Virtuality and reality combined automatic driving automobile test system and method

Abstract

The invention relates to the technical field of automatic driving, in particular to a virtuality and reality combined automatic driving automobile test system and method, which are characterized in that a simulation test of an automatic driving automobile and a real automobile road test are fused, and a virtual test scene is constructed by collecting real scene data, so that the authenticity of the test scene is improved; a virtual dangerous scene is generated through derivation of real data, and a real vehicle is combined with the derived scene to carry out closed site testing, so that the limitation of a real vehicle road testing scene is overcome, and the testing efficiency of the automatic driving vehicle is improved; synchronous interaction of data between virtual and real spaces is realized, so that a simulation test result is closer to a real vehicle test result, and the credibility of the test result is improved; in conclusion, the test system and method provided by the invention combine the advantages of simulation test and real vehicle test, and have great significance in accelerating the development of the automatic driving test technology.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a virtual-real combined autonomous driving vehicle testing system and method. Background Art

[0002] In recent years, with the annual growth of the number of cars, the number of road traffic accidents has remained high, traffic congestion and other problems have arisen. According to statistical analysis, driver errors caused by factors such as driver distraction and fatigue are one of the main causes of traffic accidents. Autonomous vehicles rely on autonomous driving systems to control vehicle driving, which to a certain extent reduces the impact of driver behavior on vehicle driving. Therefore, in order to reduce traffic accidents and improve road traffic safety, it is necessary to increase research and development of autonomous vehicles. In order to ensure the safety of autonomous vehicles, a large number of real vehicle tests and simulation tests must be carried out before autonomous vehicles are put into use on the market to verify the safety and stability of autonomous vehicles, thereby accelerating the early market entry of autonomous vehicles.

[0003] The autonomous driving system faces many challenges brought by environmental uncertainty. Factors such as complex traffic environment and changeable weather conditions increase the difficulty of autonomous driving testing. Current autonomous driving car testing methods mainly include open road testing, closed field testing, and virtual simulation testing. Both open road testing and closed field testing are carried out in real traffic environments. There are problems such as long test cycle, high cost, low safety, and non-repeatability of scenes. Virtual simulation testing is carried out in a virtual environment. Although the test cost is low and the safety is high, there is a disadvantage of large errors in virtual simulation results. Therefore, the present invention combines simulation testing with real vehicle road testing, and proposes a virtual and real combined autonomous driving car testing system and method, in order to improve the accuracy of autonomous driving test results, speed up the autonomous driving test process, etc., and provide theoretical support for the research and development of autonomous driving technology. Summary of the invention

[0004] The purpose of the present invention is to provide a virtual-reality combined autonomous driving car testing system and method, which aims at the shortcomings of the above-mentioned autonomous driving car testing method, combines the advantages of simulation testing and real vehicle testing, so as to overcome the limitations of real vehicle testing scenarios, high testing costs and low accuracy of simulation test results, and is of great significance to accelerating the development of autonomous driving testing technology.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: A virtual and real combined autonomous driving vehicle testing system, including a scenario construction module, a simulation test module, a real vehicle test module and an evaluation and analysis module; The scenario construction module is used to provide a scenario that is regionally representative and credible for testing and verifying the autonomous driving function. The scenario construction module generates a virtual test scenario based on real data. The simulation test module is used for performing simulation tests on the virtual twin vehicle in the simulation platform; The real vehicle test module is used to perform real vehicle tests on the real target vehicle in a closed field environment in combination with the virtual test scenario; the evaluation and analysis module is used to output comparative evaluation results of the simulation test and the real vehicle test of the target vehicle; The scenario construction module includes data collection, data processing, data derivation, scenario generation and scenario input to the cloud. The test scenario specifically includes twins and derivative scenarios of real traffic scenarios; The data collection is to collect point cloud data of real traffic scenes through vehicle-side sensors, laser radar, and camera equipment, and integrate the collected data and send it to data processing; The data processing is to pre-process the collected data to form a valid data set; The derivation of the data is to use a deep learning algorithm to simulate learning on real data to generate derived data; The scene generation is to extract features from valid data sets and derived data, set road types, environment and lighting types, and virtual traffic participants in the simulation software to generate twin scenes and derived scenes of the real scene; The scenario input is to input the generated test scenario into the cloud to provide a test scenario for the virtual simulation and real vehicle test modules of the autonomous driving vehicle.

[0006] The road types in the virtual test scene include highways, urban roads and rural roads, the environment and lighting types include sunny days, rainy days, snowy days, foggy days, daytime and night, and the virtual traffic participants include dynamic pedestrians, non-motor vehicles and motor vehicles.

[0007] The simulation test module includes a virtual twin of the target vehicle, a virtual test scene, and a simulation test platform; CarSim and Simulink software are used to build a virtual twin vehicle model corresponding to the target vehicle, and the appearance dimension information of the virtual vehicle is set to be consistent with the real vehicle model; the simulation test platform imports the virtual test scene in the Prescan simulation software, adds sensor models and other traffic participant models, and is used for virtual simulation testing of the target vehicle.

[0008] The real vehicle test module is used to conduct real vehicle testing in a closed venue by combining a real autonomous driving car with a virtual test scene in the cloud. The real vehicle test module mainly includes a real autonomous driving car, a virtual test scene and a closed test site.

[0009] Furthermore, data interaction between the simulation test module and the real vehicle test module is achieved through the communication transmission system, specifically including: in the simulation test, the virtual sensor model perceives the information of the surrounding environment and traffic participants, transmits the data to the real vehicle controller, and provides input for information fusion and decision control of the real vehicle. After receiving the information, the real vehicle controller sends the control decision instruction to the actual actuator, and feeds back the position coordinates, posture information, and vehicle trajectory data of the real vehicle to the simulation platform.

[0010] The evaluation and analysis module includes evaluation indicators and evaluation methods, and is used to output comparative evaluation results of simulation tests and actual vehicle tests of the target vehicle.

[0011] The present invention provides a virtual and real combined automatic driving vehicle testing method, comprising the following steps: Step S1, collecting point cloud data of real traffic scenes based on vehicle-side sensors, laser radars, and cameras; Step S2, deleting and repairing the data to obtain a valid data set, and performing scene feature extraction; Step S3, performing simulation learning on real data through a deep learning algorithm to generate derivative data; Step S4, the processed data is combined with 3D modeling technology to construct high-fidelity twin and derivative scenes of the real scene, and the generated virtual test scene is uploaded to the cloud via the network; Step S5, using CarSim and Simulink software to build a corresponding virtual twin vehicle model, and setting the consistency of the virtual vehicle's appearance dimension information with the real vehicle model; Step S6, using Prescan software to build a virtual simulation test platform for the target vehicle, importing virtual test scenarios into the simulation platform, adding sensor models, traffic participants and other traffic elements for simulation testing of the virtual twin vehicle; Step S7, the real autonomous driving car is tested in a closed field in combination with the virtual test scene, and its operation is detected to test the performance of the real vehicle combined with the virtual scene autonomous driving car; Step S8, realizing data interaction between the simulation test module and the real vehicle test module through the communication transmission system, specifically including: the perception data of the virtual sensor model is injected into the real vehicle controller to provide input for information fusion and decision control of the real vehicle. After receiving the information, the real vehicle controller sends the control decision instruction to the actual actuator, and feeds back the position coordinates, posture information, vehicle trajectory and other data of the real vehicle to the simulation platform, and guides the simulation test of the virtual twin vehicle according to the control decision instruction of the real vehicle; Step S9, select appropriate evaluation indicators and evaluation methods, and output comparative evaluation results of autonomous driving vehicle simulation and actual vehicle testing.

[0012] Furthermore, in order to ensure the synchronous mapping between the virtual and real vehicles, it is necessary to achieve the consistency of the position of the real vehicle and the virtual twin vehicle, that is, the coordinate transformation of the autonomous driving vehicle in the virtual and real space. In the coordinate system of the virtual space, assuming that the coordinates of the twin vehicle are p(x, y, z), after homogeneous transformation, they are p′(x′, y′, z′), then the transformation matrix between p and p′ is: Where T represents the translation matrix: Among them, T x , T y , T z Respectively represent the translation components of the vehicle in the directions of the three coordinate axes x, y, and z; the selected angles of the rotation matrix R on the x, y, and z axes can be determined by R x , R y , R z Represents, and has the following relationship: R=R x ·R y ·R z The selection matrix R for the rotation angle α around the x-axis x for: The selection matrix R for the rotation angle β around the y-axis y for: The selection matrix R for the rotation angle γ around the z-axis z for: Through the above-mentioned coordinate transformation, it is possible to map the posture changes of the real vehicle to the corresponding vehicle twin in the virtual space.

[0013] Furthermore, for the simulation and actual vehicle test of the target vehicle, a comprehensive comparative evaluation is conducted around the three aspects of safety, energy saving effect and traffic efficiency. The evaluation index is determined by the empirical method, and the weight of each evaluation index is obtained by the entropy method. The specific calculation formula is as follows: Calculate the entropy value H of each indicator i After that, the weight value of each evaluation index is calculated by the weight calculation formula, which is as follows: Among them, x ij It represents the jth sample value under the i-th evaluation index. After standardizing its data, the standard value is r ij , p ijis the proportion of the jth sample value under the i-th evaluation index to the index, H i is the entropy value of the i-th evaluation index. After the entropy value of each index is determined, the weight value of each evaluation index is calculated as ω i .

[0014] Furthermore, each evaluation index is normalized and combined with the weight of each evaluation index, the calculation formula of the comprehensive evaluation result is: Among them, ω i is the weight of the i-th evaluation index, S i is the normalized result of the i-th evaluation index, and R is the comprehensive evaluation result. The beneficial effects of the present invention are as follows: the present invention proposes a virtual-real combined autonomous driving vehicle testing system and method, which integrates the simulation test and real vehicle road test of the autonomous driving vehicle, and constructs a virtual test scene by collecting real scene data, thereby improving the authenticity of the test scene; by deriving the real data to generate a virtual dangerous scene, the real vehicle is combined with the derived scene for closed field testing, thereby overcoming the limitations of the real vehicle road test scene and improving the testing efficiency of the autonomous driving vehicle; realizing the synchronous interaction of data between the virtual and real space, so that the simulation test results are closer to the real vehicle test results, and increasing the credibility of the test results; in summary, the test system and method proposed by the present invention combine the advantages of simulation testing and real vehicle testing to overcome the limitations of the real vehicle test scene, high testing costs, and low accuracy of simulation test results, which is of great significance to accelerate the development of autonomous driving test technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of the steps of the virtual-real combination testing method of the automatic driving vehicle provided by the present invention; Figure 2 It is a structural diagram of the virtual-real combination test system of the automatic driving vehicle provided by the present invention; Figure 3 This is a diagram of the comprehensive verification solution for vehicle-cloud integration based on the combination of virtual and real provided by the present invention. DETAILED DESCRIPTION

[0016] Specific embodiment 1: The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. It should be noted that: in the present invention, if there is no special explanation, all the implementation methods and preferred implementation methods mentioned in this article can be combined with each other to form a new technical solution. In the present invention, if there is no special explanation, all the technical features and preferred features mentioned in this article can be combined with each other to form a new technical solution. Unless otherwise specified, the professional and scientific terms used in this article have the same meaning as those familiar to those skilled in the art.

[0017] In view of the problems existing in the existing autonomous driving car testing systems and methods, such as low testing efficiency, large errors in test results, high costs, and non-repeatable scenarios, the present invention proposes a virtual-real combined autonomous driving car testing system and method to improve the safety of autonomous driving car testing and the accuracy of test results and overcome the limitations of real vehicle testing scenarios. Specifically, the specific implementation methods of the present invention are described as follows in conjunction with the accompanying drawings.

[0018] manual Figure 1 This is a flowchart of the steps of the virtual-real combination test method for the autonomous driving vehicle provided by the present invention, as shown in the attached manual. Figure 1 As shown, the specific steps of the method of the present invention are as follows: Step S1, collecting point cloud data of real traffic scenes based on vehicle-side sensors, laser radars, cameras, etc.; Step S2, the data is deleted, repaired, etc. to obtain a valid data set, and scene feature extraction and other processing are performed; Step S3, performing simulation learning on real data through a deep learning algorithm to generate derivative data; Step S4, the processed data is combined with 3D modeling technology to construct high-fidelity twin and derivative scenes of the real scene, and the generated virtual test scene is uploaded to the cloud via the network; Step S5, using CarSim and Simulink software to build a corresponding virtual twin vehicle model, and setting the consistency of the virtual vehicle's appearance dimensions and other information with the real vehicle model; Step S6, using Prescan software to build a virtual simulation test platform for the target vehicle, importing virtual test scenarios into the simulation platform, adding sensor models, traffic participants and other traffic elements for simulation testing of the virtual twin vehicle; Step S7, the real autonomous driving car is tested in a closed field in combination with the virtual test scene, and its operation is detected to test the performance of the real vehicle combined with the virtual scene autonomous driving car; Step S8, realizing data interaction between the simulation test module and the real vehicle test module through the communication transmission system, specifically including: the perception data of the virtual sensor model is injected into the real vehicle controller to provide input for information fusion and decision control of the real vehicle. After receiving the information, the real vehicle controller sends the control decision instruction to the actual actuator, and feeds back the position coordinates, posture information, vehicle trajectory and other data of the real vehicle to the simulation platform, and guides the simulation test of the virtual twin vehicle according to the control decision instruction of the real vehicle; Step S9, select appropriate evaluation indicators and evaluation methods, and output comparative evaluation results of autonomous driving vehicle simulation and actual vehicle testing.

[0019] Instructions attached Figure 2 This is the structure diagram of the automatic driving car test system based on virtual and real combination provided by the present invention, as shown in the attached manual. Figure 2 As shown, the virtual-real combined autonomous driving vehicle testing system of the present invention includes: a scenario construction module, a simulation test module, a real vehicle test module and an evaluation and analysis module; The scenario construction module is used to provide a scenario with regional representativeness and credibility for the test and verification of the autonomous driving function, including data collection and processing of real traffic scenarios and derivation of real data. The processed data is combined with 3D modeling software to construct a high-fidelity virtual test scenario. The virtual test scenario specifically includes road type, environment type, road test equipment, other traffic participants, and traffic flow data, and the constructed virtual test scenario is uploaded to the cloud; Specifically, the scene types constructed by the scene construction module include twins and derivative scenes of real traffic scenes. Derivative scenes are mostly dangerous scenes such as accident scenes, edge scenes, tunnels and mountain roads; the road types of the virtual test scenes include highways, urban roads and rural roads, etc., and the environment types include sunny days, rainy days, snowy days, foggy days, daytime and night, etc. Other traffic participants include the forms and movements of pedestrians, non-motor vehicles and motor vehicles, etc.; The simulation test module is used for simulation testing of the virtual twin model of the target vehicle, including a virtual simulation test platform, a virtual twin model of the target vehicle and a virtual test scene; The virtual twin vehicle is a model obtained by mapping the physical model of the real target vehicle one by one; the virtual simulation test platform includes a scene road map, an environment editor, a virtual sensor model, virtual traffic elements, a tested vehicle and other traffic participant models, etc.; The real vehicle test module is used for closed-field testing of the autonomous driving vehicle in combination with the virtual test scenario, and includes a real vehicle model, a virtual test scenario, and a closed test site. The real vehicle model specifically includes a physical model of the autonomous driving vehicle, a decision-making planning algorithm, a control actuator, etc.; Specifically, the communication transmission system is used to realize the data interaction between the simulation test module and the real vehicle test module; The evaluation and analysis module includes evaluation indicators and evaluation methods, and conducts comprehensive evaluation on the three aspects of safety, energy-saving effect and traffic efficiency for simulation and actual vehicle testing of target vehicles, and outputs comparative evaluation results of the two.

[0020] Instructions attached Figure 3 This is a comprehensive verification scheme diagram of the vehicle-cloud integration based on virtual-real integration of the autonomous driving vehicle provided by the present invention, as shown in the attached manual. Figure 3 As shown in the figure, a real-time synchronized virtual operation scene is constructed based on the real vehicle road scene, and the real-time synchronization of information is maintained between the virtual and real scenes; a virtual self-driving vehicle is constructed based on the real self-driving vehicle, the real vehicle is tested in a real open road environment, and the virtual self-driving vehicle is simulated in the virtual operation scene, and the real vehicle road test results and simulation results are uploaded to the cloud; among them, the cloud-based virtual controller of the learning and training module is used to perceive and control the virtual self-driving vehicle, and provides a basis for the optimization of the real vehicle control strategy according to the simulation test results of the digital twin model; the vehicle-side real controller of the management and control module is used to perceive and control the real self-driving vehicle, and provides a basis for the verification of the digital twin model according to the vehicle-side test results; throughout the test and verification process, the real-time synchronization of information between the virtual and real spaces is always maintained, thereby realizing the interactive iterative optimization of the virtual and real spaces. Finally, the evaluation results of the experiment and evaluation module are used to verify the effectiveness of the virtual-real combination method for the comprehensive verification of the vehicle-road-cloud integration of self-driving vehicles. The following examples will be used to further describe the embodiments of the present invention. The present invention provides a virtual-real combination autonomous driving vehicle testing system and method, wherein the virtual-real combination autonomous driving vehicle testing system includes a scene construction module, a simulation test module, a real vehicle test module and an evaluation and analysis module, and the testing method is implemented based on the test system. The scene input module uses a certain autonomous driving demonstration closed park as a real vehicle test site, firstly collects the real scene data based on vehicle-side sensors, radars, cameras and other equipment, collects the point cloud data of the road environment and dynamic traffic participants, and after the data is deleted, repaired and processed, the effective data set obtained is used to extract scene element features, and the deep learning algorithm is combined with the effective data set after the above processing to simulate and learn to generate derivative data. For the real data set and derivative data, the twin scene and other derivative scenes of the autonomous driving demonstration closed park are constructed using a map editor and three-dimensional modeling technology, and the constructed virtual test scene is transmitted to the cloud for simulation testing and real vehicle testing of the target vehicle. Specifically, the road type, environment type, road test facilities and other traffic participants of the virtual test scene can be set according to the real scene data. The simulation test module builds a corresponding virtual twin model based on the physical model of the tested autonomous vehicle, edits the road, environment, traffic participants and other data of the twin test scene in the simulation test platform, and adds virtual sensor models, decision planning algorithms, controller models, etc. in the simulation platform. The virtual twin vehicle is combined with the twin scene of a demonstration park for virtual simulation testing. In the real vehicle test module, the tested autonomous vehicle is tested in a closed field in combination with the virtual test scene, and the data interaction between the virtual and real space is completed through the communication transmission system. Specifically, the actual autonomous driving car is tested in the original demonstration park scene. In order to ensure the consistency of the virtual and real vehicles, the position coordinates, vehicle trajectory, posture information, etc. of the real vehicle are mapped to the virtual simulation platform, and the perception, decision, and control data of the real vehicle are sent to the simulation platform through the communication transmission system. The simulation test of the virtual twin vehicle model is guided according to the control decision instructions of the real vehicle. Furthermore, the actual target vehicle is tested in a closed field in combination with the derived test scenario. The data detected by the virtual sensor is input to the real vehicle controller through the communication transmission system. After the real vehicle processes the perception data, the control decision instruction is sent to the actual actuator, and the position coordinates, posture information, vehicle trajectory and other data of the real vehicle are fed back to the simulation platform. The simulation test of the virtual twin vehicle is guided according to the control decision instruction of the real vehicle. For the simulation and real vehicle testing of autonomous driving vehicles, appropriate evaluation indicators and evaluation methods are selected to output the comparative evaluation results of the two.Specifically, in order to achieve consistent transformation of the position coordinates of the target vehicle and the twin model in the virtual and real space, it is necessary to establish the vehicle position coordinates in a unified coordinate system to improve the overall test accuracy; the coordinate transformation between the real vehicle and the virtual twin vehicle can be achieved through the transformation matrix. In the coordinate system of the virtual space, assuming that the coordinates of the twin vehicle are p(x, y, z), and after homogeneous transformation are p′(x′, y′, z′), then the transformation matrix between p and p′ is:. Where T represents the translation matrix: Among them, T x , T y , T z Respectively represent the translation components of the vehicle in the directions of the three coordinate axes x, y, and z; the selected angles of the rotation matrix R on the x, y, and z axes can be determined by R x , R y , R z Represents, and has the following relationship: R=R x ·R y ·R z The selection matrix R for the rotation angle α around the x-axis x for: The selection matrix R for the rotation angle β around the y-axis y for: The selection matrix R for the rotation angle γ around the z-axis z for: Through the above coordinate transformation, the position change of the real vehicle can be mapped to the corresponding vehicle twin in the virtual space. Specifically, a comprehensive evaluation is conducted around the three aspects of safety, energy saving effect and traffic efficiency. The evaluation index is determined by the empirical method, and the entropy method is used to obtain the weight of each evaluation index. The specific calculation formula is as follows: Calculate the entropy value H of each indicator i After that, the weight value of each evaluation index is calculated by the weight calculation formula, which is as follows: Among them, x ij It represents the jth sample value under the i-th evaluation index. After standardizing its data, the standard value is r ij , p ij is the proportion of the jth sample value under the i-th evaluation index to the index, Hi is the entropy value of the i-th evaluation index. After the entropy value of each index is determined, the weight value of each evaluation index is calculated as ω i . Each evaluation index is normalized and combined with the weight of each evaluation index. The calculation formula of the comprehensive evaluation result is: Among them, ω i is the weight of the i-th evaluation index, S i is the normalized result of the i-th evaluation index, and R is the comprehensive evaluation result. The above is only a preferred embodiment of the present invention, and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modify the technical content disclosed above to an equivalent embodiment of equivalent changes without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still belongs to the scope of the technical solution of the present invention.

Claims

1. A virtual and real combined autonomous driving vehicle testing system, characterized in that: It includes scenario construction module, simulation test module, real vehicle test module and evaluation and analysis module; The scenario construction module is used to provide a scenario that is regionally representative and credible for testing and verifying the autonomous driving function. The scenario construction module generates a virtual test scenario based on real data. The simulation test module is used for performing simulation tests on the virtual twin vehicle in the simulation platform; The real vehicle test module is used to perform real vehicle tests on real target vehicles in a closed field environment in combination with virtual test scenarios; The evaluation and analysis module is used to output the comparative evaluation results of the simulation test and the real vehicle test of the target vehicle; the scenario construction module includes data collection, data processing, data derivation, scenario generation and scenario input to the cloud. The test scenario specifically includes the twin and derivative scenarios of the real traffic scenario; The data collection is to collect point cloud data of real traffic scenes through vehicle-side sensors, laser radar, and camera equipment, and integrate the collected data and send it to data processing; The data processing is to pre-process the collected data to form a valid data set; The derivation of the data is to use a deep learning algorithm to simulate learning on real data to generate derived data; The scene generation is to extract features from valid data sets and derived data, set road types, environment and lighting types, and virtual traffic participants in the simulation software to generate twin scenes and derived scenes of the real scene; The scenario input is to input the generated test scenario into the cloud to provide a test scenario for the virtual simulation and real vehicle test modules of the autonomous driving vehicle.

2. The virtual-real combined automatic driving vehicle testing system according to claim 1, characterized in that: The road types in the virtual test scene include highways, urban roads and rural roads, the environment and lighting types include sunny days, rainy days, snowy days, foggy days, daytime and night, and the virtual traffic participants include dynamic pedestrians, non-motor vehicles and motor vehicles.

3. The virtual-real combined automatic driving vehicle testing system according to claim 2, characterized in that: The simulation test module includes a virtual twin of the target vehicle, a virtual test scene, and a simulation test platform; CarSim and Simulink software are used to build a virtual twin vehicle model corresponding to the target vehicle, and the appearance dimension information of the virtual vehicle is set to be consistent with the real vehicle model; the simulation test platform imports the virtual test scene in the Prescan simulation software, adds sensor models and other traffic participant models, and is used for virtual simulation testing of the target vehicle.

4. The virtual-real combined automatic driving vehicle testing system according to claim 3 is characterized in that: The real vehicle test module is used to conduct real vehicle testing in a closed venue by combining a real autonomous driving car with a virtual test scene in the cloud. The real vehicle test module mainly includes a real autonomous driving car, a virtual test scene and a closed test site.

5. The virtual-real combined automatic driving vehicle testing system according to claim 4, characterized in that: The data interaction between the simulation test module and the real vehicle test module is realized through the communication transmission system, which includes: in the simulation test, the virtual sensor model perceives the information of the surrounding environment and traffic participants, transmits the data to the real vehicle controller, and provides input for information fusion and decision control of the real vehicle. After receiving the information, the real vehicle controller sends the control decision instruction to the actual actuator, and feeds back the position coordinates, posture information, and vehicle trajectory data of the real vehicle to the simulation platform.

6. The virtual-real combined automatic driving vehicle testing system according to claim 1, characterized in that: The evaluation and analysis module includes evaluation indicators and evaluation methods, and is used to output comparative evaluation results of simulation tests and actual vehicle tests of the target vehicle.

7. The method for operating a virtual-real combined automatic driving vehicle testing system according to any one of claims 1 to 6, characterized in that: The following steps are involved: Step S1, collecting point cloud data of real traffic scenes based on vehicle-side sensors, laser radars, and cameras; Step S2, deleting and repairing the data to obtain a valid data set, and performing scene feature extraction; Step S3, performing simulation learning on real data through a deep learning algorithm to generate derivative data; Step S4, the processed data is combined with 3D modeling technology to construct high-fidelity twin and derivative scenes of the real scene, and the generated virtual test scene is uploaded to the cloud via the network; Step S5, using CarSim and Simulink software to build a corresponding virtual twin vehicle model, and setting the consistency of the virtual vehicle's appearance dimension information with the real vehicle model; Step S6, using Prescan software to build a virtual simulation test platform for the target vehicle, importing virtual test scenarios into the simulation platform, adding sensor models, and traffic elements of traffic participants for simulation testing of the virtual twin vehicle; Step S7, the real autonomous driving car is tested in a closed field in combination with the virtual test scene, and its operation is detected to test the performance of the real vehicle combined with the virtual scene autonomous driving car; Step S8, realizing data interaction between the simulation test module and the real vehicle test module through the communication transmission system, specifically including: injecting the perception data of the virtual sensor model into the real vehicle controller to provide input for information fusion and decision control of the real vehicle; after receiving the information, the real vehicle controller sends the control decision instruction to the actual actuator, and feeds back the position coordinates, posture information, and vehicle trajectory data of the real vehicle to the simulation platform, and guides the simulation test of the virtual twin vehicle according to the control decision instruction of the real vehicle; Step S9, select appropriate evaluation indicators and evaluation methods, and output comparative evaluation results of the autonomous driving vehicle simulation and actual vehicle test.

8. The method for operating the virtual-real combined automatic driving vehicle testing system according to claim 7, characterized in that: In order to ensure the synchronous mapping between the virtual and real vehicles, it is necessary to achieve the consistency of the position of the real vehicle and the virtual twin vehicle, that is, the coordinate transformation of the autonomous driving vehicle in the virtual and real space. In the coordinate system of the virtual space, assuming that the coordinates of the twin vehicle are p(x, y, z), after homogeneous transformation, they are p′(x′, y′, z′), then the transformation matrix between p and p′ is: Where T represents the translation matrix: Among them, T x , T y , T z Respectively represent the translation components of the vehicle in the directions of the three coordinate axes x, y, and z; the selected angles of the rotation matrix R on the x, y, and z axes can be determined by R x , E y , R z Represents, and has the following relationship: R=R x ·R y ·R z The selection matrix R for the rotation angle α around the x-axis x for: The selection matrix R for the rotation angle β around the y-axis y for: The selection matrix R for the rotation angle γ around the z-axis z for: Through the above-mentioned coordinate transformation, it is possible to map the posture changes of the real vehicle to the corresponding vehicle twin in the virtual space.

9. The method for operating the virtual-real combined automatic driving vehicle testing system according to claim 8, characterized in that: For the simulation and actual vehicle test of the target vehicle, a comprehensive comparative evaluation is conducted around the three aspects of safety, energy saving effect and traffic efficiency. The evaluation index is determined by the empirical method, and the weight of each evaluation index is obtained by the entropy method. The specific calculation formula is as follows: Calculate the entropy value H of each indicator i After that, the weight value of each evaluation index is calculated by the weight calculation formula, which is as follows: Among them, x ij It represents the jth sample value under the i-th evaluation index. After standardizing its data, the standard value is r ij , p ij is the proportion of the jth sample value under the i-th evaluation index to the index, H i is the entropy value of the i-th evaluation index. After the entropy value of each index is determined, the weight value of each evaluation index is calculated as ω i .

10. The method for operating the virtual-real combined automatic driving vehicle testing system according to claim 9, characterized in that: Each evaluation index is normalized and combined with the weight of each evaluation index. The calculation formula of the comprehensive evaluation result is: Among them, ω i is the weight of the i-th evaluation index, S i is the normalized result of the ith evaluation index, and R is the comprehensive evaluation result.

Images