation systems (such as Carla) are widely used in research and development and testing, providing a variety of scenarios and sensor feedback through virtual environments. The simulation platform provides an efficient testing environment for algorithm development and autonomous driving system verification, but it often lacks effective integration with real environment data, resulting in deviations between simulation results and actual performance in some cases. Direct road testing is very risky, costly, and inefficient.

[0003] Current autonomous driving simulation platforms mainly rely on simulating the vehicle and surrounding sensor data in a fully virtual environment. However, due to the simplification of simulation scenarios and modeling limitations, they cannot fully reflect the complex real driving environment, especially in edge cases such as extreme weather or rare traffic conditions. In addition, virtual sensor data is usually idealized and cannot fully reflect the various interferences or data noises faced by real sensors. Summary of the invention

[0004] The present application provides a method, device, storage medium and equipment for building an autonomous driving simulation scene, which is used to solve the problem that when using virtual sensor data for simulation, it is impossible to reflect the complex real driving environment, various interferences and data noise. The technical solution is as follows:

[0005] According to a first aspect of the present application, a method for building an autonomous driving simulation scenario is provided, which is used in an autonomous driving simulation platform, and the method includes:

[0006] Modeling real driving data recorded by a real vehicle to obtain metadata of each real object in a real driving environment, wherein the metadata includes point cloud data, category information, and posture information of the real object;

[0007] Acquire object information of a virtual object in the constructed autonomous driving simulation scene, where the virtual object is an obstacle object encountered by the autonomous driving vehicle to be simulated, and the object information includes category information and posture information of the virtual object;

[0008] Finding metadata matching the object information;

[0009] If metadata matching the object information is found, then in the autonomous driving simulation scene, the simulation data of the virtual object is replaced with the point cloud data in the matching metadata.

[0010] In a possible implementation, searching for metadata matching the object information includes:

[0011] Determining the category of the virtual object according to the category information in the object information;

[0012] Determining a sensor posture of the autonomous driving vehicle and a vehicle posture of the virtual object according to the posture information in the object information;

[0013] Metadata whose category information matches the category of the virtual object and whose posture information matches the vehicle posture and the sensor posture is searched.

[0014] In a possible implementation, the sensor posture includes position coordinates and Euler angles, and the sensor posture is used to represent the position and viewing angle of the autonomous driving vehicle.

[0015] In a possible implementation, the method further includes:

[0016] Acquire traffic light information in the autonomous driving simulation scene, wherein the traffic light information includes category information, color information and posture information of the traffic light;

[0017] Searching for metadata matching the traffic light information, the metadata including category information, color information of the traffic light, posture information, and a traffic light image;

[0018] In the autonomous driving simulation scenario, the simulated image of the traffic light is replaced with the traffic light image in the matched metadata.

[0019] In a possible implementation, the method further includes:

[0020] Obtaining a simulation trajectory of a virtual object in the autonomous driving simulation scene;

[0021] searching for metadata matching the simulated trajectory, the metadata including a real trajectory;

[0022] In the autonomous driving simulation scenario, the simulated trajectory is replaced with the real trajectory in the matched metadata.

[0023] In a possible implementation, the method further includes:

[0024] If metadata matching the object information is not found, the virtual object is simulated using simulation data of the virtual object in the autonomous driving simulation scenario.

[0025] In a possible implementation, the step of obtaining object information of a virtual object in the constructed autonomous driving simulation scene includes:

[0026] Get the configuration information of the built autonomous driving simulation scene;

[0027] Determining a virtual object to be replaced with a real object from the configuration information;

[0028] Obtain object information of the virtual object.

[0029] According to a second aspect of the present application, a device for building an autonomous driving simulation scene is provided, which is used in an autonomous driving simulation platform, and the device includes:

[0030] A modeling module is used to model the real driving data recorded by the real car to obtain metadata of each real object in the real driving environment, wherein the metadata includes point cloud data, category information and posture information of the real object;

[0031] An acquisition module, used to acquire object information of a virtual object in the constructed autonomous driving simulation scene, wherein the virtual object is an obstacle object encountered by the autonomous driving vehicle to be simulated, and the object information includes category information and posture information of the virtual object;

[0032] A search module, used to search for metadata matching the object information;

[0033] A building module is used to replace the simulation data of the virtual object with the point cloud data in the matching metadata in the autonomous driving simulation scene if metadata matching the object information is found.

[0034] According to a third aspect of the present application, a computer-readable storage medium is provided, in which at least one instruction is stored. The at least one instruction is loaded and executed by a processor to implement the method for building an autonomous driving simulation scene as described above.

[0035] According to a fourth aspect of the present application, a computer device is provided, which includes a device for building the above-mentioned autonomous driving simulation scene.

[0036] The beneficial effects of the technical solution provided by this application include at least:

[0037] By extracting metadata of real objects from real driving data recorded by real vehicles, in autonomous driving simulation scenarios, the simulation data of virtual objects is replaced with point cloud data in the matching metadata. In this way, by introducing real data, the complexity of the real driving environment can be better reflected in the simulation, especially simulating the real driving environment under extreme conditions, reducing the cost and risk in actual road testing.

[0038] The metadata of real objects can reflect possible faults and noise in real sensors, as well as some unexpected driving behaviors that people may have under real road conditions. In this way, the performance of autonomous driving algorithms under more complex conditions can be tested, thereby improving the stability and comprehensiveness of the system.

[0039] In the autonomous driving simulation platform, both simulated data and real data can be used, ensuring that the optimal data source is used in the appropriate scenario. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0041] Figure 1 is a structural schematic diagram of an autonomous driving simulation platform according to some exemplary embodiments;

[0042] Figure 2 This is a flow chart of a method for building an autonomous driving simulation scenario provided by an embodiment of the present application;

[0043] Figure 3 This is a flow chart of a method for building an autonomous driving simulation scenario provided by an embodiment of the present application;

[0044] Figure 4 It is a structural block diagram of a device for building an autonomous driving simulation scene provided by an embodiment of the present application. DETAILED DESCRIPTION

[0045] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0046] Before autonomous vehicles are put on the road, they need to be simulated using an autonomous driving simulation platform in order to identify and solve problems during the simulation process and improve the safety of autonomous driving. There are many types of autonomous driving simulation platforms, such as the Carla simulation platform, the AirSim simulation platform, or the Gazebo simulation platform, which can simulate different vehicles, sensors, and road scenarios.

[0047] like Figure 1As shown in the figure, taking the Carla simulation platform as an example of the simulation system architecture, the upper layer of the Carla simulation platform is the scene editor of the application layer, and the lower layer is the big data layer and the data modeling and storage layer. The data modeling and storage layer includes data extraction operators and metadata storage modules. The data extraction operators are used to extract point cloud data, traffic light data, trajectories, images, etc. from big data, and store these data as metadata in the metadata storage module. The metadata module can be a MySQL database, a Redis database, an MQ database, etc. The big data layer can provide operations such as adding, deleting, modifying, and querying, display operations, etc., and import the metadata in the metadata storage module into the Carla simulation platform.

[0048] like Figure 2 As shown, it shows a method flow chart of a method for building an autonomous driving simulation scene provided by an embodiment of the present application. The method for building an autonomous driving simulation scene can be applied to an autonomous driving simulation platform in a computer device. The method for building an autonomous driving simulation scene can include:

[0049] Step 201 , modeling the real driving data recorded by the real vehicle to obtain metadata of each real object in the real driving environment, where the metadata includes point cloud data, category information and posture information of the real object.

[0050] During the driving process, the real vehicle (real vehicle) will collect surrounding data through sensors to obtain real driving data. These real driving data can be data of real objects encountered by the real vehicle during driving. Real objects can be obstacles, roads, traffic lights, signs, etc. Obstacles can be different types of vehicles, pedestrians, etc. Then, the real driving data can be processed using a pre-trained recognition model to obtain metadata for each real object. Taking the real object as an obstacle as an example, the metadata can include the point cloud data, category information and posture information of the obstacle. The category information indicates the category of the obstacle, which can be a vehicle model, and the posture information indicates the posture of the obstacle.

[0051] We can obtain a large amount of real driving data recorded by real vehicles, and thus obtain a large amount of metadata, which can meet different simulation needs.

[0052] Step 202, obtaining object information of a virtual object in the constructed autonomous driving simulation scene, where the virtual object is an obstacle object encountered by the autonomous driving vehicle to be simulated, and the object information includes category information and posture information of the virtual object.

[0053] In the autonomous driving simulation scenario, in addition to simulating driving environments such as roads, grass, and buildings, it is also necessary to simulate multiple moving or stationary virtual objects (also called obstacle objects) to test the obstacle avoidance algorithm of the autonomous driving vehicle to be simulated after encountering virtual objects. These virtual objects can be different types of vehicles and can be placed anywhere in the driving environment.

[0054] When simulating each virtual object, it is necessary to obtain object information of the virtual object, determine the category of the virtual object according to category information in the object information, and determine the posture of the virtual object according to posture information in the object information.

[0055] Step 203: Find metadata matching the object information.

[0056] Step 204: If metadata matching the object information is found, then in the autonomous driving simulation scene, the simulation data of the virtual object is replaced with the point cloud data in the matching metadata.

[0057] Specifically, point cloud data of real objects can be obtained from the matched metadata, and the point cloud data can be used to replace the simulation data of virtual objects, thereby realizing the use of real objects for simulation in autonomous driving simulation scenarios.

[0058] In this embodiment, the metadata after modeling has a high instance cutting accuracy. At the same time, the simulation data in the autonomous driving simulation platform is also high-precision data. Therefore, high-precision data fusion and synchronization can be performed, thereby improving the simulation accuracy.

[0059] To sum up, the method for building an autonomous driving simulation scene provided in the embodiment of the present application extracts metadata of real objects from real driving data recorded by a real vehicle, and replaces the simulation data of virtual objects with point cloud data in the matched metadata in the autonomous driving simulation scene. In this way, by introducing real data, the complexity of the real driving environment can be better reflected in the simulation, especially simulating the real driving environment under extreme conditions, thereby reducing the cost and risk in actual road testing.

[0060] The metadata of real objects can reflect possible faults and noise in real sensors, as well as some unexpected driving behaviors that people may have under real road conditions. In this way, the performance of autonomous driving algorithms under more complex conditions can be tested, thereby improving the stability and comprehensiveness of the system.

[0061] like Figure 3 As shown, it shows a flow chart of a method for building an autonomous driving simulation scene provided by an embodiment of the present application, and the method for building an autonomous driving simulation scene can be applied to an autonomous driving simulation platform in a computer device. The method for building an autonomous driving simulation scene may include:

[0062] Step 301 , modeling the real driving data recorded by the real vehicle to obtain metadata of each real object in the real driving environment, where the metadata includes point cloud data, category information and posture information of the real object.

[0063] The real vehicle (real vehicle) will collect surrounding data through sensors during driving to obtain real driving data. These real driving data can be data of real objects encountered by the real vehicle during driving. Real objects can be obstacles, roads, traffic lights, signs, etc. Obstacles can be different types of vehicles, pedestrians, etc. Then, the real driving data can be processed using the pre-trained recognition model to obtain metadata of each real object.

[0064] Taking the real object as an obstacle as an example, the metadata may include point cloud data, category information and posture information of the obstacle. The category information indicates the category of the obstacle, which may be a vehicle model, and the posture information indicates the posture of the obstacle.

[0065] Taking the real object as a traffic light as an example, the metadata may include a traffic light image, category information, color information and posture information of the traffic light. The category information indicates the category of the traffic light, which may be a temporary traffic light, a fixed traffic light, etc. The color information may be red, yellow, or green. The posture information indicates the posture of the traffic light.

[0066] Taking the trajectory of a real object as an example, the metadata may include category information, map information, speed, acceleration, direction, time and other information.

[0067] We can obtain a large amount of real driving data recorded by real vehicles, and thus obtain a large amount of metadata, which can meet different simulation needs.

[0068] Step 302, obtaining object information of a virtual object in the constructed autonomous driving simulation scene, where the virtual object is an obstacle object encountered by the autonomous driving vehicle to be simulated, and the object information includes category information and posture information of the virtual object.

[0069] In the autonomous driving simulation scenario, in addition to simulating driving environments such as roads, grass, and buildings, it is also necessary to simulate multiple moving or stationary virtual objects (also called obstacle objects) to test the obstacle avoidance algorithm of the autonomous driving vehicle to be simulated after encountering virtual objects. These virtual objects can be different types of vehicles and can be placed anywhere in the driving environment.

[0070] In the autonomous driving simulation platform, you can configure whether to use simulated data or real data to simulate virtual objects as needed, thereby ensuring that the optimal data source is used in appropriate scenarios.

[0071] Specifically, obtaining object information of a virtual object in a constructed autonomous driving simulation scene may include: obtaining configuration information of the constructed autonomous driving simulation scene; determining a virtual object to be replaced by a real object from the configuration information; and obtaining object information of the virtual object.

[0072] For example, virtual objects that need to use real data can be labeled, or a list can be used to record the IDs of virtual objects that need to use real data, or other methods can be used to mark virtual objects that need to use real data, which is not limited in this embodiment.

[0073] Step 303: Determine the category of the virtual object according to the category information in the object information.

[0074] The category of the virtual object can represent a vehicle type, such as a sedan, a sport utility vehicle (SUV), a van, a truck, a tram, and the like.

[0075] Step 304, determining the sensor posture of the autonomous driving vehicle and the vehicle posture of the virtual object based on the posture information in the object information.

[0076] The sensor attitude includes position coordinates and Euler angles, and the sensor attitude is used to represent the position and viewing angle of the autonomous vehicle. For example, the sensor attitude can be used to know that there is a car two meters to the left of the autonomous vehicle.

[0077] The vehicle posture can represent the posture of the virtual object, such as the direction of the virtual object.

[0078] Step 305 : searching for metadata whose category information matches the category of the virtual object and whose posture information matches the vehicle posture and the sensor posture.

[0079] For each piece of metadata, it is possible to detect whether the category information in the metadata matches the category of the virtual object, and to detect whether the posture information in the metadata matches the vehicle posture and the sensor posture; if each item matches, it is determined that the metadata matches the object information; if at least one of them does not match, it is determined that the metadata does not match the object information, and continue to detect whether the next piece of metadata matches the virtual object.

[0080] Step 306: If metadata matching the object information is found, then in the autonomous driving simulation scene, the simulation data of the virtual object is replaced with the point cloud data in the matching metadata.

[0081] Specifically, point cloud data of real objects can be obtained from the matched metadata, and the point cloud data can be used to replace the simulation data of virtual objects, thereby realizing the use of real objects for simulation in autonomous driving simulation scenarios.

[0082] Step 307: If metadata matching the object information is not found, the virtual object is simulated using the simulation data of the virtual object in the autonomous driving simulation scenario.

[0083] Specifically, simulation data of virtual objects can be obtained, and the simulation data can be used to perform simulation in an autonomous driving simulation scenario.

[0084] In one embodiment, if it is necessary to simulate a traffic light in an autonomous driving simulation scenario, the traffic light information in the autonomous driving simulation scenario can be obtained, the traffic light information including category information, color information and posture information of the traffic light; metadata matching the traffic light information is searched, the metadata including category information, color information, posture information and traffic light image; in the autonomous driving simulation scenario, the simulated image of the traffic light is replaced with the traffic light image in the matched metadata.

[0085] The attitude information indicates the attitude of the sensor of the autonomous vehicle. The sensor attitude includes position coordinates and Euler angles, and the sensor attitude is used to indicate the position and viewing angle of the autonomous vehicle. For example, the sensor attitude can be used to indicate that there is a traffic light three meters in front of the autonomous vehicle.

[0086] In addition to using traffic light images collected from a real vehicle to simulate traffic lights, traffic light images can also be directly generated according to the traffic light colors required in the autonomous driving simulation scenario, or other methods can be used to simulate traffic lights, which are not limited in this embodiment.

[0087] In one embodiment, if it is necessary to simulate a trajectory in an autonomous driving simulation scene, the simulated trajectory of the virtual object in the autonomous driving simulation scene can be obtained; metadata matching the simulated trajectory is searched, and the metadata includes a real trajectory; in the autonomous driving simulation scene, the simulated trajectory is replaced with the real trajectory in the matched metadata.

[0088] In this embodiment, the metadata after modeling has a high instance cutting accuracy. At the same time, the simulation data in the autonomous driving simulation platform is also high-precision data. Therefore, high-precision data fusion and synchronization can be performed, thereby improving the simulation accuracy.

[0089] To sum up, the method for building an autonomous driving simulation scene provided in the embodiment of the present application extracts metadata of real objects from real driving data recorded by a real vehicle, and replaces the simulation data of virtual objects with point cloud data in the matched metadata in the autonomous driving simulation scene. In this way, by introducing real data, the complexity of the real driving environment can be better reflected in the simulation, especially simulating the real driving environment under extreme conditions, thereby reducing the cost and risk in actual road testing.

[0090] The metadata of real objects can reflect possible faults and noise in real sensors, as well as some unexpected driving behaviors that people may have under real road conditions. In this way, the performance of autonomous driving algorithms under more complex conditions can be tested, thereby improving the stability and comprehensiveness of the system.

[0091] In the autonomous driving simulation platform, both simulated data and real data can be used, ensuring that the optimal data source is used in the appropriate scenario.

[0092] like Figure 4 As shown, it shows a structural block diagram of a device for building an autonomous driving simulation scene provided by an embodiment of the present application, and the device for building an autonomous driving simulation scene can be applied to an autonomous driving simulation platform in a computer device. The device for building an autonomous driving simulation scene may include:

[0093] A modeling module 410 is used to model the real driving data recorded by the real vehicle to obtain metadata of each real object in the real driving environment, wherein the metadata includes point cloud data, category information and posture information of the real object;

[0094] An acquisition module 420 is used to acquire object information of a virtual object in the constructed autonomous driving simulation scene, where the virtual object is an obstacle object encountered by the autonomous driving vehicle to be simulated, and the object information includes category information and posture information of the virtual object;

[0095] A search module 430, configured to search for metadata matching the object information;

[0096] The construction module 440 is used to replace the simulation data of the virtual object with the point cloud data in the matching metadata in the autonomous driving simulation scene if metadata matching the object information is found.

[0097] In an optional embodiment, the search module 430 is further configured to:

[0098] Determining the category of the virtual object according to the category information in the object information;

[0099] Determining a sensor posture of the autonomous driving vehicle and a vehicle posture of the virtual object according to the posture information in the object information;

[0100] Metadata whose category information matches the category of the virtual object and whose posture information matches the vehicle posture and the sensor posture is searched.

[0101] In an optional embodiment, the sensor posture includes position coordinates and Euler angles, and the sensor posture is used to represent the position and viewing angle of the autonomous driving vehicle.

[0102] In an optional embodiment, the acquisition module 420 is further used to acquire traffic light information in the autonomous driving simulation scene, where the traffic light information includes category information, color information and posture information of the traffic light;

[0103] The search module 430 is further used to search for metadata matching the traffic light information, wherein the metadata includes category information, color information of the traffic light, posture information and a traffic light image;

[0104] The construction module 440 is also used to replace the simulated image of the traffic light with the traffic light image in the matched metadata in the autonomous driving simulation scene.

[0105] In an optional embodiment, the acquisition module 420 is further used to acquire a simulation trajectory of a virtual object in the autonomous driving simulation scene;

[0106] The search module 430 is further used to search for metadata matching the simulation trajectory, wherein the metadata includes a real trajectory;

[0107] The construction module 440 is also used to replace the simulated trajectory with the real trajectory in the matched metadata in the autonomous driving simulation scenario.

[0108] In an optional embodiment, the building module 440 is also used to simulate the virtual object using the simulation data of the virtual object in the autonomous driving simulation scene if metadata matching the object information is not found.

[0109] In an optional embodiment, the acquisition module 420 is further configured to:

[0110] Get the configuration information of the built autonomous driving simulation scene;

[0111] Determining a virtual object to be replaced with a real object from the configuration information;

[0112] Obtain object information of the virtual object.

[0113] To sum up, the device for building an autonomous driving simulation scene provided in the embodiment of the present application extracts metadata of real objects from real driving data recorded by a real vehicle, and replaces the simulation data of virtual objects with point cloud data in the matched metadata in the autonomous driving simulation scene. In this way, by introducing real data, the complexity of the real driving environment can be better reflected in the simulation, especially simulating the real driving environment under extreme conditions, thereby reducing the cost and risk in actual road testing.

[0114] The metadata of real objects can reflect possible faults and noise in real sensors, as well as some unexpected driving behaviors that people may have under real road conditions. In this way, the performance of autonomous driving algorithms under more complex conditions can be tested, thereby improving the stability and comprehensiveness of the system.

[0115] In the autonomous driving simulation platform, both simulated data and real data can be used, ensuring that the optimal data source is used in the appropriate scenario.

[0116] An embodiment of the present application provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is loaded and executed by a processor to implement the method for building an autonomous driving simulation scene as described above.

[0117] An embodiment of the present application provides a computer device, which includes a device for building any of the above-mentioned autonomous driving simulation scenarios.

[0118] It should be noted that: the device for building an autonomous driving simulation scene provided in the above embodiment only uses the division of the above functional modules as an example when simulating an autonomous driving simulation scene. In actual applications, the above functional distribution can be completed by different functional modules as needed, that is, the internal structure of the device for building an autonomous driving simulation scene is divided into different functional modules to complete all or part of the functions described above. In addition, the device for building an autonomous driving simulation scene provided in the above embodiment and the method for building an autonomous driving simulation scene belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0119] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0120] The above description is not intended to limit the embodiments of the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the protection scope of the embodiments of the present application.

Claims

1. A method for building an autonomous driving simulation scene, characterized in that: Used in an autonomous driving simulation platform, the method includes: Modeling real driving data recorded by a real vehicle to obtain metadata of each real object in a real driving environment, wherein the metadata includes point cloud data, category information, and posture information of the real object; Acquire object information of a virtual object in the constructed autonomous driving simulation scene, where the virtual object is an obstacle object encountered by the autonomous driving vehicle to be simulated, and the object information includes category information and posture information of the virtual object; Finding metadata matching the object information; If metadata matching the object information is found, then in the autonomous driving simulation scene, the simulation data of the virtual object is replaced with the point cloud data in the matching metadata.

2. The method for constructing an autonomous driving simulation scene according to claim 1, characterized in that: The searching for metadata matching the object information includes: Determining the category of the virtual object according to the category information in the object information; Determining a sensor posture of the autonomous driving vehicle and a vehicle posture of the virtual object according to the posture information in the object information; Metadata whose category information matches the category of the virtual object and whose posture information matches the vehicle posture and the sensor posture is searched.

3. The method for constructing an autonomous driving simulation scene according to claim 2, characterized in that: The sensor posture includes position coordinates and Euler angles, and the sensor posture is used to represent the position and viewing angle of the autonomous driving vehicle.

4. The method for constructing an autonomous driving simulation scene according to claim 1, characterized in that: The method further comprises: Acquire traffic light information in the autonomous driving simulation scene, wherein the traffic light information includes category information, color information and posture information of the traffic light; Searching for metadata matching the traffic light information, the metadata including category information, color information of the traffic light, posture information, and a traffic light image; In the autonomous driving simulation scenario, the simulated image of the traffic light is replaced with the traffic light image in the matched metadata.

5. The method for constructing an autonomous driving simulation scene according to claim 1, characterized in that: The method further comprises: Obtaining a simulation trajectory of a virtual object in the autonomous driving simulation scene; searching for metadata matching the simulated trajectory, the metadata including a real trajectory; In the autonomous driving simulation scenario, the simulated trajectory is replaced with the real trajectory in the matched metadata.

6. The method for constructing an autonomous driving simulation scene according to claim 1, characterized in that: The method further comprises: If metadata matching the object information is not found, the virtual object is simulated using simulation data of the virtual object in the autonomous driving simulation scenario.

7. The method for constructing an autonomous driving simulation scene according to any one of claims 1 to 6, characterized in that: The step of obtaining object information of a virtual object in the constructed autonomous driving simulation scene includes: Get the configuration information of the built autonomous driving simulation scene; Determining a virtual object to be replaced with a real object from the configuration information; Obtain object information of the virtual object.

8. A device for building an autonomous driving simulation scene, characterized in that: Used in an autonomous driving simulation platform, the device comprises: A modeling module is used to model the real driving data recorded by the real car to obtain metadata of each real object in the real driving environment, wherein the metadata includes point cloud data, category information and posture information of the real object; An acquisition module, used to acquire object information of a virtual object in the constructed autonomous driving simulation scene, wherein the virtual object is an obstacle object encountered by the autonomous driving vehicle to be simulated, and the object information includes category information and posture information of the virtual object; A search module, used to search for metadata matching the object information; A building module is used to replace the simulation data of the virtual object with the point cloud data in the matching metadata in the autonomous driving simulation scene if metadata matching the object information is found.

9. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the method for building an autonomous driving simulation scene as described in any one of claims 1 to 7.

10. A computer device, characterized in that: The computer device includes: the device for building the autonomous driving simulation scene as described in claim 8.

Images