Virtual test scene model generation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention relates to a virtual test scene model generation method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining scene data collected from a preset driving scene; performing feature analysis on the scene data to obtain scene description information representing features of the preset driving scene; inputting the scene description information into a pre-trained modeling language generation model to obtain modeling description information; and converting the modeling description information by using a preset test scene modeling tool to obtain a virtual test scene model. According to the embodiment of the invention, the virtual scene model can be automatically generated, the construction of the scene model does not need to be manually intervened, and the simulation effect of the constructed scene model is more real, so that the efficiency of constructing the virtual test scene model is improved, and the efficiency of testing the driving system is further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent vehicles, and in particular to a virtual test scene model generation method and device, electronic equipment and a storage medium. BACKGROUND

[0002] In the research and development process of an autonomous driving system, a comprehensive and rigorous test process is the core to ensure its safe and reliable operation in a complex and variable traffic environment. As autonomous driving technology advances to a higher level, the accuracy and comprehensiveness of testing requirements also increase day by day. At present, the main testing methods rely on pre-constructed fixed scene libraries, which usually cover common traffic conditions, such as intersection traffic and vehicle merging, aiming to provide basic test scenes for autonomous driving systems. Alternatively, test cases are manually designed by test personnel based on their own experience, who imagine various possible situations, from weather factors to road conditions, and construct corresponding test scenes one by one. However, these traditional scene coverage methods have limitations and low testing efficiency. Therefore, how to improve the authenticity of simulation test scenes and improve testing efficiency is a problem to be solved. SUMMARY

[0003] In view of this, to solve the above-mentioned part or all technical problems, the embodiments of the present application provide a virtual test scene model generation method, device, electronic equipment and storage medium.

[0004] In a first aspect, the embodiments of the present application provide a virtual test scene model generation method, which comprises: acquiring scene data collected from a preset driving scene; performing feature analysis on the scene data to obtain scene description information representing the features of the preset driving scene; inputting the scene description information into a pre-trained modeling language generation model to obtain modeling description information; and converting the modeling description information using a preset test scene modeling tool to obtain a virtual test scene model.

[0005] In one possible implementation, inputting the scene description information into the pre-trained modeling language generation model to obtain the modeling description information comprises: inputting the scene description information into the pre-trained modeling language generation model to obtain map modeling description information for generating a simulation map and scene modeling description information for generating a simulation dynamic scene, and taking the map modeling description information and the scene modeling description information as the modeling description information.

[0006] In a possible implementation, the modeling description information is converted by using a preset test scene modeling tool to obtain a virtual test scene model, including: the map modeling description information is converted by using a preset map modeling tool to obtain a simulation map model; and the scene modeling description information is converted by using a preset dynamic scene modeling tool to obtain a dynamic scene model.

[0007] In a possible implementation, after the modeling description information is converted by using a preset test scene modeling tool to obtain a virtual test scene model, the method further includes: testing the preset driving system based on the virtual test scene model to obtain a test result; if the test result indicates that the test fails, generating feedback information for indicating a reason for the test failure based on a gap between the test result and a preset expected result; and adjusting the scene description information based on the feedback information to obtain updated scene description information.

[0008] In a possible implementation, the feedback information for indicating the reason for the test failure is generated based on a gap between the test result and a preset expected result, including: scene defect analysis is performed on test data obtained after the test by using a preset test data analysis script to obtain the feedback information, wherein the scene defect includes at least one of the following: an autonomous driving decision algorithm defect, a sensor misjudgment defect, and a vehicle control strategy defect.

[0009] In a possible implementation, the preset driving system is tested based on the virtual test scene model to obtain a test result, including: test data obtained after the driving system is tested under the virtual test scene model is acquired; evaluation index quantization processing is performed on the test data to obtain at least one evaluation index data, wherein the at least one evaluation index data includes at least one of the following: safety index data, reliability index data, comfort index data, and driving efficiency data; and the test result indicating whether the driving system passes the test is generated based on the at least one evaluation index data.

[0010] In a possible implementation, the scene data is feature-analyzed to obtain scene description information indicating features of the preset driving scene, including: feature extraction and semantic extraction are performed on the scene data to obtain scene feature data and semantic description text indicating features of the preset driving scene, and the scene feature data and the semantic description text are taken as the scene description information.

[0011] In a second aspect, an embodiment of the present application provides a virtual test scene model generation apparatus, which comprises: an acquisition module configured to acquire scene data collected from a preset driving scene; an analysis module configured to perform feature analysis on the scene data to obtain scene description information representing features of the preset driving scene; a first generation module configured to input the scene description information into a pre-trained modeling language generation model to obtain modeling description information; and a second generation module configured to convert the modeling description information using a preset test scene modeling tool to obtain a virtual test scene model.

[0012] In a third aspect, an embodiment of the present application provides an electronic device, which comprises: a memory configured to store a computer program; and a processor configured to execute the computer program stored in the memory, and when the computer program is executed, the method of any one of the embodiments of the virtual test scene model generation method of the first aspect of the present application is implemented.

[0013] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method of any one of the embodiments of the virtual test scene model generation method of the first aspect described above is implemented.

[0014] In a fifth aspect, an embodiment of the present application provides a computer program, which comprises computer readable code, and when the computer readable code is run on a device, the processor in the device implements the method of any one of the embodiments of the virtual test scene model generation method of the first aspect described above.

[0015] The virtual test scene model generation method, apparatus, electronic device and storage medium provided by the embodiments of the present application, by acquiring scene data collected from a preset driving scene, performing feature analysis on the scene data to obtain scene description information representing features of the preset driving scene, inputting the scene description information into a pre-trained modeling language generation model to obtain modeling description information, and converting the modeling description information using a preset test scene modeling tool to obtain a virtual test scene model, the embodiments of the present application automatically generate modeling description information for scene modeling based on data collected in an actual driving scene by using an artificial intelligence method, automatically generate a virtual scene model by using the modeling description information, do not need human intervention in the construction of the scene model, and the simulation effect of the constructed scene model is more realistic, thereby improving the efficiency of constructing a virtual test scene model and further improving the efficiency of testing a driving system. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings, which are incorporated herein and form part of the specification, illustrate embodiments consistent with the present application and, together with the description, further serve to explain the principles of the application.

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings are only a part of the embodiments of the present application, and thus, for those of ordinary skill in the art, other drawings can be obtained based on these drawings without any creative effort.

[0018] One or more embodiments are illustrated by way of example in the drawings that are not intended to be limiting of the present application, and the same or similar reference numerals designate similar components throughout the drawings and specification. The drawings are not necessarily to scale, the emphasis instead being placed upon illustrating the principles of the embodiments.

[0019] Figure 1 A flowchart of a method for generating a virtual test scene model according to an embodiment of the present application is shown in FIG. 1.

[0020] Figure 2 A flowchart of a method for generating a virtual test scene model according to an embodiment of the present application is shown in FIG. 1.

[0021] Figure 3 A flowchart of a method for generating a virtual test scene model according to an embodiment of the present application is shown in FIG. 1.

[0022] Figure 4 A flowchart of a method for generating a virtual test scene model according to an embodiment of the present application is shown in FIG. 1.

[0023] Figure 5 A flowchart of a method for generating a virtual test scene model according to an embodiment of the present application is shown in FIG. 1.

[0024] Figure 6 A flowchart of a method for generating a virtual test scene model according to an embodiment of the present application is shown in FIG. 1.

[0025] Figure 7 A flowchart of a method for generating a virtual test scene model according to an embodiment of the present application is shown in FIG. 1. DETAILED DESCRIPTION

[0026] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. As apparent from the following description, the embodiments described are only part of the embodiments of the present application, and thus do not limit the scope of the present application. Unless otherwise defined, technical terms used in the embodiments of the present application have the same meanings as those that are well known and used commonly in the art.

[0027] Those skilled in the art can understand that the terms "first", "second", etc. in the embodiments of the present application are only used to distinguish different steps, devices or modules, and do not represent any specific technical meaning, nor represent a logical sequence between them.

[0028] It should also be understood that in the present embodiment, "a plurality of" can refer to two or more, and "at least one" can refer to one, two or more.

[0029] It should also be understood that for any component, data or structure mentioned in the embodiments of the present application, one or more can be generally understood without explicit limitation or in the context of the preceding and following.

[0030] In addition, the term "and / or" in the present application is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can mean that A exists alone, A and B exist together, and B exists alone. In addition, the character " / " in the present application generally represents an "or" relationship between the front and rear associated objects.

[0031] It should also be understood that the description of the embodiments of the present application emphasizes the differences between the embodiments, and the same or similar parts can be referred to each other, and for the sake of brevity, will not be repeated.

[0032] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the application or its application or use.

[0033] The techniques, methods and devices known to those skilled in the relevant art can not be discussed in detail, but in appropriate cases, the above-mentioned techniques, methods and devices should be regarded as part of the specification.

[0034] It should be noted that similar reference numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0035] It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict. In order to understand the embodiments of the present application, the following will be described in detail with reference to the drawings and in combination with the embodiments. Obviously, the described embodiments are part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0036] In order to solve the technical problem that the construction of virtual test scene model in the prior art needs manual intervention, resulting in low test efficiency, the present application provides a method for generating a virtual test scene model, which can automatically generate a virtual test scene model based on the data collected in the actual driving scene by using artificial intelligence method, and the efficiency of model construction and test is higher.

[0037] Figure 1A flowchart of a method for generating a virtual test scene model is provided. The method can be applied to a scenario of virtual testing of a vehicle driving system, and is executed by various electronic devices. For example, the method can be executed by a host on a test bench. In addition, the execution subject of the method can be hardware or software. When the execution subject is hardware, the execution subject can be one or more of the electronic devices. For example, a single electronic device can execute the method, or multiple electronic devices can cooperate with each other to execute the method. When the execution subject is software, the method can be implemented as multiple software or software modules, or as a single software or software module. No specific limitation is made herein.

[0038] As shown in Figure 1 , the method specifically includes:

[0039] Step 101: Obtain scene data collected from a preset driving scene.

[0040] In some embodiments, the preset driving scene can include one or more of, for example, scenes of different traffic flow conditions such as urban trunk roads, highways, and rural roads; fixed scenes of various weather conditions (such as sunny days, cloudy days, rainy days); standard scenes required for regulatory testing; and special scene materials that may be involved in long-distance generalization testing.

[0041] Generally, in order to improve the pertinence of testing, data under a failed test scene can also be obtained. For example, a scene in which the autonomous driving system collides when turning at an intersection, a scene in which a decision-making error occurs during a merging process, and the like.

[0042] Under various scenes, various scene data can be obtained through sensors, navigation positioning, the Internet, and the like. For example, vehicle speed, position, driving direction, and operation data of the vehicle control system, such as sensor output, decision-making instructions, and the like. Vehicle-mounted sensors, including cameras, ultrasonic radars, laser radars, and the like, can also be used to collect sensing data from actual roads as scene data.

[0043] After obtaining various data, the data can be preprocessed and labeled in a standardized manner to provide a data basis for subsequent scene generation and testing. For example, preprocessing can include filtering out invalid data, normalizing data, and the like. Labeling can include setting vehicle types, obstacle types, road lengths, lane widths, and the like on the road.

[0044] Step 102: Perform feature analysis on the scene data to obtain scene description information representing the features of the preset driving scene.

[0045] In some embodiments, the scene data can be analyzed in various ways. For example, the scene data includes road images captured by a camera, and a machine vision method can be used to extract features from the image data, including extracting features of road edges, lane lines, traffic signs, vehicles, pedestrians, and other targets. Key information such as vehicle speed, acceleration, and position can also be extracted by deeply mining the data collected by the sensors.

[0046] After analyzing the various information, the information can be formatted according to a predetermined data format to obtain scene description information. For example, the scene description information can be in the form of a feature vector.

[0047] In step 103, the scene description information is input into a pre-trained modeling language generation model to obtain modeling description information.

[0048] In some embodiments, the modeling language generation model can be trained based on a large language model (e.g., GPT). During training, the sample scene description information can be set with corresponding modeling description information as labeled information, and the model parameters can be updated through multiple iterations to obtain a trained modeling language generation model.

[0049] The above modeling description information is a modeling language that conforms to the syntax rules of the test scene modeling tool. For example, the test scene modeling tool can include OpenSCENARIO, which can be used to build a driving scene model, and the modeling description information can include road network labels, vehicle entity labels, and information about position, speed, orientation, and other attributes.

[0050] In step 104, the modeling description information is converted using a pre-set test scene modeling tool to obtain a virtual test scene model.

[0051] In some embodiments, after obtaining the modeling description information that conforms to the syntax rules of the test scene modeling tool, the modeling description information can be used as a parameter to build a virtual test scene model to generate a virtual test scene model.

[0052] The generated virtual test scene model can simulate a real driving scene, and the driving system can perform performance testing under the virtual test scene model.

[0053] The method for generating a virtual test scene model provided by the embodiments of the present application comprises the following steps: acquiring scene data collected from a preset driving scene, performing feature analysis on the scene data to obtain scene description information representing the features of the preset driving scene, inputting the scene description information into a pre-trained modeling language generation model to obtain modeling description information, and converting the modeling description information by using a preset test scene modeling tool to obtain a virtual test scene model. The embodiments realize automatic generation of modeling description information for scene modeling based on the data collected in an actual driving scene by using an artificial intelligence method, and automatically generate a virtual scene model by using the modeling description information, without manual intervention in the construction of the scene model. The simulation effect of the constructed scene model is more realistic, thereby improving the efficiency of constructing a virtual test scene model and further improving the efficiency of testing a driving system.

[0054] In some optional implementation manners of the embodiments, as shown in Figure 2 Step 103 comprises:

[0055] Step 1031, inputting the scene description information into a pre-trained modeling language generation model to obtain map modeling description information for generating a simulation map and scene modeling description information for generating a simulation dynamic scene, and taking the map modeling description information and the scene modeling description information as the modeling description information.

[0056] The map modeling description information can be information conforming to the syntax rules of a map modeling tool. For example, the map modeling tool can be a Houdini tool, and the required map modeling description information of the Houdini tool conforms to a corresponding language template. The language template focuses on node network construction. For example, a scene object is created by an / obj node, geometry shape data is processed by a geo node, and for a vehicle lane-changing scene, the vehicle motion path, speed change curve and other parameters are set in detail under the corresponding node.

[0057] The scene modeling description information can be information conforming to the syntax rules of a dynamic scene modeling tool. For example, the dynamic scene modeling tool can be OpenSCENARIO, and the corresponding language template of the OpenSCENARIO usually contains a scene initialization part, such as definition of a road network, and a scenario part, such as definition of a vehicle motion path. <roadnetwork>tags describing road type, number of lanes, etc.; vehicle entity part passes through <entities>under the label <vehicle>The sub-label defines the initial state of the vehicle, including attributes such as position, speed, orientation, etc.

[0058] The embodiment generates the map modeling description information and the scene modeling description information by using a modeling language to generate a model, and can automatically generate information required for constructing a simulation map and a dynamic scene without human intervention, thereby improving the efficiency of generating a virtual test scene.

[0059] In some optional implementation manners of the embodiment, as shown in Figure 2 Step 104 includes:

[0060] Step 1041, converting the map modeling description information by using a preset map modeling tool to obtain a simulation map model.

[0061] The map modeling tool is configured to generate the simulation map model by using parameters contained in the map modeling description information. The simulation map model can include map elements such as roads, terrains, and buildings. For example, the map modeling tool can be Houdini software, which can generate a simulation map model according to input map modeling description information conforming to its syntax rules. Houdini can generate a high-precision simulation map by using its three-dimensional modeling function according to road geometry information in the map modeling description information. For complex road structures such as three-dimensional intersections, Houdini can accurately model different height road layers, ramp connection relationships, and the like, and save the map in an OpenDRIVE format to provide an accurate geographic basis for subsequent scene building.

[0062] Step 1042, converting the scene modeling description information by using a preset dynamic scene modeling tool to obtain a dynamic scene model.

[0063] The dynamic scene modeling tool is configured to generate the dynamic scene model by using parameters contained in the scene modeling description information. The dynamic scene model can simulate a real vehicle driving scene. For example, the dynamic scene modeling tool can be OpenSCENARIO software, which can configure a generated scene in detail, set behavior rules of traffic participants, operation modes of traffic flow, and changes of environmental factors, and the like. For example, the driving speed range of a vehicle, the following distance, the lane changing rule, the crossing time and path of a pedestrian, and the like are specified. The virtual test scene model can be obtained by combining the dynamic scene model with the simulation map model. The virtual test scene model contains a map and rules of a vehicle driving scene on the map.

[0064] The embodiment generates a simulation map model and a dynamic scene model by using a map modeling tool and a dynamic scene modeling tool, and the two can be combined to obtain a virtual test scene model close to a real driving scene, thereby helping to more efficiently perform driving function testing in a virtual scene close to a real scene.

[0065] In some optional implementations of the embodiment, as shown in FIG. 1, after step 104, the method further includes: Figure 3

[0066] Step 105, testing the preset driving system based on the virtual test scene model to obtain a test result.

[0067] Specifically, the virtual test scene model can be imported into a preset simulation platform, and parameters of the driving system are imported into the simulation platform, so that the driving system can be tested.

[0068] For example, the CARLA simulation platform can be used, and the virtual test scene model and the automatic driving system model are synchronously loaded in the platform to start a full-process automatic testing task. Relying on the high-precision physical engine and multi-modal sensor simulation technology of the platform, the running state of the vehicle in a complex scene can be simulated in real time and dynamically, and full-link data of the automatic driving system can be synchronously collected, including key information such as environment perception original data, decision planning output results, and vehicle control instruction parameters, to build a precise and complete data support system for subsequent multi-dimensional performance evaluation.

[0069] Step 106, if the test result indicates that the test fails, feedback information indicating the reason for the test failure is generated based on the gap between the test result and the preset expected result.

[0070] Specifically, the test result can be evaluated by using preset evaluation indexes. If the gap between each data index included in the test result and the data index included in the expected result is greater than a preset threshold, it is determined that the test fails. The test result can include statistical data of multiple dimensions, and the data can be compared with the corresponding indexes to determine the reason for the test failure.

[0071] For example, if the test result includes data in an automatic parking scene, the vehicle fails to successfully park in a parking space under the operation of the automatic driving system in the simulation platform, and the planning path of the automatic driving system can be analyzed. If the planning path conflicts with the actual scene, for example, the size or position of the parking space in the virtual scene is not standardized, feedback information indicating that the size or position of the parking space causes the parking failure is generated.

[0072] Step 107, adjusting the scene description information based on the feedback information to obtain updated scene description information. ​

[0073] The feedback information can include numerical values, textual descriptions, and the like, which can be used as a reference for adjusting the scene description information. For example, in an automatic parking scenario, if the size or position of the parking space in the virtual scene is not standardized, the position and size of the parking space can be adjusted according to the feedback information, so as to obtain updated scene description information that meets the requirements of automatic parking. For another example, if a collision accident occurs in the simulation test, the speed of each vehicle in the virtual test scene can be adjusted, or the distance between surrounding vehicles can be changed, or the lane changing opportunity can be adjusted, and the like, to generate updated scene description information.

[0074] For the scene that fails the test, based on the feedback information, the updated scene description information can be generated by using parameter perturbation means, such as adjusting the vehicle speed, changing the traffic signal light period, performing structural deformation operations, such as adding road obstacles, changing the road topological structure, implementing logic extension strategies, such as adding sudden traffic events, introducing new traffic participants, and the like, to generate a series of variant scenes with similar characteristics but different parameters and structures.

[0075] The modeling language generation model can automatically generate more targeted variant scenes based on the updated scene description information, and the test is performed again. This cycle is repeated to form an efficient closed-loop test process until the driving system meets the expected high performance standard in all test scenes.

[0076] The embodiment realizes feedback on the test process, and the updated scene description information can adjust the original virtual test scene to automatically generate variant scenes, so that the test process is more targeted, the test scene is more diversified, and the accuracy and efficiency of the driving system test are improved.

[0077] In some optional implementations of the embodiment, step 106 can be as follows:

[0078] The test data obtained after the test is analyzed by using a preset test data analysis script to obtain feedback information.

[0079] The scene defects include at least one of the following: automatic driving decision algorithm defects, sensor misjudgment defects, and vehicle control strategy defects.

[0080] The test data analysis script contains pre-written analysis code. Running this script automatically analyzes the test data. The aforementioned autonomous driving decision algorithm defects refer to defects in the execution of the autonomous driving decision algorithm that cause test failures. For example, in an automated parking scenario, the autonomous driving decision algorithm implements functions such as path planning. If the resulting path results in a parking failure, feedback information indicating parking failure is generated. This feedback information may include the coordinates of the path curve, the size and position of the parking space, as well as information such as the vehicle speed during parking and the distance to obstacles.

[0081] These sensor misjudgment defects occur when sensor-collected data is incorrect, leading to test failure. For example, when a vehicle is autonomously driving and encounters a road section, the collected LiDAR and image data may be inconsistent with the actual obstacle's location, size, and other attributes, or the detected distance to another vehicle may be inconsistent with the actual distance. In this case, various data at the time of the error can be recorded as feedback information.

[0082] Vehicle control strategy defects are defects in the vehicle's controlled state that lead to test failures. For example, if the vehicle's speed, steering angle, acceleration, etc., do not match the actual road conditions, leading to a collision, the vehicle's control parameters can be recorded as feedback information.

[0083] Based on the above feedback information, the original scene description information can be modified to avoid the above defects from happening again.

[0084] This embodiment performs scenario defect analysis on the test data obtained after the test to determine the cause of the test failure, so that more accurate scenario correction can be performed for the test failure scenario, thereby improving test efficiency.

[0085] In some optional implementations of this embodiment, such as Figure 4 As shown, step 105 includes:

[0086] Step 1051 , obtaining test data obtained after testing the driving system under the virtual test scenario model.

[0087] Step 1052: perform evaluation index quantification processing on the test data to obtain at least one evaluation index data.

[0088] The at least one evaluation index data includes at least one of the following: safety index data, reliability index data, comfort index data and driving efficiency data.

[0089] The safety index data can include collision times, violation times, and the like; the reliability index data can include system failure times, decision error rates, and the like; the comfort index data can include sudden acceleration times, sudden braking times, driving smoothness evaluation data, and the like; and the driving efficiency data can include route completion times, task success rates, and the like.

[0090] At step 1053, a test result indicating whether the driving system passes the test is generated based on the at least one evaluation index data.

[0091] Specifically, the weights of the above-mentioned index data can be set, and a test result that can evaluate the performance of the driving system can be obtained through weighted calculation.

[0092] The embodiment can comprehensively and objectively evaluate the performance of the driving system in the virtual test scene through the evaluation index quantization processing on the test data, so that the driving system test can be more accurately performed.

[0093] In some optional implementations of the embodiment, step 102 can be performed as follows:

[0094] The scene data is subjected to feature extraction and semantic extraction to obtain scene feature data representing features of the preset driving scene and a semantic description text, and the scene feature data and the semantic description text are taken as scene description information.

[0095] The scene feature data can represent various features of the preset driving scene, and is usually in the form of a feature vector. For example, the scene feature data can include road geometry data (such as the number of lanes, the radius of a curve, and the like), traffic participant attributes (including vehicle types, pedestrian behavior patterns, and the like), and environmental parameters (such as weather conditions, light intensity, and the like).

[0096] The text description of the scene is subjected to in-depth analysis by using an NLP (Natural Language Processing) technology to extract key semantic information, so as to obtain the semantic description text.

[0097] The scene feature data and the semantic description text can be combined to accurately describe the scene corresponding to the scene data. The scene feature data and the semantic description text are used as input data of a modeling language generation model, and the modeling language generation model can more accurately reason the scene, so as to obtain modeling description information with higher accuracy.

[0098] Reference Figure 5 In combination with the above embodiments, the flowchart for generating a virtual test scene model and performing simulation testing is provided. As shown in Figure 5 As shown, the whole process includes three parts, which are scene data collection, AI generated scene and simulation tool execution. The scene data includes various road data, functional use case scene data, test failure scene data, risk collision scene data, etc. From the above various data, scene description information is mined, the scene description information is input into the modeling language generation model constructed based on the AI large model, and the modeling description information is obtained. Then, the Houdini map modeling tool is used to generate a simulation map simulation map model; the OpenSCENARIO dynamic scene modeling tool is used to generate a dynamic scene model. The simulation map model and the dynamic scene model are imported into the simulation test platform (Carla), and the tested driving system is tested on the simulation test platform. If the test fails, the same type of generalized scene is generated according to the feedback information, and the test is performed again until the test passes.

[0099] Figure 6 A structural schematic diagram of a virtual test scene model generation device provided by an embodiment of the present application is provided. Specifically, it includes:

[0100] The acquisition module 601 is configured to acquire scene data collected from a preset driving scene.

[0101] The analysis module 602 is configured to analyze the features of the scene data to obtain scene description information representing the features of the preset driving scene.

[0102] The first generation module 603 is configured to input the scene description information into a pre-trained modeling language generation model to obtain modeling description information.

[0103] The second generation module 604 is configured to convert the modeling description information using a preset test scene modeling tool to obtain a virtual test scene model.

[0104] In one possible implementation, the first generation module is further configured to input the scene description information into the pre-trained modeling language generation model to obtain map modeling description information for generating a simulation map and scene modeling description information for generating a simulation dynamic scene, and to take the map modeling description information and the scene modeling description information as the modeling description information.

[0105] In one possible implementation, the second generation module includes a first generation unit configured to convert the map modeling description information using a preset map modeling tool to obtain a simulation map model, and a second generation unit configured to convert the scene modeling description information using a preset dynamic scene modeling tool to obtain a dynamic scene model.

[0106] In a possible implementation, the apparatus further includes: a test module, configured to test the preset driving system based on the virtual test scene model, and obtain a test result; a third generation module, configured to if the test result indicates that the test fails, generate feedback information for indicating a reason for the test failure based on a gap between the test result and a preset expected result; and an adjustment module, configured to adjust the scene description information based on the feedback information, and obtain updated scene description information.

[0107] In a possible implementation, the third generation module includes: a feedback unit, configured to perform scene defect analysis on test data obtained after the test by using a preset test data analysis script, and obtain the feedback information, wherein the scene defect includes at least one of the following: an autonomous driving decision algorithm defect, a sensor misjudgment defect, and a vehicle control strategy defect.

[0108] In a possible implementation, the test module includes: an acquisition unit, configured to acquire test data obtained after the driving system is tested under the virtual test scene model; a quantitative processing unit, configured to perform evaluation index quantitative processing on the test data, and obtain at least one evaluation index data, wherein the at least one evaluation index data includes at least one of the following: safety index data, reliability index data, comfort index data, and driving efficiency data; and a third generation unit, configured to generate a test result indicating whether the driving system passes the test based on the at least one evaluation index data.

[0109] In a possible implementation, the analysis module is further configured to: perform feature extraction and semantic extraction on the scene data, to obtain scene feature data representing features of the preset driving scene and a semantic description text, and use the scene feature data and the semantic description text as the scene description information.

[0110] The virtual test scene model generation apparatus provided in this embodiment can be a virtual test scene model generation apparatus as shown in Figure 6 , which can perform all steps of the above virtual test scene model generation method, and further achieve the technical effects of the above virtual test scene model generation method. For brevity, details are not described herein.

[0111] Figure 7 A structural schematic diagram of an electronic device provided in an embodiment of the present application, Figure 7 The electronic device 700 shown includes at least one processor 701, a memory 702, at least one network interface 704, and other user interfaces 703. The various components of the electronic device 700 are coupled together by a bus system 705, which can include a data bus, a power bus, a control bus, and a state signal bus. For the sake of clarity, the various buses are illustrated in FIG. 7 as the bus system 705. The bus system 705 is used for the exchange of control and status signals between the components and can be implemented as a distributed bus system, in which components exchange signals via a network, for example. Figure 7

[0112] The user interface 703 can include a display, a keyboard, or a pointing device (e.g., a mouse, a trackball, a touchpad, or a touchscreen).

[0113] It is to be understood that the memory 702 in the embodiments described herein can be volatile or nonvolatile memory, or can include both volatile and nonvolatile memory. By way of example, and not limitation, nonvolatile memory can be read-only memory (ROM), programmable ROM (PROM), erasable programmable ROM (EPROM), electrically EPROM (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), which acts as external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synch link DRAM (SLDRAM), and direct Rambus RAM (DRRAM). The memory 702 described herein is intended to include, without being limited to, these and any other suitable types of memory.

[0114] In some embodiments, the memory 702 stores elements, executable instructions, or data structures or a subset thereof, or an expanded set thereof, including an operating system 7021 and applications 7022.

[0115] ​The operating system 7021 includes various system programs, such as a framework layer, a core library layer, a driver layer, and the like, for implementing various basic services and processing hardware-based tasks. The application program 7022 includes various application programs, such as a media player (Media Player), a browser (Browser), and the like, for implementing various application services. The program implementing the method of the embodiments of the present application can be included in the application program 7022.

[0116] In the present embodiment, by invoking the program or instruction stored in the memory 702, specifically, the program or instruction stored in the application program 7022, the processor 701 is configured to execute the method steps provided by the method embodiments, for example, including:

[0117] Obtaining scene data collected from a preset driving scene; performing feature analysis on the scene data to obtain scene description information representing features of the preset driving scene; inputting the scene description information into a pre-trained modeling language generation model to obtain modeling description information; and converting the modeling description information by using a preset test scene modeling tool to obtain a virtual test scene model.

[0118] The method disclosed in the embodiments of the present application can be applied to the processor 701 or implemented by the processor 701. The processor 701 can be an integrated circuit chip having a signal processing capability. In the implementation process, each step of the above method can be completed by an integrated logic circuit or an instruction in the form of software in the processor 701. The processor 701 described above can be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. Each method, step and logic block diagram disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software units in the code processor for execution. The software unit can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, or other mature storage media in the art. The storage medium is located in the memory 702, and the processor 701 reads the information in the memory 702 and combines the hardware to complete the steps of the above method.

[0119] It can be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing units can be implemented within one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSP Devices, DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), general purpose processors, controllers, micro-controllers, microprocessors, other electronic units designed to perform the functions described above, or a combination thereof.

[0120] For software implementation, the techniques described herein can be implemented with a processing unit executing program code stored in a storage medium. The program code stored in the storage medium can be executed by the processing unit.

[0121] The electronic device provided by the embodiments can be an electronic device as shown in Figure 7 The electronic device provided by the embodiments can be an electronic device as shown in

[0122] The embodiments of the present application further provide a storage medium (computer readable storage medium). The storage medium stores one or more programs. The storage medium can include a volatile memory, such as a random access memory, and / or can include a non-volatile memory, such as a read-only memory, a flash memory, a hard disk drive, or a solid-state drive. The storage medium can also include a combination of the above-mentioned memories.

[0123] When the one or more programs stored in the storage medium are executed by the one or more processors, the above-mentioned virtual test scene model generation method executed at the electronic device side can be implemented.

[0124] The above-mentioned processor is configured to execute the program stored in the memory, so as to implement the following steps of the virtual test scene model generation method executed at the electronic device side:

[0125] Acquire scene data collected from a preset driving scene; perform feature analysis on the scene data to obtain scene description information representing features of the preset driving scene; input the scene description information into a pre-trained modeling language generation model to obtain modeling description information; and convert the modeling description information by using a preset test scene modeling tool to obtain a virtual test scene model.

[0126] Those skilled in the art should further appreciate that the units and algorithm steps of each example described in connection with the embodiments disclosed herein can be embodied in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, various aspects have been described generally in terms of their functionality without loss of generality. Whether such functionality is implemented as hardware or software depends on the particular application and design constraints imposed on the overall architecture. Skilled persons can implement the described functionality in varying ways for each particular application, but such implementation does not cause a departure from the scope of the application.

[0127] The steps of a method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in random access memory (RAM), flash memory, read-only memory (ROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0128] It should be understood that the terms used herein are for the purpose of describing particular example embodiments and are not intended to be limiting. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. The terms "comprises", "comprising", "includes", "including" and "has" are to be construed as containing, comprising, including, or having, but to not exclude other features, steps, operations, elements and / or components. Method steps, processes, and operations described herein are not to be construed as necessarily requiring their performance in the particular order in which they are described unless specifically identified as an order dependent step. It is also to be understood that additional or alternative steps can be employed.

[0129] The above description is merely that of a specific implementation of the present application and as such is not to be taken in a limiting sense. Various modifications to the implementation will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other implementations without departing from the spirit or scope of the application. Accordingly, the application is not to be restricted based on the specific implementation described but is to be interpreted within the scope of the claims and their equivalents.< / vehicle> < / entities> < / roadnetwork>

Claims

1. A method for generating a virtual test scene model, characterized in that: The method comprises: Acquiring scene data collected from a preset driving scene; Performing feature analysis on the scene data to obtain scene description information representing features of the preset driving scene; Inputting the scene description information into a pre-trained modeling language generation model to obtain modeling description information; The modeling description information is converted using a preset test scenario modeling tool to obtain a virtual test scenario model.

2. The method according to claim 1, characterized in that The step of inputting the scene description information into a pre-trained modeling language generation model to obtain modeling description information includes: The scene description information is input into a pre-trained modeling language generation model to obtain map modeling description information for generating a simulation map and scene modeling description information for generating a simulation dynamic scene, and the map modeling description information and the scene modeling description information are used as the modeling description information.

3. The method according to claim 2, characterized in that The method of converting the modeling description information using a preset test scenario modeling tool to obtain a virtual test scenario model includes: Using a preset map modeling tool, converting the map modeling description information to obtain a simulation map model; The scene modeling description information is converted using a preset dynamic scene modeling tool to obtain a dynamic scene model.

4. The method according to claim 1, wherein After converting the modeling description information using a preset test scenario modeling tool to obtain a virtual test scenario model, the method further includes: Based on the virtual test scenario model, a preset driving system is tested to obtain a test result; If the test result indicates that the test fails, generating feedback information indicating the reason for the test failure based on the difference between the test result and the preset expected result; Based on the feedback information, the scene description information is adjusted to obtain updated scene description information.

5. The method according to claim 4, characterized in that The generating of feedback information indicating the reason for the test failure based on the gap between the test result and the preset expected result includes: Using a preset test data analysis script, the test data obtained after the test is subjected to scenario defect analysis to obtain the feedback information, wherein the scenario defect includes at least one of the following: an autonomous driving decision algorithm defect, a sensor misjudgment defect, and a vehicle control strategy defect.

6. The method according to claim 4, characterized in that The step of testing a preset driving system based on the virtual test scenario model to obtain a test result includes: Acquiring test data obtained after testing the driving system under the virtual test scenario model; Performing evaluation index quantification processing on the test data to obtain at least one evaluation index data, wherein the at least one evaluation index data includes at least one of the following: safety index data, reliability index data, comfort index data, and driving efficiency data; Based on the at least one evaluation index data, a test result indicating whether the driving system passes the test is generated.

7. The method according to any one of claims 1 to 6, characterized in that The feature analysis of the scene data to obtain scene description information representing the features of the preset driving scene includes: Feature extraction and semantic extraction are performed on the scene data to obtain scene feature data and semantic description text representing the features of the preset driving scene, and the scene feature data and the semantic description text are used as the scene description information.

8. A device for generating a virtual test scene model, characterized in that: The device comprises: An acquisition module, used to acquire scene data collected from a preset driving scene; an analysis module, configured to perform feature analysis on the scene data to obtain scene description information representing features of the preset driving scene; A first generation module is configured to input the scene description information into a pre-trained modeling language generation model to obtain modeling description information; The second generating module is used to convert the modeling description information using a preset test scenario modeling tool to obtain a virtual test scenario model.

9. An electronic device, characterized in that: include: Memory for storing computer programs; The processor is configured to execute the computer program stored in the memory, and when the computer program is executed, the method for generating a virtual test scene model according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for generating a virtual test scene model according to any one of claims 1 to 7 is implemented.

Images