Intelligent network connection automobile test scene construction method

Abstract

The invention relates to the technical field of intelligent network connection automobile testing, and discloses an intelligent network connection automobile testing scene construction method, which comprises the following steps of 1, interpreting a design operation range of a tested automobile; 2, screening road traffic elements based on the design operation range; the road traffic elements comprise road elements, meteorological elements, traffic elements and vehicle behavior elements; 3, permutation and combination are carried out on the screened road traffic elements, and an initial scene group is obtained; and 4, screening the initial scene group, and obtaining a target test scene. According to the invention, a global coverage test scene system can be efficiently constructed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent connected vehicle testing, and in particular to a method for constructing an intelligent connected vehicle testing scenario. Background Art

[0002] With the increasing intelligence and connectivity of connected vehicles, and the widespread adoption of advanced driver assistance and autonomous driving, the "perception-decision-execution" process, once performed by drivers, is gradually being replaced by vehicles. In this context, the focus of autonomous vehicle research and development has shifted from simply achieving breakthroughs and implementation in autonomous driving technology to testing and evaluating its capabilities.

[0003] Scenarios, as models that describe the complex dynamic relationships between people, vehicles, roads, and the environment in spatial and temporal terms, are the foundation for product development and functional testing of intelligent connected vehicles. By constructing test scenarios and applying scenario-based testing methods, it is possible to accelerate the testing and evaluation of intelligent connected vehicles and quickly verify the reliability and safety of their functions. Therefore, designing a scientific and rational test scenario construction method is essential. Especially for high-level intelligent connected vehicles, statistics show that the number of potential scenarios that L3 autonomous driving systems must cope with is exponentially increasing compared to traditional vehicles. Their functional safety verification requires covering over 17 billion kilometers of test mileage, which brings new challenges to test scenario construction.

[0004] The current mainstream test scenario construction methods in the industry are mainly based on vehicle functions and regulatory requirements, and test scenarios are constructed for vehicle dynamic control and component systems. However, these methods can no longer meet the testing needs of intelligent connected vehicles. The reason is that intelligent connected vehicles have taken over the perception and decision-making tasks from the driver, and the complexity of the intelligent driving system has increased significantly. In order to verify the safety of the intelligent driving function, it needs to undergo comprehensive testing in various scenarios. In addition, the most dangerous scenarios that challenge the effectiveness of the intelligent driving system are extremely rare in natural driving conditions on open roads. Therefore, traditional real-car open road tests based on mileage cannot actually meet and guarantee the functional safety of intelligent connected vehicles. Summary of the Invention

[0005] The present invention aims to provide a method for constructing test scenarios for intelligent connected vehicles, which can efficiently construct a test scenario system with full coverage.

[0006] The basic solution provided by the present invention is: a method for constructing a test scenario for an intelligent connected vehicle, comprising the following steps:

[0007] Step 1: Interpret the design operating range of the vehicle under test;

[0008] Step 2: Screening road traffic elements based on the design operation range; the road traffic elements include road elements, weather elements, traffic elements, and vehicle behavior elements;

[0009] Step 3: Arrange and combine the selected road traffic elements to obtain an initial scene group;

[0010] Step 4: Filter the initial scene group and obtain the target test scene.

[0011] Furthermore, the method also includes step 5, designing test cases according to the target test scenario to conduct simulation tests and closed-field test verification.

[0012] Furthermore, the method further includes step 6, performing an open road test on the vehicle under test after the verification is passed.

[0013] Furthermore, in step 1, the interpretation of the design operating range of the vehicle under test includes: identifying external environmental conditions suitable for the functional operation of the intelligent driving system of the vehicle under test.

[0014] Furthermore, a combination test tool is used to arrange and combine the screened road traffic elements.

[0015] Furthermore, in step 4, the initial scene group is screened based on historical traffic accidents and extremely complex scenes.

[0016] Furthermore, in step 4, the initial scene group is screened according to the screening principles; the screening principles are as follows: screen out scenes that cannot be realized in real road traffic scenes or closed road test sites; screen out scenes that have little impact on the movement of the test vehicle among similar scenes; for scenes that have the same impact on the movement of the test vehicle, retain one scene; based on the analysis of historical traffic accident causes, screen out scenes with a higher proportion of injuries.

[0017] Furthermore, in step 3, a Monte Carlo method is used to generate boundary scenes and add them to the initial scene group.

[0018] Furthermore, in step 4, the initial scenario group is evaluated based on a scenario value evaluation function, and scenarios with value scores higher than a threshold are screened out as target test scenarios; the scenario value evaluation function includes:

[0019] V(s)=α·H(s)+β·R(s)+γ·C(s);

[0020] Among them, V(s) is the value score of the sth scenario in the initial scenario group; H(s) is the scenario information entropy, R(s) is the historical accident correlation, C(s) is the parameter coverage completeness; α, β and γ are weight values.

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

[0022] This invention provides a method for constructing test scenarios for intelligent connected vehicles. It establishes a scenario generation mechanism with rigorous mathematical logic. By introducing an ODD (operating design range)-driven scenario construction paradigm, it ensures that scenario elements precisely match the vehicle's capability boundaries, enabling two-way empowerment of test verification and function development. In terms of technical path design, element screening and scenario construction based on ODD boundary conditions not only serve test scenario generation but also reversely verify the capability boundaries of autonomous driving systems, helping to shift vehicle design from a "function implementation first" to a "safety boundary first" approach.

[0023] In addition, the present invention constructs a multi-stage collaborative scenario generation framework, which can systematically solve the problem of balancing the breadth and depth of scenario coverage through the deep integration of scenario combination and Monte Carlo sampling. In the scenario construction, a combination test tool is used to generate a basic scenario set with strict permutations and combinations to ensure the basic verification requirements of vehicle functions; combined with the Monte Carlo method, high-risk scenarios can be generated for parameter boundary conditions and high-dimensional coupling relationships through probability density sampling, breaking through the coverage limitations of traditional methods for long-tail scenarios, and then forming a comprehensive initial scenario group, and then performing scenario optimization based on this, and incorporating historical accident correlation (historical traffic accidents) and parameter coverage completeness (extremely complex scenarios) into the screening system, which is highly targeted and can cover corner scenarios, with a wide verification range and high credibility of the test results. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The figure is a flow chart of a method for constructing a test scenario for an intelligent connected vehicle according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following is a further detailed description through specific implementation methods:

[0026] Example 1

[0027] The embodiment is basically as shown in the attached Figure 1 As shown: A method for constructing a test scenario for an intelligent connected vehicle includes the following steps:

[0028] Step 1: Interpret the design operating range of the vehicle under test.

[0029] The interpretation of the design operating range of the vehicle under test includes: identifying external environmental conditions suitable for the functional operation of the intelligent driving system of the vehicle under test.

[0030] In this embodiment, typical design operating range external conditions include road and road infrastructure conditions, weather conditions, traffic targets, and lighting conditions. Road and road infrastructure conditions include highways, urban roads, rural roads, road traffic signs and markings, and traffic lights. Weather conditions include sunny, rainy, foggy, hazy, snowy, and dusty conditions. Traffic targets include motor vehicles, non-motor vehicles, pedestrians, animals, and obstacles. Lighting conditions include daytime, nighttime, and city lights.

[0031] Step 2: Screen road traffic elements based on the design operation range; the road traffic elements include road elements, meteorological elements, traffic elements and vehicle behavior elements.

[0032] The road elements include road slope, road width, road length, road line type, road surface type, road markings, etc.; the meteorological elements include light intensity, weather conditions (including temperature, humidity, meteorological conditions), etc.; the traffic elements include traffic signs, targets, etc.; the vehicle behavior elements include the movement status, movement trajectory, movement direction, etc. of vehicles and targets.

[0033] Specifically, when screening road traffic elements, the screening of road elements includes: based on the road type characteristic parameters defined in the design operating range of the tested vehicle model, excluding road element subcategories that do not meet the parameter threshold; when the design operating range is limited to closed highways, excluding road elements containing urban road characteristics, rural road characteristics, and non-closed highway characteristics;

[0034] The screening of meteorological elements includes: filtering out meteorological element combinations that exceed the threshold range based on the meteorological condition threshold parameters set in the design operating range of the tested vehicle model. When the design operating range is limited to visibility > 500 meters and rainfall < 10mm / h, meteorological characteristic data with visibility ≤ 500 meters and meteorological characteristic combinations with rainfall ≥ 10mm / h are eliminated;

[0035] The screening of traffic elements includes: removing pedestrians, non-motor vehicles and other abnormal traffic participants that exceed the parameter limits based on the traffic participant type and density parameters specified in the design operation range;

[0036] The screening of vehicle behavior elements includes: filtering out characteristic sequences of following vehicles, lane changing, and emergency braking behaviors that do not conform to the preset behavior patterns based on the speed range, acceleration extremes, and special operating condition parameters defined within the design operating range.

[0037] Step 3: Arrange and combine the selected road traffic elements to obtain an initial scene group.

[0038] The selected road traffic elements are arranged and combined using the combination test tool (PICT).

[0039] Step 4: Filter the initial scene group and obtain the target test scene.

[0040] The initial scene group is screened based on historical traffic accidents and extremely complex scenes.

[0041] The initial scenario group is screened according to the screening principles; the screening principles are as follows: screen out scenarios that cannot be realized in real road traffic scenarios or closed road test sites; screen out scenarios that have little impact on the movement of the test vehicle among similar scenarios; retain one scenario from scenarios that have the same impact on the movement of the test vehicle; and screen out scenarios with a higher proportion of injuries based on an analysis of the causes of historical traffic accidents.

[0042] Step 5: Design test cases based on the target test scenario for simulation testing and closed field testing verification.

[0043] Specifically, based on the target test scenario, we design targeted test cases around various combinations of road traffic elements, using methods such as equivalence class partitioning and boundary value analysis. For example, for the "highway - 3 lanes - green light - straight ahead" scenario, we design test cases for vehicles traveling at different speeds (speed limit, speed limit ±10%). For curve scenarios, we consider test cases with different curve radii and slope combinations. Each test case includes the test objective, test steps, expected results, and input parameters (such as the vehicle's initial position, speed, traffic flow density, etc.).

[0044] In the simulation platform, a virtual scenario is constructed based on the test case for simulation testing. During the simulation, vehicle operation data (speed, acceleration, position, heading angle) and traffic flow data (vehicle distance, following time) are collected. In-depth analysis is performed using the simulation platform's built-in data analysis tools or by exporting the data to Python, MATLAB, or other tools to determine if the test results meet expectations. If so, the simulation test is considered passed; if not, issues are flagged and the causes are analyzed.

[0045] Furthermore, based on the test case requirements, simulated road scenarios are set up within the closed area, such as simulated highway straights, urban intersections, curves, and ramps. Real traffic signs, signal lights, road markings, and other facilities are deployed, and high-precision positioning equipment (such as differential GPS) and sensors (such as lidar and cameras) are deployed for data collection.

[0046] The vehicle under test is driven into a closed area and tested according to the conditions and procedures set by the test case. During the test, sensors collect real-time vehicle status data and environmental perception data, and a high-speed camera records the entire test process for subsequent review and analysis.

[0047] The actual test results are compared with the expected results, and an evaluation is conducted based on vehicle safety (whether a collision or loss of control occurred), functionality (whether adaptive cruise control, lane keeping, and other functions function properly), and comfort (riding experience and operational responsiveness). If all of these meet the expected results, the verification is considered passed; otherwise, it is considered failed, and a review is conducted to analyze whether there are issues with the scenario design, the vehicle, or the test execution, and targeted adjustments and optimizations are made.

[0048] Step 6: After verification, conduct an open road test of the vehicle under test.

[0049] Specifically, representative open roads are selected, covering sections of different types (urban roads, expressways, highways), different traffic flows (peak, flat, low), and different climatic conditions (sunny, rainy, foggy), to conduct open road tests on the vehicles under test.

[0050] This embodiment provides a method for constructing a test scenario for an intelligent connected vehicle, which can efficiently construct a test scenario system with full coverage.

[0051] Example 2

[0052] A method for constructing a test scenario for an intelligent connected vehicle is provided, based on the first embodiment, with the following improvements.

[0053] The interpretation of the design operating range of the vehicle under test includes: identifying the external environmental conditions suitable for the functional operation of the intelligent driving system of the vehicle under test. In this embodiment, the intelligent driving system referred to is an intelligent driving system of L3 or above.

[0054] In step 3, the Monte Carlo method is also used to generate boundary scenes and add them to the initial scene group.

[0055] It includes the following sub-steps:

[0056] Based on road traffic elements, key parameters are further selected to construct an N-dimensional space Ω = {v, μ, R, I, ρ, …}; where v is the vehicle speed, μ is the road friction coefficient, R is the curve radius, I is the light intensity, and ρ is the traffic density.

[0057] Hard and soft constraints are set to ensure the physical feasibility of the scenario. In this embodiment, the hard constraint includes R being greater than a corresponding threshold. The soft constraint includes P=0.1%, where P is the probability of extreme weather.

[0058] Equal probability intervals are divided in N-dimensional space to ensure uniform coverage of each dimension, and Latin hypercube stratified sampling is performed to generate multiple alternative boundary scenarios; and the sampling weight is increased for high-risk areas (such as low friction + curves).

[0059] Based on alternative boundary scenarios, dynamic feasibility detection and sensor signal simulation are performed to screen out scenarios that are not dynamically feasible and scenarios that the vehicle's perception system cannot detect at all.

[0060] The risk entropy value of each alternative boundary scenario is calculated, and the NSGA-II multi-objective optimization algorithm is used to select the Pareto frontier scenario as the boundary scenario.

[0061] The risk entropy value x i is the discretized state of the scenario parameter combination, i is the i-th alternative boundary scenario, and M is the total number of alternative boundary scenarios.

[0062] This embodiment provides a method for constructing test scenarios for intelligent connected vehicles. Through probabilistic sampling and physical constraint guidance, it can achieve efficient mining of boundary areas in high-dimensional parameter spaces under controllable computing resources and generate reliable boundary scenarios, which can effectively improve the global coverage of test scenario construction. Compared with the orthogonal combination method in Example 1, which only covers pairwise interaction scenarios, the Monte Carlo method in this embodiment can construct a uniformly distributed set of scene points in N-dimensional space through Latin hypercube stratified sampling, greatly improving the completeness of parameter coverage. In addition, the generated boundary scenarios have both mathematical combinatorial extremes and physical feasibility, avoiding the generation of a large number of invalid scenarios in traditional methods.

[0063] Example 3

[0064] A method for constructing a test scenario for an intelligent connected vehicle is provided, based on the first embodiment, with the following improvements.

[0065] In step 4, the initial scenario group is evaluated based on the scenario value evaluation function, and scenarios with value scores higher than a threshold are screened out as target test scenarios; the scenario value evaluation function includes:

[0066] V(s)=α·H(s)+β·R(s)+γ·C(s);

[0067] Among them, V(s) is the value score of the sth scenario in the initial scenario group; H(s) is the scenario information entropy, R(s) is the historical accident correlation, C(s) is the parameter coverage completeness; α, β and γ are weight values.

[0068] This embodiment provides a method for constructing intelligent connected vehicle test scenarios, which can optimize target test scenarios based on scenario value assessment, thereby further improving the effectiveness of the constructed scenarios.

[0069] The above is only an embodiment of the present invention. Common knowledge such as the known specific structure and characteristics in the scheme is not described in detail here. Ordinary technicians in the relevant field are aware of all common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all existing technologies in the field, and have the ability to apply conventional experimental means before that date. Ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the guidance of this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several variations and improvements can be made, which should also be regarded as the scope of protection of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A method for constructing a test scenario for an intelligent connected vehicle, characterized in that: The following steps are involved: Step 1: Interpret the design operating range of the vehicle under test; Step 2: Screen road traffic elements based on the design operation range; The road traffic elements include road elements, weather elements, traffic elements and vehicle behavior elements; Step 3: Arrange and combine the selected road traffic elements to obtain an initial scene group; Step 4: Filter the initial scene group and obtain the target test scene.

2. The method for constructing a test scenario for an intelligent connected vehicle according to claim 1, wherein: The method also includes step 5, designing test cases according to the target test scenario to conduct simulation tests and closed field test verification.

3. The method for constructing a test scenario for an intelligent connected vehicle according to claim 2, wherein: The method further includes, in step 6, conducting an open road test on the vehicle under test after the verification is passed.

4. The method for constructing a test scenario for an intelligent connected vehicle according to claim 1, wherein: In step 1, the interpretation of the design operating range of the vehicle under test includes: identifying external environmental conditions suitable for the functional operation of the intelligent driving system of the vehicle under test.

5. The method for constructing a test scenario for an intelligent connected vehicle according to claim 1, wherein: A combination testing tool is used to arrange and combine the screened road traffic elements.

6. The method for constructing a test scenario for an intelligent connected vehicle according to claim 1, wherein: In step 4, the initial scene group is screened based on historical traffic accidents and extremely complex scenes.

7. The method for constructing a test scenario for an intelligent connected vehicle according to claim 1, wherein: In step 4, the initial scene group is screened according to the screening principles; the screening principles are as follows: screen out scenes that cannot be realized in real road traffic scenes or closed road test sites; screen out scenes with little impact on the movement of the test vehicle among similar scenes; retain one scene from scenes that have the same impact on the movement of the test vehicle; and screen out scenes with a higher proportion of injuries based on analysis of historical traffic accident causes.

8. The method for constructing a test scenario for an intelligent connected vehicle according to claim 1, wherein: In step 3, the Monte Carlo method is also used to generate boundary scenes and add them to the initial scene group.

9. The method for constructing a test scenario for an intelligent connected vehicle according to claim 1, wherein: In step 4, the initial scenario group is evaluated based on the scenario value evaluation function, and scenarios with value scores higher than a threshold are screened out as target test scenarios; the scenario value evaluation function includes: V(s)=α·H(s)+β·R(s)+γ·C(s); Among them, V(s) is the value score of the sth scenario in the initial scenario group; H(s) is the scenario information entropy, R(s) is the historical accident correlation, C(s) is the parameter coverage completeness; α, β and γ are weight values.

Images