Large-scale traffic flow and road network simulation and reproduction method and system in vehicle-road collaboration scenarios

Abstract

The present invention relates to the field of automatic driving simulation technology, and discloses a method and system for simulating and reproducing large-scale traffic flow and road network in a vehicle-road cooperative scenario, wherein the method comprises the following steps: step 1, collecting basic data; step 2, converting basic data and obtaining target data; step 3, performing data fusion and preliminary screening between roads; then, performing preliminary screening of abnormal scenes; obtaining preliminary screening data; step 4, based on the preliminary screening data, performing trajectory evaluation, and screening out scene data with high-quality trajectories; step 5, based on the preliminary screening data, performing collision risk evaluation, obtaining collision risk rating, and screening out scene data without collision risk; step 6, performing scene element identification; step 7, performing scene classification, and storing the classified scenes in a scene library. The present invention can automatically complete scene-based data processing and classification using real data, and can provide real scene data for traffic flow and road network simulation and reproduction.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving simulation technology, and specifically to a method and system for simulating and reproducing large-scale traffic flow and road network in a vehicle-road collaboration scenario. Background Art

[0002] With the continuous development of autonomous driving technology, more and more vehicles equipped with autonomous driving systems are appearing in the vehicle sales market. The increasingly mature autonomous driving technology has had a huge impact on people's driving experience, and has improved safety and comfort to varying degrees. In order to make the autonomous driving system show better functions and performance levels, system-level simulation testing is essential and crucial in the development process of the autonomous driving system.

[0003] Autonomous driving simulation testing is a test verification method that uses effective test cases to quantify parameters and confirm the effects of the autonomous driving system or system components (perception unit, decision unit, execution unit) in a simulation environment based on the simulation test scenario. Simulation testing has the characteristics of authenticity, safety, accuracy, etc., and has gradually become an indispensable part of the research and development process of the autonomous driving vehicle industry.

[0004] Existing simulation scene construction methods usually require more complex calculations and are virtual scenes, which leads to bottlenecks in their construction efficiency and poor scene realism. Therefore, how to more efficiently provide a simulation scene construction method with higher realism has become a technical problem that urgently needs to be solved in the industry. Summary of the invention

[0005] The present invention aims to provide a method and system for simulating and reproducing large-scale traffic flow and road network in a vehicle-road collaboration scenario, which can automatically complete scenario-based data processing and classification using real data, and can provide real scenario data for traffic flow and road network simulation and reproduction.

[0006] To achieve the above objectives, the present invention provides the following basic solutions.

[0007] Solution 1

[0008] The large-scale traffic flow and road network simulation and reproduction method under the vehicle-road collaboration scenario includes the following steps:

[0009] Step 1, collecting basic data; the basic data includes map data, trajectory data and event data;

[0010] Step 2, converting the basic data and obtaining the target data; the target data includes Opendrive data and Opensenario data;

[0011] Step 3: Based on the target data, data fusion and preliminary screening are performed between roads; then, preliminary screening of abnormal scenes is performed to obtain preliminary screening data;

[0012] Step 4: Based on the initial screening data, perform trajectory evaluation and select scene data with high-quality trajectories;

[0013] Step 5: Based on the initial screening data, a collision risk assessment is performed to obtain a collision risk rating, and scene data without collision risk is screened out;

[0014] Step 6, based on the preliminary screening data and the processing results of steps 4 to 5, scene elements are identified; the scene elements include map information, traffic participant behavior, traffic participant type, trajectory quality, collision level and traffic flow;

[0015] Step 7: Based on the initial screening data and the processing results of step 6, the scenes are classified and stored in the scene library.

[0016] Solution 2

[0017] A large-scale traffic flow and road network simulation and reproduction system in a vehicle-road cooperative scenario, which is used to execute the large-scale traffic flow and road network simulation and reproduction method in a vehicle-road cooperative scenario as described in Scheme 1; comprising an acquisition module, a conversion module, a data processing module and a classification module;

[0018] The acquisition module is used to acquire basic data; the conversion module is used to convert basic data and obtain target data; the data processing module is used to perform data fusion and preliminary screening between roads based on the target data; then, perform preliminary screening of abnormal scenes; obtain preliminary screening data; and, based on the preliminary screening data, perform trajectory evaluation and screen out scene data with high-quality trajectories; and, based on the preliminary screening data, perform collision risk evaluation, obtain collision risk rating, and screen out scene data with no collision risk; and, perform scene element identification; the classification module is used to classify scenes based on the processing results of the data processing module, and store the classified scenes in a scene library.

[0019] The working principle and advantages of the present invention are:

[0020] First, this solution performs data analysis and scene classification based on real data. Compared with conventional virtual simulation scenes, this solution can provide more realistic data for scene construction, and the simulation scene has a higher degree of realism. Second, this solution can automatically screen and process the collected basic data; based on the connection between the basic data, it can automatically merge the data and automatically correct and screen the data with large errors, data that does not conform to the actual situation, and data that is partially missing, and the data reliability is high. Furthermore, this solution can also automatically complete trajectory evaluation and collision risk evaluation, and can provide a reference benchmark for subsequent scene classification. Third, this solution can realize multi-dimensional automatic scene classification, which is convenient for efficient completion of targeted scene search in the simulation stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic diagram of a method flow of an embodiment of a method and system for simulating and reproducing large-scale traffic flow and road network in a vehicle-road collaboration scenario of the present invention;

[0022] Figure 2 A data fusion diagram of a method and system embodiment for simulating and reproducing large-scale traffic flow and road network in a vehicle-road collaboration scenario of the present invention;

[0023] Figure 3 A schematic diagram of scene anomaly classification in an embodiment of a method and system for simulating and reproducing large-scale traffic flow and road network in a vehicle-road collaboration scenario of the present invention;

[0024] Figure 4 It is a schematic diagram of trajectory quality assessment of an embodiment of a method and system for simulating and reproducing large-scale traffic flow and road network in a vehicle-road collaboration scenario of the present invention;

[0025] Figure 5 A schematic diagram of scene processing classification of a method and system embodiment of large-scale traffic flow and road network simulation reproduction in a vehicle-road collaboration scenario of the present invention;

[0026] Figure 6 Schematic diagram of a scene classification method for a large-scale traffic flow and road network simulation and reproduction method and system implementation example in a vehicle-road collaboration scenario of the present invention. DETAILED DESCRIPTION

[0027] The following is a further detailed description through specific implementation methods:

[0028] The embodiment is basically as shown in the attached Figure 1 As shown: A large-scale traffic flow and road network simulation and reproduction method in a vehicle-road collaboration scenario includes the following steps:

[0029] Step 1, collecting basic data; the basic data includes map data, trajectory data and event data.

[0030] In this step, the map data is collected by the collection vehicle; the trajectory data and event data are collected by the V2X device.

[0031] This setting, based on the dual data collection of the vehicle side (collection vehicle) and the road side (V2X equipment), takes out the advantages of each as the basic data, which can make the collected data more real. Among them, it is easier and more accurate for the vehicle side to collect static maps, and it is easier for the road side to collect dynamic status information of traffic participants.

[0032] Step 2: convert the basic data and obtain the target data; the target data includes Opendrive data and Opensenario data.

[0033] Specifically, the map data in the basic data is converted into Opendrive data, and the trajectory data and event data are converted into Opensenario data; so as to facilitate direct import into the simulation software.

[0034] Step 3: Based on the target data, data fusion and preliminary screening are performed between roads; then, preliminary screening of abnormal scenes is performed to obtain preliminary screening data.

[0035] In this step, if Figure 2 As shown, the data fusion includes fusing the repeatedly collected data in the overlapping collection area according to the traffic participants.

[0036] Specifically, since the data collection ranges between road-side devices (V2X devices) are prone to overlap, when a traffic participant passes through section A, he may pass through the data collection ranges of two V2X devices at the same time, resulting in repeated collected data, and such data is fused. In addition, in the fusion process, since the collection time between the two V2X devices is not completely consistent, and the collected data may be lost, it is necessary to judge the collected data. Since only one traffic participant will appear at the same coordinate at the same time, and the traffic participant has a certain volume, and the range of movement of the traffic participant is limited in a continuous period of time. In addition, although the two collection devices are not completely consistent in time, the time error is at the millisecond level.

[0037] According to these characteristics, the range of action of traffic participants is limited in the millisecond level of error. Therefore, by adding upper and lower thresholds for location and time, it is determined whether the traffic participants are the same traffic participants, and then the data of the two road-end devices are fused into the data of the traffic participants with the same ID. The preliminary screening between roads includes discarding the data of this period of time (i.e., screening out this period of data) when a large amount of data is misplaced or the data of traffic participants cannot be connected and matched. Using the above method, the ID of the fused trajectory data of traffic participants does not change in the entire road section, and the data of different devices can be effectively connected.

[0038] like Figure 3 As shown, the preliminary screening of abnormal scenarios includes:

[0039] (1) Preliminary screening based on the relationship between traffic participants and the map: determine the location of the traffic participant. When the location is outside the map, if the number of data outside the map is less than N, only the data outside the map is screened out; if the number of data outside the map is greater than M, all the data is screened out.

[0040] (2) Preliminary screening based on traffic participants - judging whether the speed and posture of traffic participants are abnormal based on the posture of traffic participants, the time interval between sampling points, and the kinematic model of traffic participants; the posture includes the position of traffic participants and the orientation of traffic participants. Specifically, in this embodiment, when judging whether the speed of traffic participants is abnormal, if the change of its speed within the sampling time is greater than the maximum acceleration, it is judged that the speed of the traffic participant is abnormal; similarly, when judging whether the posture of traffic participants is abnormal, if the change of its posture within the sampling time does not conform to the kinematic model, if so, it is judged that the posture of traffic participants is abnormal. In addition, for data with abnormalities, the data of a single sampling point is directly deleted.

[0041] (3) Preliminary screening based on the relationship between traffic participants and other objects - determine whether there are two or more traffic participants at the same place at the same time (i.e., whether there are traffic participants with overlapping positions); if so, extract the data of the traffic participants and perform data analysis. The data analysis includes determining whether the traffic participants stop moving after a period of time after the overlapping positions. If so, it is determined to be a traffic accident. If not, further determine the reasonableness of the position and the completeness of the data of each traffic participant. Specifically, based on the changes in the speed and acceleration of each traffic participant, select the traffic participant with the most stable situation as the traffic participant with the highest reasonableness of position, retain its data, and screen out the data of the remaining traffic participants. If the changes in the speed and acceleration of each traffic participant are similar, select the traffic participant with the most complete data, retain its data, and screen out the data of the remaining traffic participants.

[0042] Step 4: Based on the initial screening data, perform trajectory evaluation and select scene data with high-quality trajectories, such as Figure 4 shown.

[0043] In this step, when performing trajectory evaluation, it includes: analyzing the trajectory integrity and trajectory availability of the trajectory, and combining the trajectory integrity and trajectory availability through trajectory quality influencing factors to obtain a quality evaluation of the trajectory; the trajectory integrity includes trajectory point density and trajectory point continuity; the trajectory availability includes trajectory smoothness.

[0044] In this embodiment, in the quality evaluation of the trajectory, an evaluation function F is used to represent the trajectory quality:

[0045] F(t) = α1 × C(t) + α2 × A(t);

[0046] Among them, α represents the trajectory quality influencing factor; C represents the function of trajectory integrity, A represents the function of trajectory availability, and t represents the evaluated trajectory.

[0047] Here, when both the trajectory integrity and trajectory availability reach their maximum values, the trajectory quality F is obtained. max As a benchmark, the trajectory quality F>0.8F in the selected trajectory max The trajectory is a high-quality trajectory.

[0048] Specifically, the trajectory point density is the ratio of the number of points collected to the total number of points that should be collected. The continuity of the trajectory points is determined based on the impact of the amount of lost data on the total data during the time when the intermediate data is missing. The greater the impact, the worse the continuity of the trajectory points. And the longer the time when the intermediate data is missing, the greater the impact on the total data.

[0049] The length of time when the intermediate data is missing refers to the length of sampling time when the data is continuously missing. For example, if the last sampling time point is 1:30 and the next sampling time point is 2:00, the length of continuous missing sampling time is 30 seconds. The amount of lost data refers to the sum of all the continuous missing sampling time lengths; these are inversely proportional to the trajectory continuity. The influencing factor Cont of the continuous missing sampling time length will increase with the increase of time. And Among them, Cont is the continuity function, β is the impact factor, and L is the time length function.

[0050] In addition, a smooth curve is obtained through curve fitting, and then the jitter of the trajectory is obtained by calculating the variance to determine the smoothness of the trajectory.

[0051] Step 5: Based on the initial screening data, a collision risk assessment is performed to obtain a collision risk rating, and scene data with no collision risk is screened out.

[0052] Specifically, when conducting a collision risk assessment, a collision risk rating is calculated based on the distance between traffic participants, the speed and direction of each traffic participant, and a collision risk judgment method based on TTC.

[0053] TTC mainly refers to the time required for a vehicle to reach the potential collision point at its current speed and trajectory; the smaller the TTC value, the higher the collision risk.

[0054] The calculation formula of TTC is as follows: TTC = D / V;

[0055] TTC is the time to collision, in seconds (s); D is the distance between the vehicle and the potential collision point (such as an obstacle, the vehicle in front, etc.), in meters; V is the current speed of the vehicle, in meters / s.

[0056] The TTC value is then converted into a more intuitive collision risk index. In this embodiment, a collision risk factor (CRF) based on TTC is defined as the collision risk index, and its calculation formula is as follows:

[0057] CRF = 1 / (TTC + ε);

[0058] Where: CRF is the collision risk factor. The larger the value, the higher the collision risk. ε is a very small positive number, which is used to ensure that the formula is still valid when TTC is close to 0.

[0059] Finally, low-level collision risk, medium-level collision risk, and high-level collision risk are distinguished according to the range of CRF values.

[0060] In this embodiment, if the CRF value is lower than a certain lower threshold (eg, 0.1), it is determined as a low-level collision risk.

[0061] If the CRF value is between two thresholds (eg, 0.1 to 0.3), it is determined to be a medium-level collision risk.

[0062] If the CRF value is higher than a certain higher threshold (eg, 0.3), it is determined to be a high-level collision risk.

[0063] In actual applications, the above thresholds can be set according to specific application scenarios and safety requirements, and can be adjusted and optimized during actual operations.

[0064] Step 6: Based on the preliminary screening data and the processing results of steps 4 to 5, scene elements are identified; the scene elements include map information, traffic participant behavior, traffic participant type, trajectory quality, collision level and traffic flow, such as Figure 5 shown.

[0065] In this step, the map information includes scene elements such as crossroads, T-shaped intersections, roundabouts, uphills, downhills, ramps, tunnels and highways; based on the behavior of traffic participants, it includes active behavior of the main vehicle and passive behavior of the main vehicle. Specifically, the active behavior of the main vehicle includes following the main vehicle, overtaking the main vehicle, continuous lane change of the main vehicle, merging into the main road, merging with the ramp vehicle, avoiding pedestrians, bypassing the accident scene, bypassing the construction section, etc. The passive behavior of the main vehicle includes the adjacent vehicle overtaking the main vehicle, the adjacent vehicle driving on the line, the adjacent vehicle turning around on the road, and the adjacent other traffic participants crossing the lane. In addition, when identifying the active behavior and passive behavior of the main vehicle, classification can be performed directly according to the time-series trajectory of the vehicle in the initial screening data and the map information.

[0066] Step 7: Based on the initial screening data and the processing results of step 6, the scenes are classified and stored in the scene library.

[0067] like Figure 6 As shown, in this step, the scene classification includes a first-level classification; the first-level classification includes a large-type classification according to preset classification rules; the preset classification rules include: whether there is a lane change operation, whether there is a ramp, whether it passes through an intersection, whether there are adjacent vehicles, the orientation between traffic participants, and whether there are pedestrians involved.

[0068] Through the above-mentioned first-level classification method, the data can be quickly divided into several small categories by utilizing the preset classification rules and the processing results obtained in the previous steps. The scene classification efficiency is high and it helps to improve the subsequent second-level classification effect.

[0069] Furthermore, during classification, the preset classification rules also include classification based on some scene elements, including classification based on traffic participant types, classification based on trajectory quality, classification based on collision levels, and classification based on traffic flow.

[0070] Traffic participants refer to the classification of major traffic participants, which can be subdivided into various types of vehicles. Trajectory quality classification is performed based on the trajectory evaluation results in step 4. Collision level classification is performed based on the collision risk rating results in step 5. Traffic flow is classified based on the number of vehicles passing through in a period of time. The classified scenes are stored in the scene library.

[0071] With this setting, the scheme can automatically classify based on the results evaluated in the above steps without redundant calculations. It only needs to define the classification value range, and the amount of data analysis is small, which is more suitable for large-scale traffic flow and road network simulation scenarios.

[0072] The scene classification also includes secondary classification; the secondary classification includes: based on the primary classification results, using a GRU model for secondary classification. In this way, compared with directly introducing the GRU model, this method chooses to apply the GRU model on the basis of the primary classification, and the data features input to the GRU model are more obvious, which can effectively save the model training cost and achieve a higher model accuracy.

[0073] Specifically, the input of the GRU unit is the position information, speed information and direction information of the main vehicle and the adjacent traffic participants at each moment, and the output is the final determined scene type. The GRU unit can complete the classification based on the behavior of traffic participants.

[0074] In actual applications, during autonomous driving simulation testing, the corresponding scene files are found in the scene library by classification according to the test requirements, and imported into the simulation software to achieve simulation reproduction of traffic flow and road network.

[0075] This embodiment also provides a large-scale traffic flow and road network simulation and reproduction system in a vehicle-road cooperative scenario, which is used to execute the large-scale traffic flow and road network simulation and reproduction method in a vehicle-road cooperative scenario as described above; it includes a collection module, a conversion module, a data processing module and a classification module;

[0076] The acquisition module is used to acquire basic data; the conversion module is used to convert basic data and obtain target data; the data processing module is used to perform data fusion and preliminary screening between roads based on the target data; then, perform preliminary screening of abnormal scenes; obtain preliminary screening data; and, based on the preliminary screening data, perform trajectory evaluation and screen out scene data with high-quality trajectories; and, based on the preliminary screening data, perform collision risk evaluation, obtain collision risk rating, and screen out scene data with no collision risk; and, perform scene element identification; the classification module is used to classify scenes based on the processing results of the data processing module, and store the classified scenes in a scene library.

[0077] This embodiment provides a method and system for simulating and reproducing large-scale traffic flow and road network in a vehicle-road collaboration scenario, which can automatically complete scenario-based data processing and classification using real data, and can provide real scenario data for traffic flow and road network simulation and reproduction.

[0078] In particular, this solution provides an efficient scene classification method that can efficiently classify the massive data generated in large-scale traffic flow and road network scenarios into different types of scene data required for simulation. Compared with the conventional solution based on a single large model for data classification and processing, the conventional solution performs well in small-scale scene simulations, but when faced with large-scale scene simulations and the data source is multi-type and complex real-world perception data, its data processing efficiency and accuracy are greatly reduced. In this solution, a classification method that combines rules and model classification is proposed, which has higher accuracy and processing efficiency.

[0079] Specifically, this scheme sets strict preliminary screening steps in the early stage, and preliminarily screens the scheme from different dimensions (preliminary screening based on the relationship between traffic participants and maps, preliminary screening based on traffic participants, and preliminary screening based on the relationship between traffic participants and other objects), and only retains the most reasonable and accurate data. In addition, after the preliminary screening, the data is evaluated in different dimensions, and the data can be screened again based on the evaluation results. After the above screening, the data without practical value in the basic data can be screened out, and the amount of data can be effectively streamlined, the burden of data processing can be reduced, and a lot of time can be saved for subsequent data classification and storage. Secondly, when classifying scenes, this scheme first uses rule-based classification calculation for primary classification, and then uses the method of decentralized GRU model to classify the scenes in detail according to the results of the primary classification, which can save a lot of computing time and improve the accuracy of model classification. In addition, the different types of scenes obtained under the classification method set by this scheme can cover most of the simulation test demand scenes through combination, and can provide sufficient and reliable materials for large-scale traffic flow and road network simulation reproduction.

[0080] 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 large-scale traffic flow and road network simulation and reproduction method in a vehicle-road collaboration scenario, characterized in that: The following steps are involved: Step 1, collecting basic data; the basic data includes map data, trajectory data and event data; Step 2, converting the basic data and obtaining the target data; the target data includes Opendrive data and Opensenario data; Step 3: Based on the target data, perform data fusion and preliminary screening between roads; Then, conduct a preliminary screening of abnormal scenarios; Get the initial screening data; The preliminary screening of abnormal scenes includes: preliminary screening based on the relationship between traffic participants and maps, preliminary screening based on traffic participants, and preliminary screening based on the relationship between traffic participants and other objects; The preliminary screening based on the relationship between the traffic participant and the map includes determining the location of the traffic participant. When the location is outside the map, if the amount of data outside the map is less than N, only the data outside the map is screened out; if the amount of data outside the map is greater than M, all the data is screened out. The preliminary screening based on traffic participants includes judging whether the speed of traffic participants and the posture of traffic participants are abnormal based on the posture of traffic participants, the time interval between sampling points, and the kinematic model of traffic participants; for abnormal data, directly deleting the data of a single sampling point; The preliminary screening based on the relationship between traffic participants and other objects includes determining whether there are two or more traffic participants at the same place at the same time, and if so, extracting the data of the traffic participants, performing data analysis, and selecting the traffic participant data to be retained according to the data analysis results; Step 4: Based on the initial screening data, perform trajectory evaluation and select scene data with high-quality trajectories; Step 5: Based on the initial screening data, a collision risk assessment is performed to obtain a collision risk rating, and scene data without collision risk is screened out; Step 6, based on the preliminary screening data and the processing results of steps 4 to 5, scene elements are identified; the scene elements include map information, traffic participant behavior, traffic participant type, trajectory quality, collision level and traffic flow; Step 7: Classify the scenes based on the initial screening data and the processing results of step 6, and store the classified scenes in the scene library; The scene classification includes a primary classification; the primary classification includes a large-scale classification according to a preset classification rule; the preset classification rules include: whether there is a lane change operation, whether there is a ramp, whether it passes through an intersection, whether there are adjacent vehicles, the orientation of traffic participants, and whether there are pedestrians involved; The scene classification also includes secondary classification; the secondary classification includes: based on the primary classification results, using a GRU model for secondary classification.

2. The method for simulating and reproducing large-scale traffic flow and road network in a vehicle-road cooperative scenario according to claim 1 is characterized in that: In step 1, map data is collected by a collection vehicle; The trajectory data and event data are collected by V2X devices.

3. The method for simulating and reproducing large-scale traffic flow and road network in a vehicle-road cooperative scenario according to claim 2 is characterized in that: In step 3, the data fusion includes fusing the repeatedly collected data in the overlapping collection area according to the traffic participants.

4. The method for simulating and reproducing large-scale traffic flow and road network in a vehicle-road cooperative scenario according to claim 1 is characterized in that: In step 4, when performing trajectory evaluation, it includes: analyzing the trajectory integrity and trajectory availability of the trajectory, and combining the trajectory integrity and trajectory availability through trajectory quality influencing factors to obtain a quality evaluation of the trajectory; the trajectory integrity includes trajectory point density and trajectory point continuity; the trajectory availability includes trajectory smoothness.

5. The method for simulating and reproducing large-scale traffic flow and road network in a vehicle-road cooperative scenario according to claim 1 is characterized in that: In step 6, the map information includes scene elements of intersections, T-junctions, roundabouts, uphills, downhills, ramps, tunnels and highways; and based on the behavior of traffic participants, it includes active behavior of the main vehicle and passive behavior of the main vehicle.

6. The method for simulating and reproducing large-scale traffic flow and road network in a vehicle-road cooperative scenario according to claim 1 is characterized in that: The input of the GRU model is the position information, speed information and direction information of the main vehicle and the adjacent traffic participants at each moment, and the output is the final determined scene type.

7. A large-scale traffic flow and road network simulation and reproduction system under vehicle-road collaboration scenarios, characterized by: Used to execute the large-scale traffic flow and road network simulation and reproduction method in the vehicle-road cooperative scenario as described in any one of claims 1 to 6; including a collection module, a conversion module, a data processing module and a classification module; The acquisition module is used to acquire basic data; the conversion module is used to convert basic data and obtain target data; the data processing module is used to perform data fusion and preliminary screening between roads based on the target data; then, preliminary screening of abnormal scenes is performed; Get the initial screening data; And, based on the initial screening data, trajectory evaluation is performed, and scene data with high-quality trajectories are screened out; And, based on the initial screening data, a collision risk assessment is performed to obtain a collision risk rating, and scene data without collision risk is screened out; and, performing scene element recognition; The classification module is used to classify scenes based on the processing results of the data processing module, and store the classified scenes in a scene library.

Images