CN116361175A - Method for creating test scenes of automatic driving vehicles in different safety domains - Google Patents
Method for creating test scenes of automatic driving vehicles in different safety domains Download PDFInfo
- Publication number
- CN116361175A CN116361175A CN202310338550.3A CN202310338550A CN116361175A CN 116361175 A CN116361175 A CN 116361175A CN 202310338550 A CN202310338550 A CN 202310338550A CN 116361175 A CN116361175 A CN 116361175A
- Authority
- CN
- China
- Prior art keywords
- scene
- information
- feature
- scenes
- classification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012360 testing method Methods 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000013145 classification model Methods 0.000 claims abstract description 11
- 238000011156 evaluation Methods 0.000 claims abstract description 11
- 238000004140 cleaning Methods 0.000 claims abstract description 8
- 238000012706 support-vector machine Methods 0.000 claims abstract description 7
- 230000008030 elimination Effects 0.000 claims abstract description 6
- 238000003379 elimination reaction Methods 0.000 claims abstract description 6
- 238000000605 extraction Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 9
- 230000003068 static effect Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 5
- 230000001133 acceleration Effects 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 4
- 230000007613 environmental effect Effects 0.000 claims description 4
- 238000002955 isolation Methods 0.000 claims description 3
- 238000010998 test method Methods 0.000 claims description 3
- 238000004088 simulation Methods 0.000 abstract description 3
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000007635 classification algorithm Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 238000009411 base construction Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010224 classification analysis Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
- G06F11/3672—Test management
- G06F11/3684—Test management for test design, e.g. generating new test cases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Software Systems (AREA)
- Strategic Management (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Economics (AREA)
- Computer Hardware Design (AREA)
- Game Theory and Decision Science (AREA)
- General Engineering & Computer Science (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a method for creating a test scene of an automatic driving vehicle in different safety domains, which comprises the following steps: 1) Natural driving data are obtained, data cleaning and feature extraction are carried out, and scene key information is formed into all feature sets; 2) Removing partial variables of all feature sets, and selecting an optimal feature subset of the simplified attribute set by adopting a recursive feature elimination method based on a support vector machine; 3) Constructing classification models by adopting different classifiers, and selecting an optimal classification model based on the true rate and the false positive rate as evaluation indexes; 4) Adding other attributes into the complex scene, applying an association rule method on the data subset, and carrying out scene refinement on part of the complex scene; 5) And (3) constructing test cases for the classified scenes, setting different safety thresholds on the basis of meeting coverage rate and complexity, and randomly generating three scenes of a SOTIF domain, a critical accident domain and an accident domain. The invention can meet the simulation test requirements of the automatic driving automobile in different scenes and different safety domains.
Description
Technical Field
The invention relates to the field of automatic driving vehicle testing, in particular to a method for secondarily dividing scenes and generating test cases of different safety domains.
Background
With the advancement of autopilot functionality, a major challenge is safe and efficient operation under complex traffic conditions (e.g., road intersections). This requires comprehensive testing, whether in a virtual simulation environment or under real world testing conditions. Therefore, directional construction of the autopilot scenario test case is required. The automatic driving test scene generation comprises two parts, namely scene analysis and test case generation, wherein the scene analysis is to extract a typical logic scene by adopting a multivariate statistical analysis method; the test case generation is based on the main parameters of the logic scene and the probability distribution model thereof, and parameter values are extracted by adopting random sampling, importance sampling and a machine learning algorithm model, so that the reconstruction and the derivation of the specific scene are realized.
In the proposed autopilot function assessment method, it is common to assess the function under test in the relevant scenario. It is therefore necessary to refine the classification of the scene. In the traditional scene classification frame, a middle steam center firstly proposes automatic division of driving scenes based on an inference engine, namely, a scene rule base is constructed based on expert experience and laws and regulations, specific rules are formulated according to different scenes, and basis is provided for automatic division of scenes. Because of the limitation of the rule base construction, it is proposed to classify driving scenes by using a machine learning algorithm, generate scene classifiers by training, and iteratively update the types and parameters of the classifiers by using the collected data to obtain more comprehensive classification results. However, the classification result is limited to parallel road scenes, the complex intersection scenes are not accurately classified, if the complex intersection scenes are accurately classified, the automatic feature selection classification method is used simply to increase the calculation amount of classification, and the training of each subdivision scene classifier is time-consuming and labor-consuming.
Regarding the generation of test cases, duan et al constructs a complexity-based test case generation algorithm, calculates the corresponding complexity when parameters take different values by using an Analytic Hierarchy Process (AHP), designs a test case generation algorithm (CTBC) considering the complexity, so that the generated test cases are more complex, the algorithm is finally applied to the test of the LDW function, and the test result shows that the complex test cases are easier to find faults in a tested system, thereby improving the test efficiency. But the model cannot generate use cases corresponding to different security domains, especially the current research on the field of intended functional security.
Disclosure of Invention
The invention aims to solve the defects in the prior art, and provides a method for creating a test scene of an automatic driving vehicle in different safety domains, so that the simulation test requirement of the automatic driving vehicle in different safety domains in different scenes can be met, and the simulation test of the automatic driving vehicle is facilitated.
In order to achieve the aim of the invention, the invention adopts the following technical scheme:
the invention relates to a method for creating a test scene of an automatic driving vehicle in different safety domains, which is characterized by comprising the following steps: the method comprises the following steps:
(1) Acquiring a natural driving data set, performing data cleaning and feature extraction to obtain scene key information and forming all feature sets; the scene key information comprises object information, road information, traffic facility information and environment information;
(2) Deleting the environment information and part of object information in all feature sets to obtain simplified scene key information for carrying out initial classification on scenes in the natural driving data set to obtain initial scene classification results, wherein the initial scene classification results comprise parallel road scenes and crossroad scenes; wherein, the parallel road scene includes: a free running scene, a following car running scene, an active lane changing scene and a cut-in scene; the intersection scene includes: the system comprises a left turn scene without a left turn signal lamp, a left turn scene with a left turn signal lamp, a turning scene, a straight-through crossing scene and a right turn scene;
selecting each type of scene in the initial scene classification result by adopting a support vector machine (SVM-RFE) -based recursive feature elimination method to obtain an optimal feature subset of each type of scene;
(3) Based on the optimal feature subset of each type of scene, constructing and training a classification model of each type of scene by utilizing different classifiers, and selecting an optimal classification model of each type of scene by taking a true rate and a false positive rate as evaluation indexes;
(4) Adding the deleted environment information and part of object information into the optimal feature subsets of the corresponding class of scenes to obtain a rough-classified scene feature subset, so that scene refinement is carried out on the crossroad scenes based on the rough-classified scene feature subset by applying a correlation rule method to obtain the crossroad scenes after refinement classification;
(5) Based on the intersection scenes after refinement and classification, different test cases are generated by using a combined test method;
setting a desired functional safety threshold W 1 And an accident critical threshold W 2 And taking a relative driving safety index RDSI as an evaluation index; if the evaluation index RDSI of the ith test case i <W 1 Dividing the ith test case into expected functional security domains; if W is 1 <RDSI i <W 2 Dividing the ith test case into a critical accident scene domain; if RDSI is used i >W 2 The ith test case is partitioned into the incident domain.
The method for creating the test scene of the automatic driving vehicle in different safety domains is also characterized in that: the object information in the step (1) is the geometric and motion characteristics of the vehicle and the object vehicle, and the method comprises the following steps: longitudinal running speed v of bicycle s Longitudinal acceleration a s Time to collision TTC, minimum line distance TLC min Relative front vehicleSpeed gain and target car lane change identification information;
the road information includes: width d of each lane line t Lane type, road range;
the traffic facility information is divided into static facility layer information and temporary facility layer information, wherein the static facility layer information comprises light pole information, isolation zone information and traffic light information; temporary facility layer information refers to traffic facility information temporarily changed due to construction;
the environmental information comprises weather conditions, visibility and road surface wet and slippery degree.
The association rule method in the step (4) is to search the characteristics with higher correlation with the optimal characteristic subset of each type of scene in the crossroad scene in the deleted information based on the association rule, and specifically comprises the following steps:
assuming that the A features are features added into an optimal feature subset in the crossroad scene, and the B feature set is an optimal feature subset of a certain type of scene in the crossroad scene;
if the Lift of the A feature to the B feature set is greater than the Lift threshold value Lift min And considering that the A features have a pushing effect on the occurrence of the B feature set, adding the A features into a scene corresponding to the B feature set to subdivide the scene, and otherwise, deleting the A features.
The electronic device of the invention comprises a memory and a processor, wherein the memory is used for storing a program for supporting the processor to execute any one of the test scene creation methods, and the processor is configured to execute the program stored in the memory.
The invention relates to a computer readable storage medium, on which a computer program is stored, characterized in that the computer program when run by a processor performs the steps of any of the test scene creation methods.
Compared with the prior art, the invention has the beneficial effects that:
1. in the invention, when the field Jing Chushi is classified, partial variables of all feature sets are removed, and the optimal feature subset is selected for the simplified attribute set by adopting a recursive feature elimination method based on a support vector machine. The simplified data set can reduce the huge calculation amount of the initial classification of the scene, and if an association rule method is used during the initial classification, the association rule is applied to the whole data set, so that the number of rules is reduced, the negative influence of environment information and part of object information on the classification of the scene is obviously reduced, and the reliability and the effectiveness of the classification analysis result are ensured.
2. The initial classified scenes are not suitable for automatic driving vehicle specific test due to the fact that the range is too large, and the method is applied to complex scene subclasses by adopting the association rule algorithm, so that scene classification is refined, and training and evaluation of all subsystems and the whole vehicle are facilitated.
3. With the deep research of the safety field of the expected functions of the intelligent driving automobile, the invention provides the test cases for randomly generating different safety domains, and the safety threshold takes the relative driving safety index RDSI as an evaluation index, thereby meeting the requirements of the test cases of different safety domains.
Drawings
FIG. 1 is a block diagram of the architecture of the present system;
FIG. 2 is a schematic illustration of a specific scenario featuring straight-ahead segmentation of a lane-subtended motor vehicle according to association rules;
FIG. 3 is a schematic illustration of a specific scenario featuring a right turn of a oncoming lane vehicle according to association rules;
FIG. 4 is a schematic illustration of a specific scenario featuring a straight-through division into lanes of non-motor vehicles according to association rules;
fig. 5 is a left turn scene association rule directed graph without left turn signal lights.
Detailed Description
In this embodiment, as shown in fig. 1, a method for creating a test scenario of an autopilot vehicle in different safety domains includes the following steps:
step 1, acquiring natural driving data, performing data cleaning and feature extraction, and forming all feature sets from scene key information;
step 1.1, acquiring natural driving data, road traffic accident data, driver examination data, closed test field test data and open road test data as an original scene data set;
and 1.2, analyzing scene data which does not meet the scene data cleaning standard in the original scene data by using methods such as statistics, data mining and the like according to the acquired original scene data. Analyzing the data problems by manual measurement or a professional data analysis program;
step 1.3, according to the data analysis result, formulating corresponding data cleaning rules including but not limited to checking and processing of data validity, checking and processing of data accuracy, checking and processing of data perfection, checking and processing of data consistency;
step 1.4, executing a data cleaning flow according to the data cleaning rule to obtain scene key information including object information, road information, traffic facility information and environment information;
wherein, the object information is the geometry and motion characteristics of the own vehicle and the object vehicle, comprising: longitudinal running speed v of bicycle s Longitudinal acceleration a s Time to collision TTC, minimum line distance TLC min Gain relative to the front vehicle speed and the target vehicle lane change identification information;
the road information includes: width d of each lane line t Lane type, road range;
the traffic facility information is divided into static facility layer information and temporary facility layer information, wherein the static facility layer information comprises light pole information, isolation zone information and traffic light information; temporary facility layer information refers to traffic facility information temporarily changed due to construction;
the environmental information includes weather conditions, visibility, and road surface slippery degree.
Step 2, deleting the environment information and part of object information in all feature sets to obtain simplified scene key information, and carrying out initial scene classification to obtain an initial scene classification result; selecting each type of scene in the initial scene classification result by adopting a support vector machine (SVM-RFE) -based recursive feature elimination method to obtain an optimal feature subset of each type of scene;
step 2.1 is to reduce the calculation amount of the initial classification of the scene, reduce part of the characteristics of the scene, and reduce the principle of the characteristic attribute of the scene so as not to influence the initial classification result. Reducing environmental information and partial object information, such as lane line width dt, weather conditions, visibility, road surface wet and slippery degree and the like, excluding low variance features, grouping or combining highly relevant features, features with unknown values in 30% of samples and the like;
step 2.2, carrying out initial scene classification on the simplified scene key information to obtain an initial scene classification result, wherein the initial scene classification result comprises a parallel road scene and an intersection scene; wherein, parallel road scene includes: a free running scene, a following car running scene, an active lane changing scene and a cut-in scene; the crossroad scene includes: the system comprises a left turn scene without a left turn signal lamp, a left turn scene with a left turn signal lamp, a turning scene, a straight-through crossing scene and a right turn scene;
step 2.3 in this embodiment, taking a left turn scene of a signal light intersection without left turn as an example, the optimal feature subset selection and classifier training under the scene are performed. The simplified attribute set comprises the pre-collision time TTC and the longitudinal running speed v of the target vehicle i Longitudinal running speed v of vehicle s Longitudinal acceleration a s Steering angle gamma, target vehicle lane change identification, etc. And selecting the optimal feature subset of the signal lamp scene without left turn by adopting a support vector machine (SVM-RFE) recursive feature elimination method. Firstly, training an SVM classifier by taking a classification result as a target correctly, and searching an optimal classification hyperplane. Selecting a criterion that the classification error rate after removing the feature is the feature ordering, and performing score ordering on each feature, wherein the ordering is shown in a table 1, removing the feature with the minimum feature score, namely the pre-collision time TTC, and training the classification model corresponding to the scene again by using the residual features, and performing the next iteration until the residual classification error rate is higher than 25%, so as to obtain an optimal feature subset.
Table 1 left turn scene feature ordering for intersections without left turn signal lamps
Step 3, constructing and training classification models of each type of scene by utilizing different classifiers based on the optimal feature subset of each type of scene, and selecting the optimal classification model of each type of scene by taking the true rate and the false positive rate as evaluation indexes;
in this embodiment, taking an optimal feature subset of a left-turn scene at a signal light intersection without left-turn as an example, a classifier of the scene is trained. And respectively constructing classification models by adopting a naive Bayesian classification algorithm, a support vector machine classification algorithm, a K nearest neighbor classification algorithm and other algorithms, and evaluating the effects of different classifiers by adopting two indexes of true rate and false positive rate.
In the formulas (1) and (2), TPR represents the true rate, namely the proportion of the scene correctly divided into the scenes by the classifier; FPR represents false positive rate, i.e. the proportion of non-class scenes which are divided into class scenes by the classifier error; TN is the number of scenes correctly divided into the types; TN is the number of scenes which are correctly divided into non-class scenes; FP is the number of non-such scenes divided into such scenes in error; FN is the number of the scenes which are wrongly divided into the non-scenes.
Finally, the classification model obtained by adopting a naive Bayes classification algorithm has the advantages that the classification true rate of the left turn scene at the intersection without the left turn signal lamp is 95%, the classification false positive rate is 6%, and the effect is best.
Step 4, adding the deleted environment information and part of object information into the optimal feature subsets of the corresponding class of scenes to obtain a rough-classified scene feature subset, and searching for features with higher correlation with the optimal feature subset of each class of scenes in the crossroad scenes by applying a correlation rule method in the deleted information based on the rough-classified scene feature subset to obtain the crossroad scenes after refining classification;
in this embodiment, the left turn scene of the intersection without the left turn signal lamp is refined and classified, the reduced features are supplemented into the classified simplified data set, the left turn scene data of the intersection without the left turn signal lamp is further analyzed, and other features with higher relevance to the feature subset of the scene are found based on the association rule. Determining a minimum support value Sup min Selecting a lower threshold increases the computation time and rules, and selecting a higher threshold may ignore information about the item set, where Sup is selected min Is 0.03, so that all term sets that occur in less than 3% of the samples are ignored. Determining a Lift threshold Lift min 1.25, when the Lift of the A feature to the B feature set is greater than the Lift threshold value Lift min The a features are considered to have a pushing effect on the occurrence of the B feature set, and then the a features are added to this scene for further subdivision of the scene.
Table 2 rules for association of signal lamp intersections without left turn
A schematic diagram of a specific scenario divided according to association rules 1, 2, 3 is shown in fig. 2.
A schematic diagram of a specific scenario divided according to the association rules 4, 5 is shown in fig. 3.
A schematic diagram of a specific scenario divided according to the association rules 6, 7 is shown in fig. 4.
The association rules in the table are further visualized by a directed graph, fig. 5, the nodes of the directed graph representing the association rules, the weight or thickness of each edge representing the weight of the respective front node and center node association rules, which means that the node with the bold edge represents the main no left turn signal light intersection left turn scene, thereby defining a refined scene.
Step 5, based on the intersection scene after the refinement and classification, generating different test cases by using a combined test method;
setting a desired functional safety threshold W 1 And an accident critical threshold W 2 And taking a relative driving safety index RDSI as an evaluation index; if the ith testEvaluation index RDSI of use case i <W 1 Dividing the ith test case into expected functional security domains; if W is 1 <RDSI i <W 2 Dividing the ith test case into a critical accident scene domain; if RDSI is used i >W 2 The ith test case is partitioned into the incident domain.
In this embodiment, an electronic device includes a memory for storing a program supporting the processor to execute the above method, and a processor configured to execute the program stored in the memory.
In this embodiment, a computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the method described above.
Claims (5)
1. A method for creating test scenes of an automatic driving vehicle in different safety domains is characterized by comprising the following steps of: the method comprises the following steps:
(1) Acquiring a natural driving data set, performing data cleaning and feature extraction to obtain scene key information and forming all feature sets; the scene key information comprises object information, road information, traffic facility information and environment information;
(2) Deleting the environment information and part of object information in all feature sets to obtain simplified scene key information for carrying out initial classification on scenes in the natural driving data set to obtain initial scene classification results, wherein the initial scene classification results comprise parallel road scenes and crossroad scenes; wherein, the parallel road scene includes: a free running scene, a following car running scene, an active lane changing scene and a cut-in scene; the intersection scene includes: the system comprises a left turn scene without a left turn signal lamp, a left turn scene with a left turn signal lamp, a turning scene, a straight-through crossing scene and a right turn scene;
selecting each type of scene in the initial scene classification result by adopting a support vector machine (SVM-RFE) -based recursive feature elimination method to obtain an optimal feature subset of each type of scene;
(3) Based on the optimal feature subset of each type of scene, constructing and training a classification model of each type of scene by utilizing different classifiers, and selecting an optimal classification model of each type of scene by taking a true rate and a false positive rate as evaluation indexes;
(4) Adding the deleted environment information and part of object information into the optimal feature subsets of the corresponding class of scenes to obtain a rough-classified scene feature subset, so that scene refinement is carried out on the crossroad scenes based on the rough-classified scene feature subset by applying a correlation rule method to obtain the crossroad scenes after refinement classification;
(5) Based on the intersection scenes after refinement and classification, different test cases are generated by using a combined test method;
setting a desired functional safety threshold W 1 And an accident critical threshold W 2 And taking a relative driving safety index RDSI as an evaluation index; if the evaluation index RDSI of the ith test case i <W 1 Dividing the ith test case into expected functional security domains; if W is 1 <RDSI i <W 2 Dividing the ith test case into a critical accident scene domain; if RDSI is used i >W 2 The ith test case is partitioned into the incident domain.
2. The method for creating a test scenario for an autonomous vehicle in different safety domains according to claim 1, wherein: the object information in the step (1) is the geometric and motion characteristics of the vehicle and the object vehicle, and the method comprises the following steps: longitudinal running speed v of bicycle s Longitudinal acceleration a s Time to collision TTC, minimum line distance TTC min Gain relative to the front vehicle speed and the target vehicle lane change identification information;
the road information includes: width d of each lane line t Lane type, road range;
the traffic facility information is divided into static facility layer information and temporary facility layer information, wherein the static facility layer information comprises light pole information, isolation zone information and traffic light information; temporary facility layer information refers to traffic facility information temporarily changed due to construction;
the environmental information comprises weather conditions, visibility and road surface wet and slippery degree.
3. The method for creating a test scenario for an autonomous vehicle in different safety domains according to claim 1, wherein: the association rule method in the step (4) is to search the characteristics with higher correlation with the optimal characteristic subset of each type of scene in the crossroad scene in the deleted information based on the association rule, and specifically comprises the following steps:
assuming that the A features are features added into an optimal feature subset in the crossroad scene, and the B feature set is an optimal feature subset of a certain type of scene in the crossroad scene;
if the Lift of the A feature to the B feature set is greater than the Lift threshold value Lift min And considering that the A features have a pushing effect on the occurrence of the B feature set, adding the A features into a scene corresponding to the B feature set to subdivide the scene, and otherwise, deleting the A features.
4. An electronic device comprising a memory and a processor, wherein the memory is for storing a program supporting the processor to execute the test scenario creation method of any one of claims 1-3, the processor being configured to execute the program stored in the memory.
5. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when run by a processor performs the steps of the test scenario creation method of any one of claims 1-3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310338550.3A CN116361175A (en) | 2023-03-31 | 2023-03-31 | Method for creating test scenes of automatic driving vehicles in different safety domains |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310338550.3A CN116361175A (en) | 2023-03-31 | 2023-03-31 | Method for creating test scenes of automatic driving vehicles in different safety domains |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116361175A true CN116361175A (en) | 2023-06-30 |
Family
ID=86907068
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310338550.3A Pending CN116361175A (en) | 2023-03-31 | 2023-03-31 | Method for creating test scenes of automatic driving vehicles in different safety domains |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116361175A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117719440A (en) * | 2024-02-08 | 2024-03-19 | 零束科技有限公司 | Automobile signal detection method, system and readable storage medium |
-
2023
- 2023-03-31 CN CN202310338550.3A patent/CN116361175A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117719440A (en) * | 2024-02-08 | 2024-03-19 | 零束科技有限公司 | Automobile signal detection method, system and readable storage medium |
| CN117719440B (en) * | 2024-02-08 | 2024-05-03 | 零束科技有限公司 | Automobile signal detection method, system and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110949398B (en) | Method for detecting abnormal driving behavior of first-vehicle drivers in vehicle formation driving | |
| Peng et al. | Uncertainty evaluation of object detection algorithms for autonomous vehicles | |
| CN107368787A (en) | A kind of Traffic Sign Recognition algorithm that application is driven towards depth intelligence | |
| Kuang et al. | Bayes saliency-based object proposal generator for nighttime traffic images | |
| CN112793576B (en) | Lane change decision method and system based on rule and machine learning fusion | |
| CN108960074B (en) | Small-size pedestrian target detection method based on deep learning | |
| CN112668779A (en) | Preceding vehicle motion state prediction method based on self-adaptive Gaussian process | |
| CN116361175A (en) | Method for creating test scenes of automatic driving vehicles in different safety domains | |
| He et al. | Probabilistic intention prediction and trajectory generation based on dynamic bayesian networks | |
| CN112883991A (en) | Object classification method, object classification circuit and motor vehicle | |
| Xue et al. | A context-aware framework for risky driving behavior evaluation based on trajectory data | |
| Yebes et al. | Learning to automatically catch potholes in worldwide road scene images | |
| Bouhoute et al. | On the application of machine learning for cut-in maneuver recognition in platooning scenarios | |
| Yahaya et al. | Ensemble-based model selection for imbalanced data to investigate the contributing factors to multiple fatality road crashes in Ghana | |
| Koh et al. | Autonomous road potholes detection on video | |
| CN117809458A (en) | Real-time assessment method and system for traffic accident risk | |
| CN114428717A (en) | Automatic driving test scene set updating method and system based on real data | |
| US20230350664A1 (en) | Methods and apparatus for automatically labeling data processing events in autonomous driving vehicles via machine learning | |
| Li | A scenario-based development framework for autonomous driving | |
| Sivaraman et al. | Observing on-road vehicle behavior: Issues, approaches, and perspectives | |
| CN115860461A (en) | Risk factor evaluation method for traffic conflict of non-motor vehicles at plane intersection | |
| Wen et al. | Modeling human driver behaviors when following autonomous vehicles: An inverse reinforcement learning approach | |
| Xie et al. | Developing a machine learning-based approach for predicting road surface friction using dash camera images—a City of Edmonton, Canada, case study | |
| CN114822044A (en) | Driving safety early warning method and device based on tunnel | |
| US20240092394A1 (en) | Generating unknown-unsafe scenarios, improving automated vehicles, and computer system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |