Driver-in-the-loop automatic driving scene construction method and system based on virtual reality

Through virtual reality technology combining vehicle dynamics and scene simulation, we build subjective and objective related autonomous driving scenarios, solving the problem of difficult to evaluate driving comfort in the existing technology, realizing automatic generation of test scenarios that meet passenger needs, and improving automatic Safety and comfort of the driving experience.

CN118915693APending Publication Date: 2024-11-08CHINA AUTOMOTIVE ENG RES INST
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410967154.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing autonomous driving simulation testing technology is difficult to effectively combine driving comfort evaluation, and ignores the important dimension of passenger experience, which leads to the autonomous driving system being unable to meet passenger comfort needs while improving safety and performance.

Method used

Using a virtual reality-based driver-in-ring autonomous driving scenario construction system, the driver's subjectiveness is collected through the cockpit simulator, visualization system, video injection module, simulation module and the controller under test. Scores and physiological indicators, build a comfort evaluation model with subjective and objective correlation, and generate an automated driving comfort testing scenario.

Benefits of technology

It realizes subjective and objective correlation of driving comfort evaluation, improves the authenticity and reliability of autonomous driving scenarios, and can automatically generate test scenarios that meet passenger comfort needs, improving the autonomous driving experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915693A_ABST
    Figure CN118915693A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of automatic driving system testing, and discloses a driver-in-the-loop automatic driving scene construction method and system based on virtual reality, and the system comprises a testing subsystem. The test subsystem comprises a cockpit simulator, a visualization system, a video injection module, a simulation module and a tested controller; the simulation module comprises a vehicle dynamics simulation unit and a scene simulation unit; the vehicle dynamics simulation unit is used for building a vehicle dynamics model; the scene simulation unit is used for constructing a virtual vehicle running scene, calculating a scene simulation result and outputting the scene simulation result as target scene data; the video injection module is used for converting target scene data in real time and transmitting the target scene data to the tested controller; the visualization system is used for visually displaying the target scene data; the cockpit simulator is used for providing a simulated driving environment for a driver. According to the invention, subjective and objective association of driving comfort evaluation can be realized, and construction and automatic generation of a driving comfort automatic driving scene can be realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving system testing, and in particular to a method and system for constructing an autonomous driving scene based on a virtual reality driver-in-a-loop. Background Art

[0002] In recent years, autonomous driving and intelligent network technology, as pioneers of the automotive industry revolution, are driving industry change at an unprecedented speed. The widespread application of these technologies in vehicles not only heralds a fundamental transformation in transportation modes, but also has a profound impact on urban traffic management, logistics supply chain and even social and economic structure. Through advanced perception, decision-making and control technologies, autonomous vehicles can significantly improve road use efficiency and alleviate traffic congestion. At the same time, with their precise algorithms and response mechanisms, they greatly reduce the risk of traffic accidents caused by human factors, building a new line of defense for public safety. In addition, its potential in reducing energy consumption and labor costs means a significant reduction in operating costs for the public transportation and logistics industries, creating huge economic benefits for enterprises and bringing new opportunities for environmental protection and sustainable development to society.

[0003] However, the rapid development of autonomous driving technology is accompanied by a series of challenges and concerns. As the technology gradually deepens into practical applications, a series of complex safety issues and technical difficulties have gradually surfaced, including but not limited to perception failure in extreme weather, decision-making logic conflicts in complex traffic environments, and network security threats. These issues are not only related to the maturity of the technology itself, but also directly related to the safety of life and property of the public, which has attracted the deep concern of government regulators and automakers. In order to overcome these obstacles and ensure the safe and reliable advancement of technology, large-scale, multi-dimensional testing and verification has become an indispensable part.

[0004] In this context, virtual reality simulation technology, with its unique advantages, has become an indispensable auxiliary tool for the development of autonomous driving technology. Compared with traditional real-car road tests, virtual simulation tests can simulate various complex traffic scenarios in a virtual environment, which not only greatly reduces testing costs and improves testing efficiency, but also allows repeated testing of the same scenario to verify the robustness and safety of the system. The application of this technology makes it possible for the rapid iteration and optimization of autonomous driving algorithms, accelerating the technology maturity cycle.

[0005] As autonomous vehicles gradually enter the lives of the general public, consumers' expectations for intelligent driving experience are also increasing. The complexity of the autonomous driving system is reflected in multiple aspects such as deep learning of algorithms, multi-sensor fusion, and path planning. How to improve the driving experience while ensuring safety, especially to meet the subjective comfort needs of passengers, has become a new challenge. At present, although there are simulation technologies that can reproduce specific scenes and simulate random traffic flows, these tests often focus on technical performance verification and ignore the important dimension of passenger experience. The riding experience of an autonomous driving vehicle, including but not limited to driving smoothness, smooth transition during emergency avoidance, and elegant response to complex road conditions, is closely related to the comfort and trust of passengers. Therefore, building a comprehensive scenario library that can objectively reflect technical performance and subjectively reflect driving comfort has become an urgent need in the current field of autonomous driving simulation testing. Summary of the invention

[0006] The present invention aims to provide a method and system for constructing an autonomous driving scenario based on a virtual reality driver-in-the-loop, which can realize the subjective and objective correlation of driving comfort evaluation, and realize the construction and automatic generation of driving comfort autonomous driving scenarios.

[0007] To achieve the above objectives, the present invention provides the following basic solutions.

[0008] Solution 1

[0009] A system for constructing an autonomous driving scenario based on a virtual reality driver-in-the-loop includes a test subsystem and a construction subsystem; the test subsystem includes a cockpit simulator, a visualization system, a video injection module, a simulation module, and a controller under test; the simulation module includes a vehicle dynamics simulation unit and a scenario simulation unit;

[0010] The vehicle dynamics simulation unit is used to build a vehicle dynamics model and calculate vehicle motion parameters according to the input vehicle instructions; the scene simulation unit is used to build a virtual vehicle operation scene, and calculate the scene simulation results in combination with the vehicle motion parameters, and output the result as the target scene data after overall visualization rendering;

[0011] The video injection module is used to convert the target scene data in real time and transmit it to the controller under test; the visualization system is used to visualize the target scene data;

[0012] The cockpit simulator is used to provide a simulated driving environment for the driver, and is provided with a cockpit controller and a collection device for collecting the driver's physiological indicators; the cockpit controller is used to receive vehicle motion parameters to simulate the vehicle body dynamic characteristics, and is also used to collect the driver's driving instructions and transmit them to the controller under test;

[0013] The construction subsystem includes an evaluation module, an analysis module and a scenario construction module; the evaluation module is used to collect the subjective scores of the drivers in the test subsystem under different test scenarios, and construct a subjective and objective correlation comfort evaluation model to obtain comfort scores for different test scenarios; the analysis module is used to select the target test scenario according to the comfort score, and perform deconstruction analysis on the target test scenario; the scenario construction module is used to generate a comfortable automatic driving test scenario according to the deconstruction analysis results.

[0014] Solution 2

[0015] The method for constructing an autonomous driving scene based on a virtual reality driver in the loop adopts the system for constructing an autonomous driving scene based on a virtual reality driver in the loop as described in Solution 1 to construct an autonomous driving scene; the method comprises the following steps:

[0016] Step 1, construct a benchmark scene set S;

[0017] Step 2: select n test scenarios with clear non-comfort characteristics and significant takeover intention, deconstruct and analyze each test scenario, and generalize the indicators to form a benchmark scenario library S1~Sn;

[0018] Step 3: Customize the driving scenario comfort questionnaire and obtain the driver's subjective rating of comfort and physiological indicators for specific scenarios based on the benchmark scenario library;

[0019] Step 4: construct a comfort scoring model that meets the driver's needs, and conduct a driver virtual in-the-loop test on the source database based on the model, and obtain the non-comfort basic scenario set S according to the test results. U ;

[0020] Step 5: For the scene set S U The system is deconstructed and the lateral acceleration, longitudinal acceleration and TCC of the vehicle are selected as directional indicators for scene generation. According to the comfort function requirements of the tested autonomous driving system, the scene parameters are adjusted to generate a generalized comfort scene library.

[0021] The working principle and advantages of the present invention are:

[0022] First, this solution builds a simulation test environment with the driver in the loop, which can effectively connect the driver's subjective feelings with the functional evaluation of the autonomous driving system. Among them, the cockpit simulator provides the driver with a simulated driving operation environment, and the video injection module and visualization system provide the driver with a highly realistic driving visual environment. Then, the simulation model is combined with the controller under test to achieve linkage, which can effectively complete a closed-loop test with high reliability. And the acquisition device is used to monitor and collect the driver's physiological indicators to identify the driver's driving status, which can provide sufficient data support for comfort evaluation.

[0023] Second, this solution addresses the problem of lack of systematic integration of ride comfort in existing autonomous driving scenario construction methods, and provides a new evaluation scenario for ride comfort for autonomous driving evaluation. In particular, this solution creates a comfort scoring model that meets the driver's needs, and combines the in-loop test results to construct a non-comfort basic scenario set, with a high reliability of data sources; it can achieve the subjective and objective correlation of ride comfort evaluation, and realize the construction and automatic generation of ride comfort autonomous driving scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of the system structure of an embodiment of a method and system for constructing an autonomous driving scene based on a virtual reality driver in the loop of the present invention;

[0025] Figure 2 A schematic diagram of a system monitoring data collection architecture of an embodiment of a method and system for constructing an autonomous driving scenario based on a virtual reality driver in the loop of the present invention;

[0026] Figure 3 It is a schematic diagram of the operation flow of the driver-in-the-loop system in an embodiment of the method and system for constructing a driver-in-the-loop autonomous driving scenario based on virtual reality of the present invention;

[0027] Figure 4 It is a method flow chart of an embodiment of the method and system for constructing an autonomous driving scene based on a virtual reality driver in the loop of the present invention. DETAILED DESCRIPTION

[0028] The following is a further detailed description through specific implementation methods:

[0029] The embodiment is basically as shown in the attached Figure 1 , Figure 2 , Figure 3 As shown: A system is constructed based on a virtual reality driver-in-the-loop autonomous driving scenario, including a test subsystem and a construction subsystem.

[0030] The test subsystem includes a cockpit simulator, a visualization system, a video injection module, a simulation module, a controller under test and a real-time machine.

[0031] The simulation module includes a vehicle dynamics simulation unit and a scenario simulation unit. The vehicle dynamics simulation unit is used to build a vehicle dynamics model and calculate vehicle motion parameters according to the input vehicle instructions. In this embodiment, the vehicle dynamics simulation unit can be replaced by vehicle dynamics simulation software, and its parameters are adjusted and calibrated according to the actual vehicle model to meet the high-precision simulation of vehicle dynamics parameters; after the vehicle dynamics simulation unit is compiled, the real-time machine calculates various vehicle motion parameters according to the input vehicle instructions.

[0032] The scene simulation unit is used to construct a virtual vehicle operation scene (such as defining required road network elements, different obstacles and weather, etc.), and calculate the scene simulation results in combination with vehicle motion parameters, and output the result as target scene data after overall visualization rendering. In this embodiment, the scene simulation unit can be replaced by scene simulation software.

[0033] The scene simulation unit also includes a variety of sensor simulation models required for autonomous driving. In addition, each sensor simulation model is also customized according to the sensors actually used by the vehicle to ensure the simulation accuracy of the sensor simulation model. When the scene simulation unit calculates the scene simulation results in combination with the vehicle motion parameters, it also synchronously calculates the sensor simulation results of the sensor simulation model, and outputs them as target scene data after visual rendering together with the scene simulation results. Among them, the calculation results of each sensor simulation model (i.e., sensor simulation results) are converted by a real-time machine and transmitted to the controller under test through protocols such as CAN and Ethernet.

[0034] The video injection module is used to convert the target scene data in real time and transmit it to the controller under test. Specifically, the target scene data is transmitted to the video injection module via HDMI, and after the real-time conversion is completed in the video injection module, it is transmitted to the controller under test via the required interface protocol.

[0035] The visualization system is used to visualize the target scene data. Specifically, the scene simulation unit divides the environment picture in the simulated target scene data into several pictures, adjusts the size and angle, and projects them to the visualization system respectively to realize the visualization display of the target scene data. In this embodiment, the visualization system can adopt a CAVE system or a VR system to provide the driver with an immersive visual experience, which helps to improve the reliability of the test.

[0036] The cockpit simulator is used to provide a simulated driving environment for the driver. The cockpit simulator is provided with a cockpit controller and a collection device for collecting the driver's physiological indicators; the cockpit controller is used to receive vehicle motion parameters to simulate the vehicle's body dynamic characteristics, and is also used to collect the driver's driving instructions and transmit them to the controller under test to realize the driver's in-loop simulation test.

[0037] Specifically, the cockpit simulator also includes a simulation device and a hydraulic motion platform. The simulation device is mounted on the hydraulic motion platform. The simulation device includes a steering simulator, a brake pedal, an accelerator pedal, a gear controller, an instrument panel, etc., and the position layout of each device is carried out with reference to the actual vehicle; the hydraulic motion platform is a six-degree-of-freedom hydraulic motion platform that can simulate a variety of vehicle motion postures. The cockpit controller establishes control and communication connections with the simulation device, the hydraulic motion platform and the acquisition device.

[0038] The physiological indicators include eye movement indicators and physiological indicators; the eye movement indicators include gaze duration, frequency and number of long gazes (a gaze duration greater than 1.6s or 2s is determined as a long gaze); the physiological indicators include electrocardiogram parameters, electrodermal parameters, electroencephalogram parameters and pupil diameter. In this embodiment, the acquisition device includes conventional electrocardiogram, electrodermal and electroencephalogram collectors, high-definition cameras or eye trackers (for collecting eye movement indicators and pupil diameter), etc., which can fully monitor the driver's physiological state.

[0039] The construction subsystem includes an evaluation module, an analysis module and a scenario construction module.

[0040] The evaluation module is used to collect the subjective scores of the driver in the test subsystem under different test scenarios, and to construct a subjective and objective correlation comfort evaluation model to obtain comfort scores for different test scenarios. Specifically, the evaluation module is also provided with a reference scenario set; the reference scenario set is constructed according to the degree of takeover intention; the reference scenario set is used to calibrate the relationship and threshold between the driver's subjective feelings and his physiological indicators.

[0041] In this embodiment, the benchmark scene set contains scene data of multiple benchmark scenes to meet the full-scene development requirements of the autonomous driving system, so that the scenes are covered as comprehensively as possible. The benchmark scenes include forced lane changes with small gaps, illegal pedestrians crossing the road, abnormal front vehicles, multi-vehicle accidents, construction areas ahead, and road closures caused by landslides. Optionally, in actual applications, based on their own development needs, scenes with a high degree of correlation with the needs can be selected as benchmark scenes through expert experience to construct a benchmark scene set; and the benchmark scene set can be described according to the OpenX series standards.

[0042] When calibrating the relationship and threshold between the driver's subjective feelings and his physiological indicators, based on the benchmark scenario set, the driver's subjective score of comfort for the benchmark scenario is obtained for the specific benchmark scenario to confirm the driver's subjective feelings; as well as the physiological indicators collected at the takeover moment, which are used as the comfort threshold corresponding to the takeover moment of the benchmark scenario.

[0043] In the subjective scoring, the driver makes subjective scores related to comfort for different test scenarios according to the benchmark scenario set. Based on the subjective scores and physiological indicators, subjective and objective correlation indicators are constructed, and subjective and objective correlation comfort evaluation models corresponding to different driving styles are constructed through a multivariate linear regression model, so as to score the comfort of the test scenarios through the subjective and objective correlation comfort evaluation model.

[0044] In this embodiment, by customizing the driving scenario comfort questionnaire, the subjective scores of comfort of different types of subjects (such as subjects with aggressive driving style, subjects with relaxed driving style, etc.) for specific scenarios are obtained based on the benchmark scenario library. The subjective score expression is shown in Table 1.

[0045] Table 1 Subjective rating table (example)

[0046] Points Comfort level Reference Standards 1 Very uncomfortable Strong willingness to take over or direct takeover 2 Very uncomfortable Poor physical feeling or high sense of insecurity about the system 3 Discomfort Poor physical feeling or low acceptance of driving style, but acceptable 4 generally No obvious discomfort 5 Comfort Meet personal driving expectations and feel comfortable

[0047] The analysis module is used to select a target test scenario according to the comfort score, and to deconstruct and analyze the target test scenario (in this embodiment, deconstruction is performed according to the OpenX series standard). The target scenario is a test scenario with a comfort score lower than a preset threshold, that is, a driving scenario with a high degree of discomfort. In practical applications, the preset threshold can be set according to actual evaluation requirements.

[0048] When performing deconstruction analysis, the analysis module deconstructs the scene through scene elements and their related attributes; the scene elements include roads (related attributes include urban roads, highways, etc.), sign facilities (related attributes include speed limits, U-turn signs, etc., and their relative positions to the tested vehicle), tested vehicles (related attributes include initial vehicle speed, vehicle dynamics attributes, etc.), and traffic participants (related attributes include other vehicles, electric vehicles, bicycles, etc.).

[0049] The scenario construction module is used to generate a comfortable autonomous driving test scenario based on the deconstruction analysis results.

[0050] When generating a comfortable autonomous driving test scenario, the scenario construction module analyzes and summarizes the differences between different scenario data based on the deconstruction analysis results and the comfort function requirements of the tested autonomous driving system, and designs scenario classification rules; clusters the scenario data in combination with the scenario classification rules, and then calculates the probability distribution of different scenarios based on the clustering results, and randomly generates a comfortable autonomous driving test scenario based on the results. The clustering process can use a hierarchical analysis method or a K-means clustering method.

[0051] In this embodiment, the comfort function requirements of the tested autonomous driving system include driving style requirements of the autonomous driving system, including: aggressive, normal, comfortable, etc. Each driving style corresponds to a subjective and objective correlation comfort evaluation model. In practical applications, different subjective and objective correlation comfort evaluation models are used to evaluate the comfort of autonomous driving according to the actual driving style requirements.

[0052] When analyzing and summarizing the differences between different scenario data, a data-driven approach is adopted to mine the driving data corresponding to the scenario data to ensure the objectivity of the data analysis.

[0053] The set scene classification rules are shown in Table 2. Combined with the scene classification rules, feature parameters are selected from the deconstruction analysis results; then the feature parameters are standardized. In this embodiment, the parameters are encoded and standardized by one-hot encoding, and then the correlation analysis of the parameters is performed in combination with the comfort score of the scene, and independent parameters that have a greater impact on non-comfortable scenes (such as benchmark scenes) are selected as target feature parameters, such as longitudinal acceleration, lateral acceleration, target obstacle distance, etc.; then the screened non-comfortable scenes are clustered to achieve dimensionality reduction of non-comfortable scenes, and obtain a non-comfortable typical scene set that can characterize the scene construction requirements; then combined with the scene classification rules, the scene data is clustered, and then the probability distribution of different scenes is calculated according to the clustering results, and a typical scene set and feature parameters with a higher probability of occurrence are obtained, so that the feature parameters can be generalized (such as changing the relative distance and relative speed of obstacles); and based on this result, a comfortable automatic driving test scene is randomly generated.

[0054] Table 2 Scene classification rules table (example)

[0055]

[0056]

[0057] As attached Figure 4 As shown, this embodiment also provides a method for constructing an autonomous driving scene based on a virtual reality driver in the loop, using the above-mentioned system for constructing an autonomous driving scene based on a virtual reality driver in the loop to construct an autonomous driving scene; the method comprises the following steps:

[0058] Step 1: Construct a benchmark scene set S.

[0059] Specifically, a benchmark scenario set S is constructed based on source data. In this embodiment, the source data is obtained by the implementation entity based on China's large-scale natural driving data, standards and regulations, traffic accident data, and multi-source data such as functional failures. Combined with different autonomous driving function definitions and test requirements, a typical scenario library set S suitable for autonomous driving product development, performance and function test verification, and regulatory certification is constructed.

[0060] Step 2: select n test scenarios with clear non-comfort characteristics and significant takeover intentions, deconstruct and analyze each test scenario, and generalize indicators to form a benchmark scenario library S1~Sn.

[0061] In this embodiment, the selected test scenarios include "forced lane change with small gap" s1, "illegal pedestrian crossing the road" s2, "abnormal vehicle in front" s3, "multi-vehicle accident" s4, "construction area ahead" s5, "road closure due to hillside collapse" s6 and other scenarios. In actual applications, the test scenarios can be adjusted according to project implementation experience and research results. The scenarios are deconstructed according to the OpenDrive standard. Specifically, key scenario elements are selected for deconstruction. The key scenario elements include the longitudinal speed of the vehicle, longitudinal acceleration, steering wheel angle, etc. In the indicator generalization, indicators such as traffic participant type, relative lateral distance, relative longitudinal distance, relative lateral speed and relative longitudinal speed are selected for generalization, thereby forming a benchmark scenario library S1~S6.

[0062] Step 3: Customize a driving scenario comfort questionnaire to collect subjective scores of drivers with different driving scenarios and styles; obtain the driver's subjective scores on comfort for specific scenarios in the benchmark scenario library (such as a benchmark scenario) based on the benchmark scenario library, as well as the driver's physiological parameters collected by the test subsystem.

[0063] Physiological indicators here include pupil diameter, skin conductivity, heart rate, etc., as well as their statistical parameters (such as average heart rate, maximum heart rate, heart rate variability extremely low frequency power, skin conductivity level, pupil diameter, blinking frequency, etc.). These physical quantities are used to describe the driver's willingness to take over, that is, the degree of "discomfort".

[0064] Step 4: Use multiple linear regression to build a subjective and objective correlation comfort evaluation model that meets the driver's needs. Based on this model, perform a driver virtual in-the-loop test on the source database, and obtain the non-comfort basic scenario set S according to the test results. U .

[0065] The source database refers to a database formed based on China's large-scale natural driving data, standards and regulations, traffic accident data, functional failures and other multi-source data.

[0066] Step 5: Based on the Open Drive standard, the scene set S U The scene is deconstructed, and during the deconstruction, the scene elements such as roads, sign facilities, tested vehicles, traffic participants and their related attributes are deconstructed. Considering that the comfort of vehicle drivers and passengers is mainly affected by balance and vision, the lateral acceleration, longitudinal acceleration and TCC of the vehicle are selected as the directional indicators for scene generation. According to the comfort function requirements of the tested autonomous driving system, the scene parameters are adjusted (such as changing the relative distance and relative speed of obstacles, etc.) to generate a generalized comfort scene library.

[0067] The present embodiment provides a method and system for constructing an autonomous driving scene based on a virtual reality driver-in-the-loop, which can realize the subjective and objective correlation of driving comfort evaluation, and realize the construction and automatic generation of driving comfort autonomous driving scenes.

[0068] The above is only an embodiment of the present invention. The common sense such as the known specific structure and characteristics in the scheme is not described in detail here. The ordinary technicians in the relevant field are aware of all the common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement the scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for the ordinary technicians in the relevant field to implement this application. It should be pointed out that for the technicians in this field, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A system for building autonomous driving scenarios based on a virtual reality driver in the loop, characterized by: It includes a test subsystem and a construction subsystem; the test subsystem includes a cockpit simulator, a visualization system, a video injection module, a simulation module and a controller under test; the simulation module includes a vehicle dynamics simulation unit and a scene simulation unit; The vehicle dynamics simulation unit is used to build a vehicle dynamics model and calculate vehicle motion parameters according to input vehicle instructions; The scene simulation unit is used to construct a virtual vehicle operation scene, and calculate the scene simulation result in combination with the vehicle motion parameters, and output the result as the target scene data after overall visualization rendering; The video injection module is used to convert the target scene data in real time and transmit it to the controller under test; the visualization system is used to visualize the target scene data; The cockpit simulator is used to provide a simulated driving environment for the driver, and is provided with a cockpit controller and a collection device for collecting the driver's physiological indicators; the cockpit controller is used to receive vehicle motion parameters to simulate the vehicle body dynamic characteristics, and is also used to collect the driver's driving instructions and transmit them to the controller under test; The construction subsystem includes an evaluation module, an analysis module and a scenario construction module; the evaluation module is used to collect the subjective scores of the driver in the test subsystem under different test scenarios, and to construct a subjective and objective correlation comfort evaluation model to obtain comfort scores for different test scenarios; the analysis module is used to select a target test scenario according to the comfort score, and to deconstruct and analyze the target test scenario; The scenario construction module is used to generate a comfortable autonomous driving test scenario based on the deconstruction analysis results.

2. The system for constructing autonomous driving scenarios based on a virtual reality driver in a loop according to claim 1, characterized in that: The physiological indicators include eye movement indicators and physiological indicators; the eye movement indicators include gaze duration, frequency and number of long gaze times; the physiological indicators include electrocardiogram parameters, electrodermal parameters, electroencephalogram parameters and pupil diameter.

3. The system for constructing autonomous driving scenarios based on a virtual reality driver in a loop according to claim 1, characterized in that: The scene simulation unit also includes a variety of sensor simulation models required for autonomous driving; when the scene simulation unit calculates the scene simulation results in combination with the vehicle motion parameters, it also synchronously calculates the sensor simulation results of the sensor simulation model, and outputs them as target scene data after visual rendering together with the scene simulation results.

4. The system for constructing autonomous driving scenarios based on a virtual reality driver in a loop according to claim 1, characterized in that: The evaluation module is also provided with a benchmark scenario set; the benchmark scenario set is constructed according to the degree of takeover intention; the benchmark scenario set is used to calibrate the relationship and threshold between the driver's subjective feelings and his physiological indicators.

5. The system for constructing autonomous driving scenarios based on a virtual reality driver in a loop according to claim 4, characterized in that: In the subjective scoring, drivers make subjective ratings related to comfort in different test scenarios based on the benchmark scenario set.

6. The system for constructing autonomous driving scenarios based on a virtual reality driver in a loop according to claim 5, characterized in that: Based on subjective scores and physiological indicators, subjective and objective correlation indicators are constructed. A subjective and objective correlation comfort evaluation model corresponding to different driving styles is constructed through a multivariate linear regression model, so as to score the comfort of the test scenarios through the subjective and objective correlation comfort evaluation model.

7. The system for constructing autonomous driving scenarios based on a virtual reality driver in a loop according to claim 1, characterized in that: The target scene is a test scene whose comfort score is lower than a preset threshold.

8. The system for constructing autonomous driving scenarios based on a virtual reality driver in a loop according to claim 1, characterized in that: When performing deconstruction analysis, the analysis module deconstructs the scene through scene elements and their related attributes; the scene elements include roads, sign facilities, tested vehicles, and traffic participants.

9. The system for constructing autonomous driving scenarios based on a virtual reality driver in a loop according to claim 1, characterized in that: When generating a comfortable autonomous driving test scenario, the scenario construction module analyzes and summarizes the differences between different scenario data based on the deconstruction analysis results and the comfort function requirements of the autonomous driving system under test, and designs scenario classification rules; clusters the scenario data in combination with the scenario classification rules, and then calculates the probability distribution of different scenarios based on the clustering results, and randomly generates a comfortable autonomous driving test scenario based on this result.

10. A method for constructing an autonomous driving scenario based on a virtual reality driver-in-the-loop, characterized in that: The autonomous driving scene construction system based on virtual reality driver in the loop as claimed in any one of claims 1 to 9 is used to construct the autonomous driving scene; the steps include: Step 1, construct a benchmark scene set S; Step 2: select n test scenarios with clear non-comfort characteristics and significant takeover intention, deconstruct and analyze each test scenario, and generalize the indicators to form a benchmark scenario library S1~Sn; Step 3: Customize a driving scenario comfort questionnaire to collect subjective scores of drivers with different driving scenarios and styles; obtain the subjective scores of drivers on comfort for specific scenarios in the benchmark scenario library based on the benchmark scenario library, as well as the driver's physiological parameters collected by the test subsystem; Step 4: construct a subjective and objective correlation comfort evaluation model that meets the driver's needs, and conduct a driver virtual in-the-loop test on the source database based on the model. According to the test results, the non-comfortable basic scenario set S is obtained. U ; Step 5: For the scene set S U The system is deconstructed and the lateral acceleration, longitudinal acceleration and TCC of the vehicle are selected as directional indicators for scene generation. According to the comfort function requirements of the tested autonomous driving system, the scene parameters are adjusted to generate a generalized comfort scene library.

Citation Information

Cited By

  • Method and system for testing acceleration performance of unmanned motorcycle under extreme working condition

    CN119845597A

  • Test method and system for acceleration performance of unmanned motorcycle under extreme working condition

    CN119845597B

  • Scene generation method and system for testing and evaluating automatic driving cold-chain logistics vehicle

    CN122366191A