Digital-real fusion test digital scene rapid generation method

Abstract

The invention discloses a digital-real fusion test digital scene rapid generation method, and belongs to the field of high-end equipment digital test verification, and the method comprises the steps: setting a test overall target, and functions and performance indexes of a to-be-tested object; according to characteristics of a test target and a to-be-tested object, in combination with an object test process, various kinds of basic unit scene data needed by a digital test scene in a digital-real fusion test are prepared; when the data is not enough to meet the requirement of high-precision scene model generation, data enhancement operation needs to be carried out; constructing a unit scene generation model based on deep learning by using the data set and the two-dimensional diffusion model; on the basis of the generated unit scene, constructing a test scene multi-element database and constructing a test rule constraint set; and a comprehensive scene test model is rapidly generated by using deep reinforcement learning, and the complexity and coverage of the generated comprehensive test scene are evaluated and verified, so that the requirement for the overall test target is met.

Description

Technical Field

[0001] The present invention belongs to the field of digital testing and verification of high-end equipment, and in particular relates to a method for quickly generating digital scenes for digital-real fusion testing. Background Art

[0002] With the continuous development of digitalization and automation technology, simulation testing has become a key link in modern system design and verification. Especially in complex scenarios such as unmanned systems, intelligent driving, and aerospace, real-world testing faces high costs and potential risks, and digital scene simulation has become an important alternative. Traditional digital scene generation methods are usually based on simple geometric modeling and physical simulation, which makes it difficult to accurately capture the various elements and their interactions in complex scenes, and cannot meet the needs of high-precision simulation. In addition, the coverage and complexity evaluation methods of existing test scenarios are limited, making it difficult to effectively verify whether the generated scenarios can fully meet the test requirements, further affecting the reliability and comprehensiveness of the test results. Therefore, there is an urgent need for a method that can quickly generate complex digital test scenarios with high fidelity and high coverage, and this method can adapt to changing test requirements, and verify the rationality and effectiveness of the scenarios through automation, thereby accelerating the verification process of the function, performance, safety, and reliability of the test object. Summary of the invention

[0003] The present invention aims to solve the problems of insufficient accuracy in digital test scene generation, and difficulty in complexity and coverage assessment in the prior art, and proposes a method for rapid generation of digital scenes for digital-real fusion testing. Through the method of the present invention, it is possible to automatically generate high-fidelity three-dimensional scenes from a variety of data sources, and optimize the scene generation process through deep learning and deep reinforcement learning algorithms, thereby meeting the functional and performance requirements of different types of test objects. The present invention solves its technical problems by adopting the following technical solutions: A method for rapid generation of digital scenes for digital-real fusion testing, comprising the following steps:

[0004] Step 1. Design a test scenario data construction module. This module generates various basic data required for model training based on the purpose and demand analysis of digital-physical fusion testing, including: Test goal setting: Set the overall test goal based on the type, function, and performance characteristics of the test object to ensure that the generated test scenario can cover all key requirements of the test object; Basic data preparation and enhancement: According to the test goals and the characteristics of the object to be tested, prepare the basic unit scenario data required for digital-physical fusion testing in combination with the test process, including geography, topography, meteorology, physical properties, model data, and road network data. When the existing data is not sufficient to meet the needs of generating high-precision scenario models, expand the diversity and accuracy of the data set through data enhancement operations, such as combining test rule embedding, data annotation, and data expansion;

[0005] Step 2, design a unit-level scene model generation module, which generates unit-level models for various test scenes required by the test object during the test process. The specific implementation is as follows: Unit scene generation: Based on the prepared data set, the two-dimensional diffusion model is injected with features in the test scene to achieve transfer training; the prompt words input to the two-dimensional image diffusion model are used to generate multi-angle pictures of each unit-level scene, and pictures that lack consistency are screened out; the SFM algorithm is used to process multi-view image data for light field reconstruction, and the three-dimensional point cloud space and viewpoint relationship of the scene are encoded; the scene is divided into multi-scale voxel grids through an adaptive grid division method, and the resolution is dynamically adjusted to adapt to complex areas; the layered volume rendering method is used to render optical features at different depths to achieve high-fidelity scene generation; the density calculation and rendering method in three-dimensional Gaussian splashing is used to achieve the generation of unit-level three-dimensional scenes;

[0006] Step 3, design a comprehensive scenario model generation module, which uses rule constraints to form a comprehensive scenario model that conforms to the actual situation and rules from various unit-level scenario models required by the test object during the test process. The specific implementation is as follows: Multi-factor database and rule constraint set construction: Based on the generated unit scenario, a multi-factor database of the test scenario is constructed, which covers scenario environment variables, test object parameters, dynamically changing data, etc. At the same time, a test rule constraint set is constructed, including expert knowledge sets, explicit and implicit constraint sets, to ensure that the test scenario can meet multi-faceted test standards; Deep reinforcement learning optimization: Using the deep reinforcement learning algorithm, a preliminary comprehensive test scenario of a variety of complex scenario combinations is generated through an adaptive sampler. Through multiple iterative evaluations and adjustments, the preliminary scenario model is further optimized, and finally a comprehensive test scenario model including standard test scenarios, extreme test scenarios, rare test scenarios and edge test scenarios is generated; Complexity and coverage verification: The complexity and coverage of the generated comprehensive test scenario are verified. Verification is performed through the complexity and coverage evaluation algorithm, and the evaluation criteria include the complexity of the environment type of the scenario, the complexity of the object behavior, and the coverage breadth and depth of the test scenario.

[0007] The present invention has the following beneficial technical effects: the system architecture of the present invention includes a test scenario data construction model module, a unit-level scenario model generation module, a unit scenario generation module, and a comprehensive-level scenario model generation module, and each module cooperates with each other to achieve efficient generation and verification of digital scenarios. Through the present invention, high-fidelity, high-complexity digital test scenarios can be quickly constructed to meet the test requirements of various complex systems and improve the accuracy, breadth and reliability of the test environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1This is a schematic diagram of a system for rapidly generating digital scenes for digital-real fusion testing according to the present invention. DETAILED DESCRIPTION

[0009] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above-mentioned purpose, the present invention adopts the following technical scheme.

[0010] The present invention proposes a method for quickly generating digital scenes for digital-real fusion testing. Figure 1 As shown, the method includes:

[0011] Step 1: Design the test scenario data construction module, and realize the generation of various basic data required for model training according to the purpose and demand analysis of digital-real fusion testing. The specific implementation is as follows:

[0012] Analyze the test object, and set the overall test goal according to the type, function and performance characteristics of the test object to ensure that the generated test scenario can cover all key requirements of the test object, including: propose test requirements / goals, performance indicators, and formulate a preliminary test implementation plan; according to the test objectives and the characteristics of the object to be tested, combined with the object test process, realize the preparation of various basic unit scene data required for digital test scenarios in digital-physical fusion testing, including geography, topography, meteorology, physical properties, model data and road network data; when the data is not sufficient to meet the needs of high-precision scene model generation, data enhancement operations are required, such as combining test rule embedding, data annotation, data expansion, etc., to expand the diversity and accuracy of the data set and build a data set;

[0013] Step 2: Design a unit-level scenario model generation module to quickly generate unit-level models of various test scenarios required by the test object during the test process. The specific implementation is as follows:

[0014] Using the constructed dataset, the image diffusion model that realizes image generation is subjected to transfer training by injecting features in the test scenario. The prompt words input to the image diffusion model are used to generate multi-angle pictures of each unit-level scene, and pictures that lack consistency are screened out. The SFM algorithm is used to process multi-view image data for preliminary scene reconstruction to obtain the three-dimensional point cloud space and camera extrinsics of the encoded scene. According to the characteristics of the point cloud scene, the scene is divided into multi-scale point cloud areas by using an adaptive point cloud partitioning method, and the resolution is dynamically adjusted to adapt to complex areas. A layered volume rendering method is used to render optical features at different depths to achieve high-fidelity scene generation. The density calculation and rendering method in three-dimensional Gaussian splashing is used to generate each three-dimensional unit-level scene.

[0015] Step 3: Design a comprehensive scenario model generation module, which uses rule constraints to form a comprehensive scenario model that conforms to the actual situation and rules based on various unit-level models required during the test of the test object. The specific implementation is as follows:

[0016] Based on the generated unit-level scenarios and test requirements, a multi-factor database of test scenarios and a test rule constraint set are constructed; the database covers scenario environment variables, test object parameters, dynamically changing data, etc. The test rule constraint set includes expert knowledge set, explicit and implicit constraint sets; deep reinforcement learning is used, combined with the test rule constraint set and the test target, and the optimal scenario elements and the relative position relationship between the elements are selected through an adaptive sampler to generate a preliminary comprehensive test scenario of multiple complex scenario combinations. Through multiple iterative evaluations and adjustments, a comprehensive-level scenario test model is quickly generated. Through multiple iterative evaluations and adjustments, the preliminary scenario model is further optimized, and finally a comprehensive test scenario model including standard test scenarios, extreme test scenarios, rare test scenarios and edge test scenarios is generated; finally, the complexity and coverage of the comprehensive test scenario are evaluated and verified, and the complexity and coverage evaluation algorithm is used for verification. The evaluation criteria include the complexity of the environment type of the scenario, the complexity of the object behavior, and the coverage breadth and depth of the test scenario.

[0017] Furthermore, the basic unit-level scene data in step 1 includes, but is not limited to, geographic environment data, terrain data, meteorological data, physical property data, object function and performance data, model data, road network data and related historical test data.

[0018] Furthermore, the data enhancement operation in step 1 includes combining test rule embedding, data annotation, data expansion, etc. to expand the diversity and accuracy of existing data and build a data set covering a wide range of scenario variables.

[0019] Furthermore, the consistency in step 2 includes geometric consistency, lighting consistency, and material consistency, ensuring that the generated three-dimensional scene complies with the laws of physics.

[0020] Furthermore, the test rule constraint set in step 3 includes an expert knowledge set, a test rule set, an explicit constraint set, and an implicit constraint set to ensure that the test scenario meets various test requirements and standards.

[0021] Furthermore, the comprehensive-level scenarios in step 3 include various comprehensive test scenario models including standard test scenarios, extreme test scenarios, rare test scenarios and edge test scenarios.

[0022] Furthermore, the complexity and coverage in step 3 include the complexity of the scene's environment type, the complexity of object behavior, and the coverage breadth and depth of the test scene.

[0023] More specifically, the present invention provides the following illustrative embodiments: The overall test goal is set according to the type, function and performance of the test object.

[0024] The test subjects may be autonomous vehicles, drones, or other equipment systems with complex functions.

[0025] The overall goal defines the scenario conditions required for testing, including routine and extreme operating scenarios, testing requirements for different environmental variables (such as weather, terrain, etc.), and evaluation of the system's performance under various conditions.

[0026] The setting of test objectives is the basis of the scenario generation process, ensuring that the final generated test scenario meets the requirements of expected functions and performance verification.

[0027] According to the requirements of the test objectives, the system needs to prepare a variety of basic unit scene data. These data include but are not limited to geographic data, terrain data, meteorological data, model data, road network data, building data, etc.

[0028] By analyzing the functional and performance requirements of the test object, select the appropriate data set for preliminary processing.

[0029] Terrain data can come from actual surveying and mapping data, meteorological data can include historical meteorological data and future weather forecast information, and model data may cover three-dimensional models of objects such as vehicles and drones in physical scenes and their dynamic characteristics.

[0030] In some cases, existing datasets may not fully meet the requirements for generating high-fidelity scene models.

[0031] Therefore, data enhancement is required. Data enhancement can be achieved by embedding test rules, data annotation, data expansion, etc.

[0032] For example, for scenarios where data is scarce, the system can combine rule embedding, such as simulating scenarios where there is no actual data through physical modeling or synthetic data generation methods; it can also improve the accuracy and diversity of scenario generation by annotating existing data and expanding the data set.

[0033] Data augmentation ensures that test scenarios with high complexity can be generated even with incomplete or low-quality data.

[0034] The diffusion model is transferred and trained using a two-dimensional dataset in flight scenarios, injecting characteristics relevant to drone testing (such as wind speed, temperature, and terrain) into the model.

[0035] Generate multi-angle pictures of unit-level scenes based on prompt words, such as simulating the flight status of drones in different time periods and weather conditions.

[0036] Use the consistency distillation method to screen and generate images to ensure the consistency of geometry, lighting, and materials, so that the scene conforms to the laws of physics.

[0037] After data preparation and data enhancement are completed, an algorithm is used to generate high-precision 3D unit scenes. NeRF generates a 3D representation of the scene by inputting 2D images from multiple perspectives into a neural network.

[0038] Specifically, the algorithm uses a volume rendering approach to encode the scene's geometry and lighting.

[0039] By predicting the density and color of each voxel in the voxel grid, high-precision reconstruction of each object and its details in the scene can be achieved.

[0040] For example, in complex urban environments, NeRF can capture the appearance of buildings, road details, and occlusion relationships between objects, and generate three-dimensional scenes that are highly consistent with actual scenes.

[0041] Based on the generated unit scenario, the system builds a multi-factor database and creates a set of rule constraints for the test scenario. The multi-factor database includes scenario environment variables, dynamic data (such as the speed and acceleration of moving objects), the model of the test object and its physical parameters. These elements are the basis for the subsequent generation of comprehensive scenarios.

[0042] In addition, the rule constraint set includes the expert knowledge set, the test rule set, the explicit constraint set and the implicit constraint set.

[0043] The expert knowledge set sets the physical constraints for scenario generation through expert experience, such as the maximum centrifugal force when a vehicle turns; the test rule set defines the operating behavior of the test object in the scenario, such as the route planning and reaction time of an autonomous vehicle; the explicit constraint set is set for explicit conditions in the scenario, and the implicit constraint set covers scenario rules that cannot be directly observed, such as implicit system vulnerabilities.

[0044] After the comprehensive test scenario is generated, the system verifies its complexity and coverage.

[0045] First, the complexity verification analyzes the comprehensive scene element model and calculates the weight of each element in the scene, and combines the information entropy method to evaluate the complexity of the scene. The specific formula of information entropy is as follows:

[0046] ,

[0047] ,

[0048] ,

[0049] Among them, H(X) represents the information entropy of the source X, H(XY) represents the joint entropy of X and Y, Represents the information entropy of the system.

[0050] This process can quantify the interrelationships and uncertainties among multiple variables in the scenario, ensuring that the complexity of the scenario meets the actual application requirements.

[0051] The coverage verification process includes test requirement analysis, effective test scenario screening and coverage calculation.

[0052] The system first analyzes the test objectives and screens out test scenarios that are closely related to the objectives; then it evaluates the capabilities of each test scenario and calculates the coverage of the scenario test to ensure that the generated test scenarios can effectively cover all functional and performance requirements, thereby ensuring the comprehensiveness and reliability of the test.

[0053] The system architecture of the present invention includes a data preparation module, a data enhancement module, a unit scene generation module, a rule constraint module, a deep reinforcement learning module and an evaluation and verification module. The modules cooperate with each other to achieve efficient generation and verification of digital scenes.

[0054] The contents not described in detail in the specification of the present invention belong to the prior art known to the professional and technical personnel in this field.

[0055] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for rapidly generating digital scenes for digital-real fusion testing, characterized in that: include: Step 1: Design a test scenario data construction module. This module generates various basic data required for model training based on the purpose and demand analysis of digital-real fusion testing. The specific implementation is as follows: Analyze the test object, propose test requirements or goals, performance indicators, and formulate a preliminary test implementation plan; According to the test objectives and the characteristics of the object to be tested, combined with the object test process, prepare various basic unit scenario data required for digital test scenarios in digital-real fusion testing; When the data is insufficient to meet the needs of scenario model generation, perform data enhancement operations; Step 2, design a unit-level scene model generation module, which generates unit-level models for various test scenes required by the test object during the test process. The specific implementation is as follows: using the constructed data set, the image diffusion model that realizes image generation is injected with features in the test scene to achieve migration training; using the prompt words input to the image diffusion model to generate multi-angle pictures of each unit-level scene, and filtering out pictures that lack consistency; using the SFM algorithm to process multi-view image data, perform preliminary scene reconstruction, and obtain the three-dimensional point cloud space and camera extrinsics of the encoded scene; according to the characteristics of the point cloud scene, by using the adaptive point cloud partitioning method, the scene is divided into multi-scale point cloud areas, and the resolution is dynamically adjusted to adapt to complex areas; using the density calculation and rendering method in the three-dimensional Gaussian splashing, the generation of each three-dimensional unit-level scene is achieved; Step 3. Design a comprehensive scenario model generation module. This module uses rule constraints to form a comprehensive scenario model that conforms to actual conditions and laws for various unit-level scenario models required during the test of the test object. The specific implementation is as follows: Based on the generated unit-level scenarios and test requirements, a multi-factor database of test scenarios and a test rule constraint set are constructed; deep reinforcement learning is used to combine the test rule constraint set and the test objectives to quickly generate a comprehensive scenario test model that meets the requirements; finally, the complexity and coverage of the comprehensive test scenario are evaluated and verified.

2. The method for rapidly generating digital scenes for digital-real fusion testing according to claim 1 is characterized in that: The basic unit-level scene data in step 1 includes, but is not limited to, geographic environment data, terrain data, meteorological data, physical property data, object function and performance data, model data, road network data and related historical test data.

3. The method for rapidly generating digital scenes for digital-real fusion testing according to claim 1 is characterized in that: The data enhancement operation in step 1 includes combining test rule embedding, data annotation, and data expansion methods to expand the diversity and accuracy of existing data and build a data set covering a wide range of scenario variables.

4. The method for rapidly generating digital scenes for digital-real fusion testing according to claim 1 is characterized in that: The consistency in step 2 includes geometric consistency, lighting consistency, and material consistency, ensuring that the generated 3D scene complies with the laws of physics.

5. The method for rapidly generating digital scenes for digital-real fusion testing according to claim 1 is characterized in that: The test rule constraint set in step 3 includes an expert knowledge set, a test rule set, an explicit constraint set, and an implicit constraint set to ensure that the test scenario meets various test requirements and standards.

6. The method for rapidly generating digital scenes for digital-real fusion testing according to claim 1 is characterized in that: The comprehensive test scenarios in step 3 include various comprehensive test scenario models including standard test scenarios, extreme test scenarios, rare test scenarios and edge test scenarios.

7. The method for rapidly generating digital scenes for digital-real fusion testing according to claim 1 is characterized in that: The comprehensive test scenario complexity and coverage in step 3 include the environment type complexity of the scenario, the object behavior complexity, and the coverage breadth and depth of the test scenario.

Images