A method and device for simulating automatic driving based on real traffic data set

Abstract

The present invention provides an automatic driving simulation method based on a real traffic data set, which can be applied to the field of intelligent driving technology. The method includes: performing data parsing and formatting on the real traffic data set, initializing entities in the simulation environment based on the obtained dictionary format traffic data set, and managing the initialized multi-category entity instances; using the multi-category entity instances to calculate the geometric features of the entities in the simulation environment and the geometric relationships between the entities, and using a predefined real-time detection and tracking algorithm to process the multi-category entity instances to obtain a global planning scheme in the simulation environment; based on the geometric relationship and the global planning scheme, using the automatic driving algorithm to be tested to obtain the expected motion strategy of the controlled vehicle entity in the simulation environment, and solving the vehicle kinematic model based on the expected motion strategy to obtain the future motion state of the controlled vehicle entity and thus realize the automatic driving simulation of the controlled vehicle entity.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and more specifically to an automatic driving simulation method and device based on a real traffic data set. Background Art

[0002] At present, the development and testing of high-level autonomous driving algorithms cannot be separated from the support of autonomous driving virtual simulators. A fully functional virtual simulation environment can provide simulations of various terrains, streets, traffic participants, vehicle sensor data and even weather factors similar to those in the real environment. In addition, the function of supporting ultra-real-time accelerated simulation greatly accelerates the iteration cycle of the algorithm. These advantages are not available in real scenes or require a lot of manpower, material resources and time. Therefore, a good virtual simulation environment should simulate the real world as realistically as possible, and the simulation scenes should be as rich as possible to cover most of the real world. The intelligent driving decision-making control algorithm trained based on this environment can have better robustness and safety in the real road environment. At present, about 90% of autonomous driving algorithm tests are completed through simulation platforms, about 9% are completed through closed test sites, and 1% are completed through actual road tests. With the improvement of simulation technology, it is expected that 99.9% of the tests will be completed by simulators to achieve more efficient and economical R&D and testing.

[0003] At present, simulation testing mainly implements algorithm testing by building a simulation scene library. Specifically for urban road scenes, there are many long-tail situations due to the complex road structure, easy obstruction of environmental perception, and complex behavior of surrounding vehicles. Therefore, the fidelity of virtual simulation scenes has always been one of the research focuses of major scientific research institutions and autonomous driving companies. In order to study high-level autonomous driving problems in urban road scenes, on the one hand, the support of relevant data sets is needed, and for reinforcement learning-based decision-making and control algorithms, their training and test verification also require a simulator that can interact with them.

[0004] However, for offline training of reinforcement learning for autonomous driving, especially for tasks such as decision-making in complex urban road scenarios, there is a lack of a complete technical solution from real data sets to virtual simulation scenarios. At the same time, many existing autonomous driving simulation technology solutions have technical problems such as incomplete traffic data sets, single simulation modes, inaccurate simulation results, and lack of generalization of simulation technology solutions. Summary of the invention

[0005] In view of the above problems, the present invention provides an automatic driving simulation method and device based on a real traffic data set.

[0006] According to a first aspect of the present invention, there is provided an automatic driving simulation method based on a real traffic data set, the method comprising:

[0007] Parse and format the real traffic data set to obtain a dictionary-format traffic data set, initialize entities in the simulation environment based on the dictionary-format traffic data set, and manage the multi-category entity instances obtained by initialization;

[0008] The geometric features of entities in the simulation environment and the geometric relationships between entities are calculated using multi-category entity instances, and the multi-category entity instances are processed using a predefined real-time detection and tracking algorithm to obtain a global planning solution in the simulation environment;

[0009] Based on the geometric relationship and the global planning scheme, the expected motion strategy of the controlled vehicle entity in the simulation environment is obtained by using the autonomous driving algorithm to be tested, and the vehicle kinematic model is solved based on the expected motion strategy to obtain the future motion state of the controlled vehicle entity;

[0010] Based on the user's selection information, the controlled vehicle entity's own state information and related information are calculated and output in real time, and the simulation environment is rendered in real time based on the user's personalized demand information, providing the user with a visual customized simulation environment;

[0011] Based on user setting information and the configuration information required by the autonomous driving algorithm to be tested, the simulation environment is controlled by data set selection, visualization, and scene simulation to meet the needs of reinforcement learning tasks, and information interaction is achieved through a request-question-answer mode.

[0012] According to an embodiment of the present invention, the above-mentioned data parsing and format processing of the real traffic data set to obtain the dictionary format traffic data set includes:

[0013] Analyze the real traffic data to obtain the real traffic data files taken by drones and the real traffic data files taken by traffic cameras;

[0014] The real traffic data files taken by drones and traffic cameras are formatted to obtain a dictionary format traffic data set.

[0015] The dictionary format traffic data set includes map files, vehicle trajectory files, vehicle surrounding environment data, and traffic facility data;

[0016] Among them, the vehicle trajectory file includes vehicle number, time frame number, timestamp, vehicle type, x-coordinate in the global coordinate system, y-coordinate in the global coordinate system, x-direction speed in the global coordinate system, y-direction speed in the global coordinate system, vehicle orientation in the global coordinate system, vehicle length, and vehicle width.

[0017] According to an embodiment of the present invention, the above-mentioned initialization of entities in the simulation environment based on the dictionary format traffic data set and management of the multi-category entity instances obtained by initialization include:

[0018] Read the map file and traffic facility data in the dictionary format traffic data set, and initialize the map entity by setting the map pointer;

[0019] Read the vehicle trajectory file in the dictionary format traffic data set to initialize the controlled vehicle entity;

[0020] Read the vehicle surrounding environment data in the dictionary format traffic data set to initialize the surrounding environment vehicle entity in the surrounding environment of the controlled vehicle entity;

[0021] Manage multiple categories of entity instances including map entities, controlled vehicle entities, and surrounding vehicle entities.

[0022] According to an embodiment of the present invention, the above-mentioned use of multiple categories of entity instances to calculate the geometric features of entities and the geometric relationships between entities in the simulation environment includes:

[0023] Reading a map entity in a multi-category entity instance, and calculating a plane geometric feature on the map in a simulation environment, wherein the plane geometric feature includes a lane curvature and a centerline position;

[0024] Read map entities, controlled vehicle entities and surrounding vehicle entities in the multi-category entity instance, and calculate the plane geometric relationship between the entities, where the plane geometric relationship includes the distance between the controlled vehicle entity and the surrounding vehicle entities, the distance between the controlled vehicle entity and the lane boundary, and the distance between the controlled vehicle entity and the lane centerline.

[0025] According to an embodiment of the present invention, the above-mentioned use of a predefined real-time detection and tracking algorithm to process multiple categories of entity instances to obtain a global planning solution in a simulation environment includes:

[0026] The map entities in the multi-category entity instances are segmented, and based on the segmentation results, the predefined real-time detection and tracking algorithm is used to perform path planning, road segment relationship search, road conflict judgment, search for vehicle entities in the multi-category entity instances, and target matching road segment search for vehicle entities in the simulation environment.

[0027] According to an embodiment of the present invention, the above-mentioned method based on geometric relationship and global planning scheme uses the autonomous driving algorithm to be tested to obtain the expected motion strategy of the controlled vehicle entity in the simulation environment, and solves the vehicle kinematic model based on the expected motion strategy to obtain the future motion state of the controlled vehicle entity, including:

[0028] Construct a vehicle kinematics model, and construct a vehicle inverse kinematics model based on the vehicle kinematics model, and obtain the expected acceleration and expected yaw angle by analyzing the expected motion strategy;

[0029] The vehicle kinematics model is solved based on the expected acceleration and the expected yaw angle to obtain the future motion state of the controlled vehicle entity;

[0030] The actual motion state information of the controlled vehicle is obtained by parsing the dictionary format traffic data set, and the vehicle inverse kinematics model is solved based on the actual motion state information to obtain the predicted motion strategy of the controlled vehicle entity;

[0031] The future motion state is used to control the controlled vehicle entity in real time in a simulation environment, and the autonomous driving algorithm to be tested is evaluated based on the real-time control results.

[0032] According to an embodiment of the present invention, the above-mentioned real-time calculation and output of the self-state information and associated information of the controlled vehicle entity based on the user selection information includes:

[0033] Selecting a target vehicle as a controlled vehicle entity from multiple categories of entity instances, and determining the self-state information and associated information of the controlled vehicle entity to be output based on user selection information;

[0034] Based on the state information and associated information of the controlled vehicle entity, the local bird's-eye view of the controlled vehicle entity, real-time motion state information and real-time motion strategy, local map information and information of the surrounding vehicle entities of the controlled vehicle entity are output.

[0035] According to an embodiment of the present invention, the above-mentioned real-time image rendering of the simulation environment based on the user's personalized demand information to provide the user with a visual customized simulation environment includes:

[0036] Determine the user visualization results by analyzing the user's personalized demand information, where the user visualization results include visualization of the planned path, visualization of the vehicle highlighting in the intersection range environment, visualization of the trajectory file comparison display corresponding to the data set of the controlled vehicle entity, and visualization of the controlled vehicle entity highlighting;

[0037] Based on the user's visualization results, the simulation environment is rendered in real time to obtain the visualized customized simulation environment required by the user.

[0038] According to an embodiment of the present invention, based on user setting information and configuration information required by the autonomous driving algorithm to be tested, the simulation environment is controlled to select a data set, to be visualized, and to be used for scene simulation to meet the needs of the reinforcement learning task, and information interaction is achieved through a request-question-answer mode, including:

[0039] Based on the user settings and the configuration information required by the autonomous driving algorithm to be tested, define the simulation environment class, simulation logic scheme, communication mode and communication protocol, and complete the scene initialization of the simulation environment;

[0040] Based on the scene initialization results of the simulation environment, the map entities, controlled vehicle entities, and surrounding vehicle entities required by the autonomous driving algorithm to be tested are selected from multiple categories of entity instances;

[0041] Visualize and output relevant information of the controlled vehicle entity in the simulation environment at each moment and information associated with the autonomous driving algorithm to be tested;

[0042] Dynamically start or stop scenario simulation and interact with users through the request-question-answer mode.

[0043] According to a second aspect of the present invention, there is provided an automatic driving simulation device based on a real traffic data set, the device comprising:

[0044] The data reading module is used to parse and format the real traffic data set to obtain the dictionary format traffic data set;

[0045] The scene entity module is used to initialize entities in the simulation environment based on the dictionary format traffic data set and manage the multi-category entity instances obtained by initialization;

[0046] A general geometric relationship calculation module is used to calculate the geometric features of entities in the simulation environment and the geometric relationships between entities using multi-category entity instances;

[0047] The map relationship calculation module is used to process multi-category entity instances using predefined real-time detection and tracking algorithms to obtain a global planning solution in the simulation environment;

[0048] The vehicle motion module is used to obtain the expected motion strategy of the controlled vehicle entity in the simulation environment based on the geometric relationship and the global planning scheme, using the autonomous driving algorithm to be tested, and solve the vehicle kinematic model based on the expected motion strategy to obtain the future motion state of the controlled vehicle entity;

[0049] An output information calculation module is used to calculate and output the self-state information and related information of the controlled vehicle entity in real time based on the user selection information;

[0050] The visualization module is used to render the simulation environment in real time based on the user's personalized demand information, and provide the user with a visualized customized simulation environment;

[0051] The simulation environment logic and communication interface module is used to control the data set selection, visualization and scene simulation of the simulation environment based on user setting information and the configuration information required by the autonomous driving algorithm to be tested to meet the needs of reinforcement learning tasks, and realize information interaction through the request-question-answer mode.

[0052] The above-mentioned autonomous driving simulation method based on real traffic data set provided by the present invention has constructed a complete technical solution from real traffic data set to virtual simulation scene; it can provide global and real traffic scene simulation for the autonomous driving algorithm to be tested, and improve the training and testing efficiency of the autonomous driving algorithm to be tested; and it provides a customized simulation environment through visualization and customization functions, providing users with targeted simulation model training scenes. In addition, due to the concise communication mode of the above-mentioned method provided by the present invention, lightweight end-to-end deployment is achieved, which expands the application scenarios of the autonomous driving simulation technical solution provided by the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The above contents and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0054] Figure 1 is a flow chart of an automatic driving simulation method based on a real traffic data set according to an embodiment of the present invention;

[0055] Figure 2 is an architecture diagram of an automatic driving simulation device based on a real traffic data set according to an embodiment of the present invention;

[0056] Figure 3 is a motion state and motion strategy analysis diagram of a controlled vehicle entity according to an embodiment of the present invention;

[0057] Figure 4 A block diagram of an electronic device suitable for implementing an autonomous driving simulation method based on a real traffic data set according to an embodiment of the present invention is schematically shown. DETAILED DESCRIPTION

[0058] Below, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of embodiments of the present invention. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of concepts of the present invention.

[0059] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0060] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0061] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0062] Due to the limitations of high-precision maps of urban roads and the openness of traffic data, as well as the difficulty and cost of collection, the current open-source traffic data in the autonomous driving industry is largely concentrated on data collected from the perspective of the main vehicle. For example, some data sets are collected on highways and only contain a small number of strong interactive scenes. Some data sets are released earlier and the collection technology (the main collection method is to set up cameras at high altitudes to record, while some newer data sets have adopted drone aerial photography to better eliminate the errors caused by tilted perspectives and camera edge distortion) is not yet mature, and there is a certain proportion of serious data errors. Some data sets contain a large amount of road information and vehicle trajectory information, but the map scenes are relatively simple, mainly intersections or straight roads. It can be seen that the data sets commonly used in existing autonomous driving simulation technology solutions have problems such as single data, old data, and incomplete data.

[0063] Some existing autonomous driving simulation technology solutions are derived from racing simulators. They are not suitable for real urban scene simulation needs due to their simple track scene maps and lack of sensor simulation support. Although some autonomous driving simulation technology solutions are rooted in actual driving scenarios such as urban scenes, they can only provide hardware-in-the-loop dynamic simulation and visualization, that is, they do not support algorithm-in-the-loop dynamic simulation. Therefore, functionally speaking, they are not suitable for building virtual simulation scenes based on real traffic data and for the development of reinforcement learning algorithms, but their complete functional modules can provide real-car online testing functions for algorithms.

[0064] From the above analysis, it can be seen that there are many troubles in realizing the data set to build a virtual simulation scene. From the perspective of offline training of reinforcement learning algorithms, there are also requirements for simulators. First of all, the transferability of reinforcement learning training results depends on the degree of simulation of the training environment provided by the simulator. If there is a large gap between the simulator and the actual test environment, the transferability is poor. This fidelity includes the underlying motion and control model of the vehicle on the one hand, and the surrounding environment such as road structure and surrounding vehicle behavior patterns on the other hand. For the surrounding vehicle behavior patterns, the current mainstream simulators all adopt the surrounding vehicle behavior setting method based on data playback. Through real data playback, the difference with the real world can be minimized. However, only data playback without feedback on the behavior of the autonomous driving main vehicle may cause the distribution offset of the interactive data. This is far from enough for some scientific research directions that study the interactive behavior of surrounding vehicles. The surrounding vehicles must have a certain degree of intelligent interaction capabilities, which requires the simulator to also support the individual control or joint control mode of the surrounding vehicles. In addition, algorithm training for reinforcement learning usually requires the collection of a large amount of experience playback data. In order to improve the training speed, the simulator needs to be able to meet multi-process parallel collection, that is, it has a port communication mechanism. Finally, the simulator itself should be as lightweight as possible to support fast simulation iteration training.

[0065] For offline training of autonomous driving reinforcement learning, especially for decision-making tasks in complex urban road scenarios, there is currently a lack of a complete set of construction technologies from data sets to virtual simulation scenarios. Therefore, it is necessary to develop a set of technologies from real traffic data to virtual simulators based on actual needs.

[0066] In order to solve the above-mentioned problems, the present invention has developed a set of technologies for constructing virtual simulation scenes from real traffic data sets based on real traffic data sets, and based on this, an autonomous driving simulator that meets the requirements of reinforcement learning training functions is formed.

[0067] Figure 1 4 is a flowchart of an automatic driving simulation method based on a real traffic data set according to an embodiment of the present invention.

[0068] like Figure 1 As shown, the above-mentioned automatic driving simulation method based on the real traffic data set includes operations S110 to S150.

[0069] In operation S110, the real traffic data set is parsed and formatted to obtain a dictionary-format traffic data set, and entities in the simulation environment are initialized based on the dictionary-format traffic data set, and the initialized multi-category entity instances are managed.

[0070] The above-mentioned real traffic dataset is optionally an Interaction dataset. It should be noted that the dataset used in the present invention is a public, open source dataset, which has been authorized by the relevant owner, and the use of the dataset in the present invention is based on scientific research purposes, not for commercial applications.

[0071] The above-mentioned Interaction dataset includes various types of real traffic data, such as map information, vehicle information, traffic facility information, and vehicle surrounding environment information (such as surrounding vehicle information, surrounding pedestrian information, etc.).

[0072] The above real traffic data set is saved in dictionary form according to the id number.

[0073] The above multi-category entity examples include map entities, controlled vehicle entities (or target vehicle entities, i.e., vehicles to be simulated by the autonomous driving algorithm), other vehicle entities in the surrounding environment of the controlled vehicle, etc. Those skilled in the art can materialize other data types in the data set according to actual simulation requirements, such as traffic facility entities, pedestrian entities, etc.

[0074] In operation S120, geometric features of entities and geometric relationships between entities in the simulation environment are calculated using multi-category entity instances, and the multi-category entity instances are processed using a predefined real-time detection and tracking algorithm to obtain a global planning solution in the simulation environment.

[0075] The above-mentioned geometric features include curve length calculation; the above-mentioned geometric relationships are generally plane geometric relationships, such as the intersection judgment of two rectangles, the shortest distance calculation of two rectangles, the intersection judgment of a straight line and a rectangle, the shortest distance calculation of a straight line and a rectangle, the judgment of whether a point is inside or outside a polygon, the calculation of the vector angle, the calculation of the overlapping area of ​​two polygons or other specific geometric relationships, and based on the above-mentioned geometric relationships, more complex traffic relationships between entities in the simulation environment can be calculated, such as the calculation of the vehicle entry and exit area judgment, the position relationship judgment of two vehicles, the collision judgment of two vehicles, the vehicle orientation calculation, and the angle calculation between the vehicle and the centerline of the road.

[0076] It should be noted that the above-mentioned geometric features or geometric relationships are merely exemplary. Those skilled in the art can calculate other geometric features or geometric relationships according to actual simulation requirements, or calculate geometric features and geometric relationships on a solid or curved surface based on the technical inspiration provided by the present invention.

[0077] In operation S130, based on the geometric relationship and the global planning scheme, the desired motion strategy of the controlled vehicle entity in the simulation environment is obtained using the autonomous driving algorithm to be tested, and the vehicle kinematic model is solved based on the desired motion strategy to obtain the future motion state of the controlled vehicle entity.

[0078] Through basic vehicle kinematic model calculations, such as supporting particle models, bicycle models, and custom models, as well as data set recording trajectory inverse kinematics solution (i.e., inversely solving the control action based on the vehicle state information at two moments before and after), the real-time motion state of the controlled vehicle entity is obtained, thereby realizing the verification of the autonomous driving algorithm to be tested.

[0079] The future motion state refers to the motion state of the controlled vehicle entity at the next moment in the simulation environment, which is the motion state predicted based on the autonomous driving algorithm to be tested and the vehicle kinematic model.

[0080] In operation S140, the self-state information and associated information of the controlled vehicle entity are calculated and output in real time based on the user selection information, and the simulation environment is rendered in real time based on the user's personalized demand information to provide the user with a visual customized simulation environment.

[0081] In operation S150, based on user setting information and configuration information required by the autonomous driving algorithm to be tested, data set selection control, visualization control, and scene simulation control are performed on the simulation environment to meet the needs of the reinforcement learning task, and information interaction is achieved through a request-question-answer mode.

[0082] The above-mentioned autonomous driving simulation method based on real traffic data set provided by the present invention has constructed a complete technical solution from real traffic data set to virtual simulation scene; it can provide global and real traffic scene simulation for the autonomous driving algorithm to be tested, and improve the training and testing efficiency of the autonomous driving algorithm to be tested; and it provides a customized simulation environment through visualization and customization functions, providing users with targeted simulation model training scenes. In addition, due to the concise communication mode of the above-mentioned method provided by the present invention, lightweight end-to-end deployment is achieved, which expands the application scenarios of the autonomous driving simulation technical solution provided by the present invention.

[0083] According to an embodiment of the present invention, the above-mentioned data parsing and formatting of the real traffic data set to obtain the dictionary format traffic data set includes: parsing the real traffic data to obtain the real traffic data file taken by the drone and the real traffic data file taken by the traffic camera; formatting the real traffic data file taken by the drone and the real traffic data file taken by the traffic camera to obtain the dictionary format traffic data set; wherein the dictionary format traffic data set includes a map file, a vehicle trajectory file, vehicle surrounding environment data and traffic facility data; wherein the vehicle trajectory file includes a vehicle number, a time frame number, a timestamp, a vehicle type, an x ​​coordinate in the global coordinate system, a y coordinate in the global coordinate system, an x-direction speed in the global coordinate system, a y-direction speed in the global coordinate system, a vehicle orientation in the global coordinate system, a vehicle length and a vehicle width.

[0084] According to an embodiment of the present invention, the above-mentioned initialization of entities in the simulation environment based on the dictionary format traffic data set and management of the multi-category entity instances obtained by initialization include: reading the map file and traffic facility data in the dictionary format traffic data set, and initializing the map entity by setting the map pointer; reading the vehicle trajectory file in the dictionary format traffic data set to initialize the controlled vehicle entity; reading the vehicle surrounding environment data in the dictionary format traffic data set to initialize the surrounding environment vehicle entity in the surrounding environment of the controlled vehicle entity; managing the multi-category entity instances including the map entity, the controlled vehicle entity and the surrounding environment vehicle entity.

[0085] According to an embodiment of the present invention, the above-mentioned use of multi-category entity instances to calculate the geometric features of entities in a simulation environment and the geometric relationships between entities includes: reading map entities in the multi-category entity instances, calculating plane geometric features on the map in the simulation environment, wherein the plane geometric features include lane curvature and centerline position; reading map entities, controlled vehicle entities and surrounding vehicle entities in the multi-category entity instances, and calculating plane geometric relationships between entities, wherein the plane geometric relationships include the distance between the controlled vehicle entity and the surrounding vehicle entities, the distance between the controlled vehicle entity and the lane boundary, and the distance between the controlled vehicle entity and the centerline of the lane.

[0086] The above-mentioned plane geometric features and plane geometric relationships are merely exemplary. Those skilled in the art may set other plane geometric features and plane geometric relationships according to actual needs.

[0087] According to an embodiment of the present invention, the above-mentioned use of a predefined real-time detection and tracking algorithm to process multi-category entity instances to obtain a global planning solution in a simulation environment includes: segmenting map entities in the multi-category entity instances, and based on the segmentation results, using a predefined real-time detection and tracking algorithm to perform path planning, road segment relationship search, road conflict judgment, search for vehicle entities in the multi-category entity instances, and target matching road segment search for vehicle entities in the simulation environment.

[0088] According to an embodiment of the present invention, the above-mentioned method based on geometric relationships and global planning schemes uses the autonomous driving algorithm to be tested to obtain the expected motion strategy of the controlled vehicle entity in the simulation environment, and solves the vehicle kinematic model based on the expected motion strategy to obtain the future motion state of the controlled vehicle entity, including: constructing a vehicle kinematic model, and constructing a vehicle inverse kinematic model based on the vehicle kinematic model, and obtaining the expected acceleration and expected yaw angle by analyzing the expected motion strategy; solving the vehicle kinematic model based on the expected acceleration and the expected yaw angle to obtain the future motion state of the controlled vehicle entity; obtaining the actual motion state information of the controlled vehicle by parsing a dictionary-format traffic data set, and solving the vehicle inverse kinematic model based on the actual motion state information to obtain the predicted motion strategy of the controlled vehicle entity; using the future motion state to control the controlled vehicle entity in real time in the simulation environment, and evaluating the autonomous driving algorithm to be tested based on the real-time control results.

[0089] According to an embodiment of the present invention, the above-mentioned real-time calculation and output of the self-state information and related information of the controlled vehicle entity based on the user selection information includes: selecting a target vehicle as the controlled vehicle entity from multiple categories of entity instances, and determining the self-state information and related information of the controlled vehicle entity to be output based on the user selection information; based on the self-state information and related information of the controlled vehicle entity, outputting a local bird's-eye view of the controlled vehicle entity, real-time motion state information and real-time motion strategy, local map information and information of vehicle entities in the surrounding environment of the controlled vehicle entity.

[0090] According to an embodiment of the present invention, the above-mentioned real-time image rendering of the simulation environment based on the user's personalized demand information to provide the user with a visual customized simulation environment includes: determining the user visualization results by parsing the user's personalized demand information, wherein the user visualization results include visualization of the planned path, visualization of vehicle highlighting in the intersection range environment, visualization of trajectory file comparison display in the corresponding data set of the controlled vehicle entity, and visualization of highlighted display of the controlled vehicle entity; based on the user visualization results, real-time image rendering of the simulation environment is performed to obtain the visual customized simulation environment required by the user.

[0091] According to an embodiment of the present invention, the above-mentioned data set selection control, visualization control and scene simulation control of the simulation environment based on user setting information and configuration information required by the autonomous driving algorithm to be tested to meet the needs of the reinforcement learning task, and information interaction is realized through a request-question-answer mode, including: based on user setting information and configuration information required by the autonomous driving algorithm to be tested, the simulation environment class, simulation logic scheme, communication mode and communication protocol are defined, and the scene initialization of the simulation environment is completed; based on the scene initialization result of the simulation environment, the map entity, the controlled vehicle entity and the surrounding vehicle entity required by the autonomous driving algorithm to be tested are selected from multiple categories of entity instances; the relevant information of the controlled vehicle entity in the simulation environment at each moment and the information associated with the autonomous driving algorithm to be tested are visualized and output; the scene simulation is dynamically started or stopped, and information interaction is realized with the user through a request-question-answer mode.

[0092] Figure 2 4 is an architectural diagram of an automatic driving simulation device based on a real traffic data set according to an embodiment of the present invention.

[0093] like Figure 2 As shown, the above-mentioned autonomous driving simulation device based on real traffic data sets includes a data reading module, a scene entity module, a general geometric relationship calculation module, a map relationship calculation module, a vehicle motion module, an output information calculation module, a visualization module, and a simulation environment logic and communication interface module.

[0094] The data reading module is used to parse and format the real traffic data set to obtain a dictionary format traffic data set.

[0095] The scene entity module is used to initialize entities in the simulation environment based on the dictionary format traffic data set and manage the multi-category entity instances obtained by initialization.

[0096] The general geometric relationship calculation module is used to calculate the geometric features of entities and the geometric relationships between entities in the simulation environment using multi-category entity instances.

[0097] The map relationship calculation module is used to process multi-category entity instances using a predefined real-time detection and tracking algorithm to obtain a global planning solution in the simulation environment.

[0098] The vehicle motion module is used to obtain the expected motion strategy of the controlled vehicle entity in the simulation environment based on geometric relationships and global planning schemes using the autonomous driving algorithm to be tested, and to solve the vehicle kinematic model based on the expected motion strategy to obtain the future motion state of the controlled vehicle entity.

[0099] The output information calculation module is used to calculate and output the self-state information and related information of the controlled vehicle entity in real time based on the user selection information.

[0100] The visualization module is used to render real-time images of the simulation environment based on the user's personalized demand information, and provide users with a visualized customized simulation environment.

[0101] The simulation environment logic and communication interface module is used to control the data set selection, visualization and scene simulation of the simulation environment based on user setting information and the configuration information required by the autonomous driving algorithm to be tested to meet the needs of reinforcement learning tasks, and realize information interaction through the request-question-answer mode.

[0102] The following is a specific example using Interaction as a traffic dataset and combining the attached Figure 2 The functions or effects of each of the above modules are further explained in detail.

[0103] like Figure 2 As shown, the data reading module is used to read the original Interaction dataset, where the original Interaction dataset contains two parts of file data, namely INTERACTION-Dataset-DR and INTERACTION-Dataset-TC, which represent the traffic files captured by drone aerial photography and traffic cameras respectively, and each file is composed of maps data and recorded_trackfiles data, representing map data and recorded vehicle track data. For example, a name such as DR_CHN_Merging_ZS.osm is a map file, DR represents the abbreviation of drone aerial photography, CHN represents the country of map collection is China, Merging represents the map scene is a convergence scene, ZS represents the map name, and .osm is the suffix of the map data. Vehicle_tracks_000.csv represents the track file of all vehicles in the map within a period of time. The track file includes 11 data items, track_id, frame_id, timestamp_ms, agent_type, x, y, vx, vy, psi_rad, length, and width, which respectively represent the vehicle number, time frame number, timestamp, vehicle type, x coordinate in the global coordinate system, y coordinate in the global coordinate system, x-direction speed in the global coordinate system, y-direction speed in the global coordinate system, vehicle orientation in the global coordinate system, vehicle length, and vehicle width. The data reading module will read the track csv file of the specified track number of the specified map in the recorded_trackfiles folder, and the function of reading the map is completed by the map initialization function in the scene entity module.

[0104] like Figure 2 As shown in the figure, the scene entity module is responsible for initializing and managing various entities in the simulation scene, including map entities, controlled vehicle entities, surrounding vehicle entities, etc. Specifically, the map entity initialization will call the load function in the lanelet2_io library in the Lanelet2 related support library to read the specified map data file. The function returns a pointer to the LaneletMap data type. The most important internal variables of the LaneletMap type are PointLayer, LineStringLayer, LaneletLayer, and AreaLayer. These internal variables can be simply regarded as dictionaries, which respectively store the structures of points, lines, and sections in the map. Through these variables, any map element of interest can be found in the form of ID and the relevant attributes can be calculated. The controlled vehicle entity is an important element defined by the simulator. Its static attributes such as length, width, and type are obtained when the trajectory file is read, while dynamic attributes such as position and speed are calculated through the vehicle motion model. The standard vehicle definition includes vehicle length and width, ID number, vehicle status information, timestamp, and other information. The difference between the surrounding vehicle entity and the controlled vehicle is that its dynamic state attributes are not calculated by the vehicle motion model, but by reading the real data in the trajectory file.

[0105] like Figure 2 As shown, the general geometric relationship calculation module provides basic plane geometric calculation functions (such as intersection judgment of two rectangles, shortest distance calculation of two rectangles, intersection judgment of a straight line and a rectangle, shortest distance calculation of a straight line and a rectangle, judgment of whether a point is inside or outside a polygon, curve length calculation, vector angle calculation, and overlap area calculation of two polygons) as well as the functions described above (such as vehicle entry and exit area judgment, two-vehicle position relationship judgment, two-vehicle collision judgment, vehicle orientation calculation, and vehicle and road centerline orientation angle calculation).

[0106] like Figure 2 As shown, the map relationship calculation module calls the open source Lanelet2 map-related calculation library to implement path planning, road segment relationship search, road conflict judgment and search, and vehicle position best matching road segment search. Specifically, the Lanelet2 map divides the map road into multiple segments, and each segment is defined as a lanelet. The attributes of the lanelet include the left boundary of the road, the right boundary, the center line of the road, the adjacent lanelet set (a total of four reachable direction sets in front, back, left, and right), and the overlapping / conflicting lanelet set (there are overlapping areas between lanelets, such as intersections and merging areas). Through the attributes of the lanelet, the planned path from the specified starting lanelet to the end lanelet and the various position relationship calculations described above can be obtained by depth-first search or breadth-first search.

[0107] Figure 3 It is a motion state and motion strategy analysis diagram of a controlled vehicle entity according to an embodiment of the present invention.

[0108] like Figure 3 As shown in the figure, the vehicle motion module provides basic vehicle kinematic model calculation (supports mass model and bicycle model), and inverse kinematics solution of data set record trajectory (inversely solve the control action based on the vehicle state information at two moments before and after). Specifically, taking the motion model bicycle model as an example: the vehicle state includes position coordinate, Coordinates, and yaw angle and speed value , the vehicle kinematic model is shown in equation group (I):

[0109] (I),

[0110] in, , represents the slip angle under the center of gravity of the vehicle. The control action of the vehicle is the acceleration value acc and the front wheel deflection angle .

[0111] The vehicle inverse kinematics model is shown in equation group (II):

[0112] (II).

[0113] like Figure 2 As shown, the output information calculation module provides a variety of optional controlled vehicle status information output configurations and calculations, including a local bird's-eye view of the vehicle, control vehicle status information (position, speed, direction), local map information (the distance of the controlled vehicle from the center line of the road, discrete path points of the planned path ahead of the road), surrounding vehicle information (distance, speed), whether there is a collision, the type of collision, and whether the end of the path has been reached.

[0114] like Figure 2 As shown, the visualization module provides image rendering for each simulation moment and supports a variety of custom visualization functions (such as visualization of planned paths, highlighting of vehicles in the intersection environment, comparison display of trajectories of controlled vehicles in corresponding data sets, and highlighting of controlled vehicles).

[0115] like Figure 2As shown, the simulation environment logic and communication interface module provides simulation environment class definition, simulation logic (initialization and reset, execution of actions, rendering, scene termination condition judgment) and communication interface (transmission of initialization, reset, action execution results, scene termination Boolean values ​​and other messages). The simulation environment logic here focuses on reinforcement learning tasks. According to the requirements of reinforcement learning, there must be map selection (specified ID method or random method), trajectory file selection (specified ID method or random method), controlled vehicle selection (specified ID method or random method), scene initialization or reset (such as map, vehicle state initialization, etc.), execution of actions (referring to controlling the controlled vehicle according to the control strategy and using the motion model to calculate the future vehicle motion state), rendering refers to calling the visualization module to visualize various information at the current moment, and the scene termination condition judgment uses the user-defined termination judgment function to judge the output Boolean value and related auxiliary information (such as collision information, etc.). The main function of the communication interface is to communicate the environment isolated from the simulator and the algorithm through port communication so that the external algorithm can call the simulator environment logic star to control the simulator externally, and pass the simulator's output information to the external program for processing. This communication mode adopts the request-question-answer mode.

[0116] Figure 4 A block diagram of an electronic device suitable for implementing an autonomous driving simulation method based on a real traffic data set according to an embodiment of the present invention is schematically shown.

[0117] like Figure 4 As shown, the electronic device 400 according to an embodiment of the present invention includes a processor 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage part 408 to a random access memory (RAM) 403. The processor 401 may include, for example, a general-purpose microprocessor (such as a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (for example, an application-specific integrated circuit (ASIC)), etc. The processor 401 may also include an onboard memory for caching purposes. The processor 401 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0118] In RAM 403, various programs and data required for the operation of electronic device 400 are stored. Processor 401, ROM 402 and RAM 403 are connected to each other via bus 404. Processor 401 performs various operations of the method flow according to the embodiment of the present invention by executing the programs in ROM 402 and / or RAM 403. It should be noted that the program can also be stored in one or more memories other than ROM 402 and RAM 403. Processor 401 can also perform various operations of the method flow according to the embodiment of the present invention by executing the programs stored in the one or more memories.

[0119] According to an embodiment of the present invention, the electronic device 400 may further include an input / output (I / O) interface 405, which is also connected to the bus 404. The electronic device 400 may further include one or more of the following components connected to the input / output (I / O) interface 405: an input portion 406 including a keyboard, a mouse, etc.; an output portion 407 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 408 including a hard disk, etc.; and a communication portion 409 including a network interface card such as a LAN card, a modem, etc. The communication portion 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output (I / O) interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed, so that a computer program read therefrom is installed into the storage portion 408 as needed.

[0120] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiment; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present invention is implemented.

[0121] According to an embodiment of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, it may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device. For example, according to an embodiment of the present invention, the computer-readable storage medium may include the ROM 402 and / or RAM 403 described above and / or one or more memories other than ROM 402 and RAM 403.

[0122] The embodiment of the present invention also includes a computer program product, which includes a computer program, and the computer program includes a program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the automatic driving simulation method based on a real traffic data set provided by the embodiment of the present invention.

[0123] The computer program executes the above functions defined in the system / device of the embodiment of the present invention when the processor 401 executes the computer program. According to the embodiment of the present invention, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0124] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 409, and / or installed from the removable medium 411. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0125] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 409, and / or installed from the removable medium 411. When the computer program is executed by the processor 401, the above functions defined in the system of the embodiment of the present invention are performed. According to the embodiment of the present invention, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0126] According to an embodiment of the present invention, the program code for executing the computer program provided by the embodiment of the present invention can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0127] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0128] It will be appreciated by those skilled in the art that the features described in the various embodiments of the present invention may be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, without departing from the spirit and teachings of the present invention, the features described in the various embodiments of the present invention may be combined and / or combined in various ways. All of these combinations and / or combinations fall within the scope of the present invention.

[0129] The embodiments of the present invention are described above. However, these embodiments are only for the purpose of illustration, and are not intended to limit the scope of the present invention. Although each embodiment is described above, it does not mean that the measures in each embodiment cannot be used in combination advantageously. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. An automatic driving simulation method based on a real traffic data set, characterized in that: The method comprises: Performing data parsing and format processing on a real traffic data set to obtain a dictionary-format traffic data set, initializing entities in a simulation environment based on the dictionary-format traffic data set, and managing multi-category entity instances obtained by initialization; Calculating geometric features of entities and geometric relationships between entities in the simulation environment using the multi-category entity instances, and processing the multi-category entity instances using a predefined real-time detection and tracking algorithm to obtain a global planning solution in the simulation environment; Based on the geometric relationship and the global planning scheme, the desired motion strategy of the controlled vehicle entity in the simulation environment is obtained by using the autonomous driving algorithm to be tested, and the vehicle kinematic model is solved based on the desired motion strategy to obtain the future motion state of the controlled vehicle entity; Based on the user selection information, the self-state information and related information of the controlled vehicle entity are calculated and output in real time, and based on the user's personalized demand information, the simulation environment is rendered in real time to provide the user with a visual customized simulation environment; Based on user setting information and the configuration information required by the autonomous driving algorithm to be tested, the simulation environment is controlled by data set selection, visualization and scene simulation to meet the needs of reinforcement learning tasks, and information interaction is achieved through a request-question-answer mode.

2. The method according to claim 1, characterized in that The real traffic data set is parsed and formatted to obtain a dictionary-formatted traffic data set including: Parsing the real traffic data to obtain a real traffic data file taken by a drone and a real traffic data file taken by a traffic camera; Performing format processing on the real traffic data file taken by the drone and the real traffic data file taken by the traffic camera to obtain a dictionary format traffic data set; The dictionary format traffic data set includes map files, vehicle trajectory files, vehicle surrounding environment data and traffic facility data; The vehicle trajectory file includes vehicle number, time frame number, timestamp, vehicle type, x-coordinate in the global coordinate system, y-coordinate in the global coordinate system, x-direction speed in the global coordinate system, y-direction speed in the global coordinate system, vehicle orientation in the global coordinate system, vehicle length, and vehicle width.

3. The method according to claim 1, characterized in that: Initializing entities in a simulation environment based on the dictionary format traffic data set and managing the multi-category entity instances obtained by initialization include: Reading the map file and traffic facility data in the dictionary format traffic data set, and initializing the map entity by setting a map pointer; Read the vehicle trajectory file in the dictionary format traffic data set to initialize the controlled vehicle entity; Read the vehicle surrounding environment data in the dictionary format traffic data set to initialize the surrounding environment vehicle entity in the surrounding environment of the controlled vehicle entity; Multi-category entity instances including the map entity, the controlled vehicle entity and the surrounding vehicle entity are managed.

4. The method according to claim 1, characterized in that: Calculating geometric features of entities in the simulation environment and geometric relationships between entities using the multi-category entity instances includes: Reading a map entity in the multi-category entity instance, and calculating a plane geometric feature on the map in the simulation environment, wherein the plane geometric feature includes a lane curvature and a centerline position; The map entities, controlled vehicle entities and surrounding vehicle entities in the multi-category entity instance are read, and the plane geometric relationship between the entities is calculated, wherein the plane geometric relationship includes the distance between the controlled vehicle entity and the surrounding vehicle entities, the distance between the controlled vehicle entity and the lane boundary, and the distance between the controlled vehicle entity and the lane centerline.

5. The method according to claim 1, characterized in that: Processing the multi-category entity instances using a predefined real-time detection and tracking algorithm to obtain a global planning solution in the simulation environment includes: The map entities in the multi-category entity instances are segmented, and based on the segmentation results, the predefined real-time detection and tracking algorithm is used to perform path planning, road segment relationship search, road conflict judgment, search for vehicle entities in the multi-category entity instances, and target matching road segment search for the vehicle entities in the simulation environment.

6. The method according to claim 1, characterized in that Based on the geometric relationship and the global planning scheme, the expected motion strategy of the controlled vehicle entity in the simulation environment is obtained by using the autonomous driving algorithm to be tested, and the vehicle kinematic model is solved based on the expected motion strategy to obtain the future motion state of the controlled vehicle entity, including: Constructing the vehicle kinematics model, and constructing a vehicle inverse kinematics model based on the vehicle kinematics model, and obtaining an expected acceleration and an expected yaw angle by analyzing the expected motion strategy; Solving the vehicle kinematics model based on the expected acceleration and the expected yaw angle to obtain a future motion state of the controlled vehicle entity; The actual motion state information of the controlled vehicle is obtained by parsing the dictionary format traffic data set, and the vehicle inverse kinematics model is solved based on the actual motion state information to obtain a predicted motion strategy of the controlled vehicle entity; The future motion state is used to control the controlled vehicle entity in real time in the simulation environment, and the automatic driving algorithm to be tested is evaluated based on the real-time control result.

7. The method according to claim 1, characterized in that The real-time calculation and output of the self-state information and associated information of the controlled vehicle entity based on the user selection information includes: Selecting a target vehicle as a controlled vehicle entity from the multi-category entity instances, and determining the self state information and associated information of the controlled vehicle entity to be output based on the user selection information; Based on the self-state information and associated information of the controlled vehicle entity, a local bird's-eye view of the controlled vehicle entity, real-time motion state information and real-time motion strategy, local map information and information of vehicle entities in the surrounding environment of the controlled vehicle entity are output.

8. The method according to claim 1, characterized in that: Based on the user's personalized demand information, the simulation environment is rendered in real time to provide the user with a visual customized simulation environment, including: Determine the user visualization result by analyzing the user's personalized demand information, wherein the user visualization result includes the visualization of the planned path, the visualization of the vehicle highlighting in the intersection range environment, the visualization of the trajectory file comparison display corresponding to the data set of the controlled vehicle entity, and the visualization of the highlighted display of the controlled vehicle entity; Based on the user visualization result, the simulation environment is rendered in real time to obtain the visualized customized simulation environment required by the user.

9. The method according to claim 1, characterized in that: Based on user setting information and configuration information required by the autonomous driving algorithm to be tested, the simulation environment is controlled by data set selection, visualization and scene simulation to meet the needs of the reinforcement learning task, and information interaction is achieved through a request-question-answer mode, including: Based on the user setting information and the configuration information required by the autonomous driving algorithm to be tested, define the simulation environment class, simulation logic scheme, communication mode and communication protocol, and complete the scene initialization of the simulation environment; Based on the scene initialization result of the simulation environment, selecting the map entity, the controlled vehicle entity and the surrounding vehicle entity required by the autonomous driving algorithm to be tested from the multi-category entity instances; Visually outputting relevant information of the controlled vehicle entity in the simulation environment at each moment and information associated with the autonomous driving algorithm to be tested; Dynamically start or stop the scenario simulation, and implement information interaction with the user through the request-question-answer mode.

10. An automatic driving simulation device based on a real traffic data set, characterized in that: The device comprises: The data reading module is used to parse and format the real traffic data set to obtain the dictionary format traffic data set; A scene entity module, used to initialize entities in a simulation environment based on the dictionary format traffic data set, and manage multi-category entity instances obtained by initialization; A general geometric relationship calculation module, used for calculating geometric features of entities in the simulation environment and geometric relationships between entities using the multi-category entity instances; A map relationship calculation module, used to process the multi-category entity instances using a predefined real-time detection and tracking algorithm to obtain a global planning solution in the simulation environment; A vehicle motion module, for obtaining an expected motion strategy of the controlled vehicle entity in the simulation environment using the autonomous driving algorithm to be tested based on the geometric relationship and the global planning scheme, and solving the vehicle kinematic model based on the expected motion strategy to obtain a future motion state of the controlled vehicle entity; An output information calculation module, used for calculating and outputting the self-state information and associated information of the controlled vehicle entity in real time based on the user selection information; A visualization module is used to render the simulation environment in real time based on the user's personalized demand information, and provide the user with a visualized customized simulation environment; The simulation environment logic and communication interface module is used to perform data set selection control, visualization control and scene simulation control on the simulation environment based on user setting information and the configuration information required by the autonomous driving algorithm to be tested to meet the needs of reinforcement learning tasks, and realize information interaction through a request-question-answer mode.

Images