Intersection scene generation method and device, electronic equipment and storage medium

Abstract

The invention discloses an intersection scene generation method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the dividing and splitting of a target scene, and obtaining a split scene; mapping is carried out according to the split scene, and an abstract space is constructed; a preset agent embedding strategy is called according to the abstract space, an intersection scene in the abstract space is generated in a combined mode, and the preset agent embedding strategy takes effect in the intersection scene. On one hand, the complexity of intersection interaction scene generation is reduced, and on the other hand, the authenticity and diversity of the scene are ensured.

Description

Technical Field

[0001] The present application relates to the technical field of autonomous driving simulation testing, and in particular to a method, device, electronic device, and storage medium for generating intersection scenes. Background Art

[0002] Scenario-based simulation testing plays an important role in the verification of autonomous driving systems. Through scenario-based simulation testing, the feasibility of algorithms in autonomous driving can be tested or the advantages and disadvantages of different strategies in the scenario can be pre-labeled.

[0003] In the face of complex urban environments, especially intersections with a large number of traffic participants and multiple conflict areas, the generation of high-fidelity and diverse interactive traffic scenes remains a huge challenge. Summary of the invention

[0004] The embodiments of the present application provide a method, device, electronic device, and storage medium for generating an intersection scene, so as to reduce the complexity of generating an intersection interaction scene while ensuring the authenticity and diversity of the scene.

[0005] The present application embodiment adopts the following technical solutions:

[0006] In a first aspect, an embodiment of the present application provides a method for generating a road intersection scene, wherein the method comprises:

[0007] Divide and split the target scene to obtain the split scene;

[0008] Mapping is performed based on the split scenes to construct an abstract space;

[0009] The preset intelligent agent embedding strategy is called according to the abstract space, and the intersection scene under the abstract space is generated by combination, and the preset intelligent agent embedding strategy is effective in the intersection scene.

[0010] In some embodiments, dividing and splitting the target scene to obtain the split scenes includes:

[0011] Dividing the target scene into a real intersection scene in the world space;

[0012] Based on the driving conditions of vehicles in the real intersection scene, two types of basic intersection scenes are divided, including one-to-many crossing scenes and one-to-many merging scenes. The one-to-many crossing scenes refer to that the vehicle and other traffic participants will eventually be separated in the lateral space, and the one-to-many merging scenes refer to that the vehicle and other traffic participants will eventually merge in the lateral space.

[0013] In some embodiments, dividing and splitting the target scene to obtain split scenes further includes:

[0014] Setting a one-to-many walking interaction strategy based on the one-to-many walking scenario;

[0015] The ego vehicle executes a deceleration and parking strategy before reaching the conflict area, during which it checks whether there is a gap;

[0016] If there is a gap, a passing strategy is executed, wherein the passing strategy includes giving way to vehicles on the right and simultaneously grabbing vehicles on the left;

[0017] and / or,

[0018] Setting a one-to-many import interaction strategy based on the one-to-many import scenario;

[0019] The ego vehicle executes a deceleration and parking strategy before reaching the conflict area, during which it checks whether there is a gap;

[0020] If there is a gap, a merging strategy is executed, which includes following the vehicle in front and simultaneously overtaking the vehicle behind.

[0021] In some embodiments, mapping is performed according to the split scenes to construct an abstract space, including:

[0022] The trajectory of the ego vehicle is abstracted as a straight line and the interaction between the ego vehicle and other intelligent agents is abstracted as an interactive game or an intersection game;

[0023] Mapping is performed based on the split scenes, and the constructed abstract space only contains the interactive scenes, which are used as the preset scenes for the first intersection;

[0024] Mapping is performed based on the split scenes, and the constructed abstract space includes both interaction and intersection scenes, with intersection first and interaction later, but only one intersection is included and used as the preset scene for the second intersection;

[0025] Mapping is performed based on the split scenes, and the constructed abstract space includes both interaction and intersection scenes, with interaction occurring before intersection, but only one intersection is included and is used as the preset scene for the third intersection;

[0026] Mapping is performed based on the split scenes, and the constructed abstract space includes both interaction and intersection scenes, with interaction occurring before intersection, but including multiple intersections, and serving as the fourth intersection preset scene;

[0027] The first intersection preset scene, the second intersection preset scene, the third intersection preset scene and the fourth intersection preset scene all include multiple game trajectories.

[0028] In some embodiments, the abstract space only includes an interactive scene, and the calling of a preset agent embedding strategy according to the abstract space to combine and generate an intersection scene under the abstract space, wherein the preset agent embedding strategy takes effect in the intersection scene, includes:

[0029] According to the interactive scenario, the preset agent embedding strategy is configured as a car-following strategy, wherein the car-following strategy at least includes a car-following model;

[0030] Based on the vehicle-following strategy, a corresponding traffic flow is generated.

[0031] In some embodiments, the abstract space includes a scene with intersection in front and interaction in the back, and the calling of a preset agent embedding strategy according to the abstract space to combine and generate an intersection scene under the abstract space, wherein the preset agent embedding strategy takes effect in the intersection scene, includes:

[0032] According to the scenario of the intersection first and the interaction later, the preset agent embedding strategy is configured as a car-following strategy and a one-to-many crossing strategy, and the car-following strategy at least includes a car-following model;

[0033] Based on the vehicle following strategy and the one-to-many traveling strategy, a corresponding traffic flow is generated.

[0034] In some embodiments, the abstract space includes a scene in which interaction precedes and intersection follows, and only one intersection is included. The calling of a preset agent embedding strategy according to the abstract space to combine and generate an intersection scene under the abstract space, wherein the preset agent embedding strategy is effective in the intersection scene, includes:

[0035] According to the scenario in which the interaction occurs first and the intersection occurs later and the scenario only includes one intersection, the preset agent embedding strategy is configured as a car-following strategy, and the car-following strategy at least includes a car-following model;

[0036] Based on the vehicle-following strategy, a corresponding traffic flow is generated.

[0037] In some embodiments, the abstract space includes a scene in which interaction precedes and intersection follows, and includes two intersections. The calling of a preset agent embedding strategy according to the abstract space to combine and generate an intersection scene under the abstract space, wherein the preset agent embedding strategy takes effect in the intersection scene, includes:

[0038] According to the scenario in which the interaction comes first and the intersection comes later and includes two intersections, the preset agent embedding strategy is configured as a following strategy and a one-to-many merging strategy, and the following strategy at least includes a following model;

[0039] A corresponding traffic flow is generated according to the vehicle following strategy and the one-to-many merging strategy.

[0040] In some embodiments, the method further comprises: a step of controlling the data diversity of the intersection scene:

[0041] In the process of calling the preset intelligent agent embedding strategy according to the abstract space and combining to generate the intersection scene in the abstract space, the traffic flow is obtained by controlling the driving along the preset path and randomly generating it.

[0042] In a second aspect, an embodiment of the present application further provides a device for generating a road intersection scene, wherein the device comprises:

[0043] A scene division and splitting module is used to divide and split the target scene to obtain the split scene;

[0044] A mapping module is used to map the split scenes and construct an abstract space;

[0045] A generation module is used to call a preset intelligent agent embedding strategy according to the abstract space, and combine and generate an intersection scene under the abstract space, wherein the preset intelligent agent embedding strategy takes effect in the intersection scene.

[0046] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor; and a memory arranged to store computer executable instructions, wherein the executable instructions, when executed, cause the processor to perform the above method.

[0047] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes the above method.

[0048] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: first, the target scene is divided and split to obtain the split scenes. Then, mapping is performed according to the split scenes to construct an abstract space. Finally, the preset intelligent agent embedding strategy is called according to the abstract space to combine and generate the intersection scene under the abstract space, and the preset intelligent agent embedding strategy is effective in the intersection scene. The above method reduces the complexity of the intersection interaction scene generation while ensuring the authenticity and diversity of the scene. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0050] Figure 1 This is one of the schematic diagrams of the world space to abstract space in the embodiment of the present application;

[0051] Figure 2 This is the second schematic diagram of the world space to abstract space in the embodiment of the present application;

[0052] Figure 3 This is a schematic diagram of one-to-many travel in an embodiment of the present application;

[0053] Figure 4 A schematic diagram of one-to-many import in an embodiment of the present application;

[0054] FIG5( a ) is one of the various intersection game scenarios in the embodiments of the present application;

[0055] FIG5( b ) is the second of the various intersection game scenarios in the embodiment of the present application;

[0056] FIG5( c ) is a third of various intersection game scenarios in an embodiment of the present application;

[0057] FIG5( d ) is a fourth of various intersection game scenarios in the embodiments of the present application;

[0058] Figure 6 A schematic diagram of the flow of a method for generating a road intersection scene in an embodiment of the present application;

[0059] Figure 7 This is a schematic diagram of the structure of a device for generating a road intersection scene in an embodiment of the present application;

[0060] Figure 8 This is a schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0062] When verifying the autonomous driving system, simulation testing is usually divided into knowledge (or rule)-based methods and data-driven methods for intersection scene generation, thereby giving interaction strategies to traffic participants in each scene.

[0063] (1) Knowledge-based (or rule-based) methods usually rely on a large number of manually designed rules when generating intersection scenarios. These manually designed rules may lead to insufficient authenticity of the interaction, and a large number of rules makes system maintenance more difficult, especially when the number of participants at the intersection increases.

[0064] (2) Data-driven methods usually generate more realistic scenes. However, due to the limitations of the dangerousness or rarity of certain data, it is difficult for data-driven methods to generate sufficiently diverse scene data.

[0065] In addition, there are some world model-based methods, such as directly generating scenes based on video data, but this method still cannot get rid of the defects of data-driven methods, and the technology is more difficult to implement.

[0066] In view of the shortcomings of the above methods, an intersection scene generation method based on abstract space is proposed in the embodiment of the present application, which only needs to implement a following algorithm, or a one-to-many crossing algorithm (specifically, a certain intelligent agent passes through a certain traffic flow), or a one-to-many merging algorithm (specifically, a certain intelligent agent merges into a certain traffic flow), thereby greatly reducing the complexity of generating intersection interaction scenes. Through the intersection scene generation method in the embodiment of the present application, the complex intersection multi-agent game is simplified to include no game scene or only one basic game scene, reducing the complexity of generating intersection interaction scenes, while ensuring the authenticity and diversity of the scene.

[0067] The verification of the autonomous driving system requires simulation testing based on scene generation. First, simulation testing can verify the feasibility of the algorithm in the autonomous driving system. Secondly, in the process of reinforcement learning training, feedback from the environment is usually required. If there is no feedback, training and testing are required on the actual vehicle. Finally, the disadvantage of modeling based on real space is that it is limited by the type and amount of collected data, and it is more difficult to collect dangerous scenes. However, it can be achieved in abstract space, and data that is not easy to occur in real scenes can be obtained. In other words, the intersection scenes generated based on abstract space are random and cover more areas. In addition, it can also be used for pre-labeling, which makes it easier to obtain true values ​​during reinforcement learning training, thereby realizing a learning process based on self-supervision.

[0068] In order to illustrate the definition of abstract space, a one-to-one correspondence between world space and abstract space is given by taking an unprotected left turn as an example, as follows: Figure 1 As shown in the figure, in the scenario of an unprotected left turn of the ego vehicle, the ego vehicle will engage in games with other agents at most in three key areas at the same time, where the trajectory of the ego vehicle is abstracted as a straight line, and the interaction with other agents is abstracted as interactive games or games. However, the interactions between other agents are either unaffected by each other, that is, only the following behavior on the same route (non-game behavior). Alternatively, it only includes interactive games between each other, or only includes intersection games between each other. Figure 2As shown in the figure, in another scenario of turning left at an intersection, we only focus on how to randomly reflect interactivity. The vertical line in the world space represents the driving trajectory of the vehicle. In the abstract space, only the trajectory of the vehicle is abstracted, and the interaction with other intelligent agents is abstracted as an interactive game or game.

[0069] The most complex intersection scenario occurs in the unprotected left turn scenario. That is, straight vehicles from left to right, left-turning vehicles from right to left (or straight vehicles), and straight vehicles from top to bottom (or right-turning vehicles) exist at the same time, and there may be multiple game vehicles on each trajectory, thus splitting into two basic intersection scenarios, namely: one-to-many crossing scenario and one-to-many merging scenario. The schematic diagram of one-to-many crossing is as follows: Figure 3 As shown, the one-to-many interaction diagram is as follows Figure 4 shown.

[0070] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0071] The present application embodiment provides a method for generating a road intersection scene, such as Figure 6 As shown, a schematic flow chart of a method for generating a road intersection scene in an embodiment of the present application is provided, and the method at least includes the following steps S610 to S630:

[0072] Step S610, dividing and splitting the target scene to obtain split scenes.

[0073] The target scene may be selected by taking the most complex intersection scene as an example, and by dividing and splitting the complex scene, a split scene may be obtained.

[0074] For intersection scenarios, the most complex intersection scenario occurs in unprotected left turns. That is, straight vehicles from left to right, left-turning vehicles from right to left (or straight vehicles), and straight vehicles from top to bottom (or right-turning vehicles) exist at the same time, and there may be multiple game vehicles on each trajectory. Based on the most complex intersection scenario, the basic scenario is obtained, and then the basic scenario is split.

[0075] Step S620: Mapping is performed according to the split scenes to construct an abstract space.

[0076] Please refer to Figure 1 or Figure 2, a one-to-many game in an abstract space can have a variety of most complex scenarios, which include interaction or intersection scenarios. It can be understood that in the most complex scenarios, some interaction trajectories may not exist or do not contain other agents. For the interaction scenario, it can be understood that the ego vehicle and other traffic participants will eventually separate in the lateral space. For the intersection scenario, it can be understood that the ego vehicle and other traffic participants will eventually converge in the lateral space. For the abstract space, it can be considered that the state of other traffic participants is abstracted as the distance from the conflict area and the time of entering and leaving.

[0077] Step S630, calling a preset agent embedding strategy according to the abstract space, combining and generating an intersection scene under the abstract space, and the preset agent embedding strategy takes effect in the intersection scene.

[0078] According to the constructed abstract space, the corresponding agent embedding strategy is called to combine and generate the intersection scene under the corresponding abstract space, that is, the multi-agent spatiotemporal joint trajectory. According to the abstract space and the preset agent embedding strategy, the intersection scene under the abstract space is combined and the traffic data takes effect.

[0079] It should be noted that the above-mentioned agent embedding strategies mainly include but are not limited to following vehicle strategy, serial strategy, and merging into traffic flow strategy.

[0080] Please continue to refer to Figure 1 In the unprotected left turn scenario, the ego vehicle will engage in games with other agents at most in three key areas at the same time, where the trajectory of the ego vehicle is abstracted as a straight line (vertical direction), and the interaction with other agents is abstracted as interactive games or games. At the same time, the interactions between other agents at the intersection are not affected, and only the following behavior (non-game behavior) on the same route is involved. Alternatively, it only includes interactive games between each other, or it only includes intersection games between each other.

[0081] Through the above method, the target scene is divided and split to obtain the split scene, which overcomes the complexity of rule design based on knowledge (rules) to generate intersection scenes and simplifies the complex intersection multi-agent game to contain no game scene or only one basic game scene.

[0082] Through the above method, the preset intelligent agent embedding strategy is called according to the abstract space, and the intersection scene under the abstract space is generated in combination. The preset intelligent agent embedding strategy is effective in the intersection scene, overcoming the defects of data-driven generation of intersection scenes, and can control the complexity and diversity of the scene to a certain extent.

[0083] Through the above method, mapping is performed according to the split scenes to construct an abstract space. To a certain extent, it can provide optimized passage trajectories for the learning-based intersection traffic model and perform self-supervisory evaluation for different trajectories based on the designed loss function, thereby facilitating model pre-training.

[0084] In one embodiment of the present application, the target scene is divided and split to obtain the split scenes, including: dividing the target scene into a real intersection scene in the world space; based on the driving conditions of the vehicles in the real intersection scene, splitting into two types of basic intersection scenes, the basic intersection scenes including a one-to-many passing scene and a one-to-many merging scene, the one-to-many passing scene means that the vehicle and other traffic participants will eventually separate in the lateral space, and the one-to-many merging scene means that the vehicle and other traffic participants will eventually merge in the lateral space.

[0085] The real intersection scene of dividing the target scene in the world space is described in detail by taking the most complex intersection scene as an example. It can be understood that it is not used to limit the actual protection scope of the present application.

[0086] The most complex intersection scenario occurs in the unprotected left turn scenario, that is, straight vehicles from left to right, left-turning vehicles from right to left (or straight vehicles), and straight vehicles from top to bottom (or right-turning vehicles) exist at the same time, and there may be multiple game vehicles on each trajectory, thus splitting into two basic intersection scenarios.

[0087] That is: one-to-many travel scenario and one-to-many import scenario. The one-to-many travel diagram is as follows: Figure 3 As shown, the one-to-many interaction diagram is as follows Figure 4 As shown. It can be understood that the one-to-many passing scenario can be considered as a certain agent passing through a certain traffic flow. The one-to-many merging scenario can be considered as a certain agent merging into a certain traffic flow.

[0088] In one embodiment of the present application, the target scene is divided and split to obtain the split scenes, and further includes: setting a one-to-many walking interaction strategy based on the one-to-many walking scenario; the vehicle executes a deceleration and stopping strategy before reaching the conflict area, and checks whether there is a gap during this process; if there is a gap, the walking strategy is executed, and the walking strategy includes giving way to the vehicle on the right and simultaneously overtaking the vehicle on the left; and / or, setting a one-to-many merging interaction strategy based on the one-to-many merging scenario; the vehicle executes a deceleration and stopping strategy before reaching the conflict area, and checks whether there is a gap during this process; if there is a gap, the merging strategy is executed, and the merging strategy includes following the vehicle in front and simultaneously overtaking the vehicle behind.

[0089] Specifically, when designing an interactive strategy for one-to-many weaving, the vehicle executes a deceleration and parking strategy before reaching the conflict area. During this process, it checks whether there is a gap: it can give way to the vehicle on the right and rush to the left at the same time. At this time, the acceleration of giving way acc is greater than the acceleration of rushing acc. If there is a solution to the interactive game, the weaving is executed.

[0090] Specifically, when designing an interactive strategy for a one-to-many merging situation, the ego vehicle executes a deceleration and stopping strategy before reaching the conflict area. During this process, it is checked whether there is a gap: that is, the vehicle can follow the vehicle in front and at the same time overtake the vehicle behind. At this time, the following vehicle acceleration acc is greater than the overtaking acceleration acc. If there is a solution to the merging game, the merging strategy is executed.

[0091] From the above, we can see that for the weaving strategy, the gap refers to the space between two cars on the left and right. For the merging strategy, the gap refers to the space between two cars in front and behind. Specifically, when weaving, you should first slow down and wait for a gap to appear. This gap is sufficient to let the car on the right pass and rush the car on the left at the same time, and then the weaving strategy is established.

[0092] In addition, regarding the specific construction and calculation of the interactive game, the following method can be referred to. It can be understood that the specific construction and calculation of the interactive game described below is only a feasible implementation method and does not limit the protection scope of this application.

[0093] In one embodiment of the present application, the constructed constraint items include: pre-assuming that the merging traffic flow always requires the vehicle to overtake other vehicles; based on the assumption that the merging traffic flow always requires the vehicle to overtake other vehicles, a condition is obtained based on the speed of the other vehicle after accelerating within a preset time, that the vehicle must exit the merging area before the rear vehicle enters the merging area.

[0094] In order to better improve the safety of the process of merging into traffic, in the above-constructed intersection game problem, the merging traffic always requires the ego vehicle to overtake the following vehicle, that is, the constraint item is that the ego vehicle must exit the intersection area before the following vehicle enters the intersection area. The intersection game problem takes into account the uncertainty of the following vehicle's behavior, that is, the following vehicle may also overtake the ego vehicle during the process of overtaking, which improves the safety of the intersection game process.

[0095] In order to cope with the uncertainty of the predicted trajectory and the unpredictability of the following vehicle's behavior (such as sudden acceleration), the vehicle needs to always assume that the following vehicle has accelerated (such as taking the maximum acceleration) within a certain period of time to calculate the speed when trying to overtake the vehicle. The "certain period of time" is usually as short as 1-2 seconds.

[0096] In one embodiment of the present application, the calculation of the intersection area includes: calculating the positions of the vehicle and other vehicles that ultimately meet longitudinal safety, the other vehicles including the rear vehicle; determining the intersection area information based on the maximum braking amount taken by the rear vehicle, and calculating the time for the rear vehicle to arrive at and exit the intersection area based on the current speed information of the rear vehicle.

[0097] When calculating the final longitudinal safety position of the ego vehicle and the following vehicle (the condition satisfied during cut-in), the intersection area information is obtained after assuming the maximum braking amount taken by the following vehicle. The time for the following vehicle to arrive at and leave the intersection area is calculated based on the current speed information of the following vehicle.

[0098] In order to be compatible with the courtesy of merging into the traffic flow, in the above-constructed intersection game problem, when the longitudinal safety of the self-vehicle and the following vehicle is satisfied, it is assumed that the maximum braking taken by the following vehicle meets certain conditions.

[0099] In the construction of the above-mentioned intersection game problem, the maximum braking (configurable) adopted by the rear vehicle is assumed when the longitudinal safety of the self-vehicle and the rear vehicle is satisfied, which is compatible with the politeness in the process of merging into the traffic flow.

[0100] It can be understood that the following distance of the rear vehicle is calculated by calculating the time when the rear vehicle arrives at and leaves the intersection area according to the current speed information of the rear vehicle.

[0101] In one embodiment of the present application, the following vehicle problem is constructed based on the self-vehicle information and the preceding vehicle information, and the second acceleration information is obtained by solving the problem, including: according to the self-vehicle information and the preceding vehicle information, a virtual obstacle with a speed of 0 is set at the location where the self-vehicle and the preceding vehicle have their first lateral collision, to construct a virtual following vehicle problem; according to the self-vehicle information and the preceding vehicle information, a preceding vehicle following vehicle problem is constructed.

[0102] In one embodiment of the present application, the mapping is performed according to the split scenes to construct an abstract space, including: abstracting the trajectory of the vehicle as a straight line and abstracting the interaction between the vehicle and other intelligent bodies as an interactive game or an intersection game; mapping is performed according to the split scenes, and the abstract space constructed only includes the interaction scene, and is used as the first intersection preset scene; mapping is performed according to the split scenes, and the abstract space constructed includes both interaction and intersection scenes, and the intersection occurs first and then the interaction, but only includes one intersection, and is used as the second intersection preset scene; mapping is performed according to the split scenes, and the abstract space constructed includes both interaction and intersection scenes, and the interaction occurs first and then the intersection, but only includes one intersection, and is used as the third intersection preset scene; mapping is performed according to the split scenes, and the abstract space constructed includes both interaction and intersection scenes, and the interaction occurs first and then the intersection, but only includes one intersection, and is used as the fourth intersection preset scene; the first intersection preset scene, the second intersection preset scene, the third intersection preset scene, and the fourth intersection preset scene all include multiple game trajectories.

[0103] By mapping the split scenes, an abstract space can be constructed. There are at most four types of one-to-many games in the abstract space (the most complex scenarios), among which some interaction trajectories may not exist or contain other intelligent agents.

[0104] Please refer to Figure 5(a), Scenario 1: All three game trajectories are interactive games, that is, they only include interactive scenarios. Interactive scenarios: The ego vehicle and other traffic participants will eventually separate in the lateral space.

[0105] Please refer to Figure 5(b), Scenario 2: For the three game trajectories, it includes both interaction and intersection scenarios, and first intersection and then interaction, but only one intersection. Intersection scenario: The vehicle and other traffic participants finally converge in the lateral space.

[0106] Please refer to Figure 5(c), scenario three: the three game trajectories are: including both interaction and intersection scenarios, and interaction comes first and then intersection, but only one intersection is included.

[0107] Please refer to Figure 5(d), scenario 4: the three game trajectories are: including both interaction and intersection scenarios, and interaction comes first and then intersection, but contains multiple intersections.

[0108] Through the above method, the complex intersection multi-agent game is simplified to contain no game scenario or only one basic game scenario. In addition, the complexity and diversity of the scenario can be controlled, specifically the number of traffic flows participating in the game and the number of agents in each traffic flow can be controlled.

[0109] In one embodiment of the present application, the abstract space only contains interaction scenes, and the preset intelligent agent embedding strategy is called according to the abstract space to combine and generate an intersection scene under the abstract space. The preset intelligent agent embedding strategy is effective in the intersection scene, including: according to the only interaction scene, the preset intelligent agent embedding strategy is configured as a following strategy, and the following strategy at least includes a following model; based on the following strategy, a corresponding traffic flow is generated.

[0110] Please refer to Figure 5(a). For the interactive scenario only, for the ego vehicle, the abstract space is represented by three traffic flows that do not affect each other. At this time, it is only necessary to generate traffic flows based on a simple following model.

[0111] It can be understood that the above-mentioned car-following models mainly refer to Adaptive Cruise Control (ACC) and Cooperative Adaptive Cruise Control (CACC). ACC and CACC are important longitudinal following technologies in the study of traffic flow of autonomous vehicles.

[0112] The above-mentioned intersection scene generation based on abstract space can, to a certain extent, provide optimized passage trajectories for the learning-based intersection traffic model and perform self-supervised evaluation for different trajectories based on the designed loss function, thereby facilitating model pre-training.

[0113] In one embodiment of the present application, the abstract space includes a scenario with intersection in front and interaction in the back, and the preset intelligent agent embedding strategy is called according to the abstract space to combine and generate an intersection scene under the abstract space. The preset intelligent agent embedding strategy is effective in the intersection scene, including: according to the scenario with intersection in front and interaction in the back, the preset intelligent agent embedding strategy is configured as a following strategy and a one-to-many passing strategy, and the following strategy at least includes a following model; based on the following strategy and the one-to-many passing strategy, the corresponding traffic flow is generated.

[0114] Please refer to Figure 5(b). For the scenario with intersection in front and interaction in the back, the abstract space is represented by two traffic flows that do not affect each other (which can be generated based on the car-following model), and a traffic flow containing one-to-many crossing (which can be generated based on the car-following model and the one-to-many crossing strategy). The one-to-many crossing strategy can be considered as a crossing strategy for a certain intelligent agent to pass through a certain traffic flow.

[0115] The above-mentioned intersection scene generation based on abstract space can, to a certain extent, provide optimized passage trajectories for the learning-based intersection traffic model and perform self-supervised evaluation for different trajectories based on the designed loss function, thereby facilitating model pre-training.

[0116] In one embodiment of the present application, the abstract space includes a scenario in which interaction comes first, convergence comes later and only one convergence is included, and the preset intelligent agent embedding strategy is called according to the abstract space to combine and generate an intersection scene under the abstract space, and the preset intelligent agent embedding strategy is effective in the intersection scene, including: according to the scenario in which interaction comes first, convergence comes later and only one convergence is included, the preset intelligent agent embedding strategy is configured as a following strategy, and the following strategy at least includes a following model; based on the following strategy, a corresponding traffic flow is generated.

[0117] Please refer to Figure 5(c). For the scenario with interaction first and intersection later (only one intersection is included), it is the same as the scenario with only interaction in the abstract space, which is manifested as three traffic flows that do not affect each other. The intelligent agents in each traffic flow only need to complete the interaction based on a simple following model.

[0118] The above-mentioned intersection scene generation based on abstract space can, to a certain extent, provide optimized passage trajectories for the learning-based intersection traffic model and perform self-supervised evaluation for different trajectories based on the designed loss function, thereby facilitating model pre-training.

[0119] In one embodiment of the present application, the abstract space includes a scenario in which interaction occurs first, convergence occurs later, and two convergences are included. The preset intelligent agent embedding strategy is called according to the abstract space to combine and generate an intersection scene under the abstract space. The preset intelligent agent embedding strategy is effective in the intersection scene, including: according to the scenario in which interaction occurs first, convergence occurs later, and two convergences are included, the preset intelligent agent embedding strategy is configured as a following strategy and a one-to-many merging strategy, and the following strategy at least includes a following model; and corresponding traffic flow is generated according to the following strategy and the one-to-many merging strategy.

[0120] Please refer to Figure 5(d). For the scenario of interaction first and merging later (including two merging), the abstract space is represented by two traffic flows that do not affect each other and a traffic flow with one-to-many merging (which can be generated based on the car-following model and the one-to-many merging strategy). The one-to-many merging strategy can be considered as a merging strategy for a certain agent to merge into a certain traffic flow.

[0121] The above-mentioned intersection scene generation based on abstract space can, to a certain extent, provide optimized passage trajectories for the learning-based intersection traffic model and perform self-supervised evaluation for different trajectories based on the designed loss function, thereby facilitating model pre-training.

[0122] In one embodiment of the present application, the method also includes: a step of controlling the data diversity of the intersection scene: in the process of calling the preset intelligent agent embedding strategy according to the abstract space and combining to generate the intersection scene under the abstract space, the traffic flow is obtained by controlling the driving along the preset path and randomly generating it.

[0123] The above generation of abstract spatial intersection scenes can achieve the randomness of the behavior of the intelligent agents in the scene to a certain extent. For example, by controlling the randomness of the behavior of the leading vehicle in each traffic flow participating in the game, the data diversity of the intersection scene can be controlled.

[0124] The embodiment of the present application also provides a road intersection scene generation device 700, such as Figure 7 As shown, a schematic diagram of the structure of the intersection scene generation device in an embodiment of the present application is provided, and the intersection scene generation device 700 at least includes: an acquisition module 710, a mapping module 720 and a generation module 730, wherein:

[0125] In one embodiment of the present application, the acquisition module 710 is specifically used to divide and split the target scene to obtain the split scenes.

[0126] The target scene may be selected by taking the most complex intersection scene as an example, and by dividing and splitting the complex scene, a split scene may be obtained.

[0127] For intersection scenarios, the most complex intersection scenario occurs in unprotected left turns. That is, straight vehicles from left to right, left-turning vehicles from right to left (or straight vehicles), and straight vehicles from top to bottom (or right-turning vehicles) exist at the same time, and there may be multiple game vehicles on each trajectory. Based on the most complex intersection scenario, the basic scenario is obtained, and then the basic scenario is split.

[0128] In one embodiment of the present application, the mapping module 720 is specifically used to: perform mapping according to the split scenes to construct an abstract space.

[0129] Please refer to Figure 1 or Figure 2 , a one-to-many game in an abstract space can have a variety of most complex scenarios, which include interaction or intersection scenarios. It can be understood that in the most complex scenarios, some interaction trajectories may not exist or do not contain other agents. For the interaction scenario, it can be understood that the ego vehicle and other traffic participants will eventually separate in the lateral space. For the intersection scenario, it can be understood that the ego vehicle and other traffic participants will eventually converge in the lateral space. For the abstract space, it can be considered that the state of other traffic participants is abstracted as the distance from the conflict area and the time of entering and leaving.

[0130] In one embodiment of the present application, the generation module 730 is specifically used to: call a preset intelligent agent embedding strategy according to the abstract space, combine and generate an intersection scene under the abstract space, and the preset intelligent agent embedding strategy is effective in the intersection scene.

[0131] According to the constructed abstract space, the corresponding agent embedding strategy is called to combine and generate the intersection scene under the corresponding abstract space, that is, the multi-agent spatiotemporal joint trajectory. According to the abstract space and the preset agent embedding strategy, the intersection scene under the abstract space is combined and the traffic data takes effect.

[0132] It should be noted that the above-mentioned agent embedding strategies mainly include but are not limited to following vehicle strategy, serial strategy, and merging into traffic flow strategy.

[0133] Please continue to refer to Figure 1 In the unprotected left turn scenario, the ego vehicle will engage in games with other agents at most in three key areas at the same time, where the trajectory of the ego vehicle is abstracted as a straight line (vertical direction), and the interaction with other agents is abstracted as interactive games or games. At the same time, the interactions between other agents at the intersection are not affected, and only the following behavior (non-game behavior) on the same route is involved. Alternatively, it only includes interactive games between each other, or it only includes intersection games between each other.

[0134] In one embodiment of the present application, the scene division and splitting module 710 is also used to

[0135] Dividing the target scene into a real intersection scene in the world space;

[0136] Based on the driving conditions of vehicles in the real intersection scene, two types of basic intersection scenes are divided, including one-to-many crossing scenes and one-to-many merging scenes. The one-to-many crossing scenes refer to that the vehicle and other traffic participants will eventually be separated in the lateral space, and the one-to-many merging scenes refer to that the vehicle and other traffic participants will eventually merge in the lateral space.

[0137] In one embodiment of the present application, the scene division and splitting module 710 is also used to

[0138] Setting a one-to-many walking interaction strategy based on the one-to-many walking scenario;

[0139] The ego vehicle executes a deceleration and parking strategy before reaching the conflict area, during which it checks whether there is a gap;

[0140] If there is a gap, a passing strategy is executed, wherein the passing strategy includes giving way to vehicles on the right and simultaneously grabbing vehicles on the left;

[0141] and / or,

[0142] Setting a one-to-many import interaction strategy based on the one-to-many import scenario;

[0143] The ego vehicle executes a deceleration and parking strategy before reaching the conflict area, during which it checks whether there is a gap;

[0144] If there is a gap, a merging strategy is executed, which includes following the vehicle in front and simultaneously overtaking the vehicle behind.

[0145] In one embodiment of the present application, the mapping module 720 is also used to

[0146] The trajectory of the ego vehicle is abstracted as a straight line and the interaction between the ego vehicle and other intelligent agents is abstracted as an interactive game or an intersection game;

[0147] Mapping is performed based on the split scenes, and the constructed abstract space only contains the interactive scenes, which are used as the preset scenes for the first intersection;

[0148] Mapping is performed based on the split scenes, and the constructed abstract space includes both interaction and intersection scenes, with intersection first and interaction later, but only one intersection is included and used as the preset scene for the second intersection;

[0149] Mapping is performed based on the split scenes, and the constructed abstract space includes both interaction and intersection scenes, with interaction occurring before intersection, but only one intersection is included and is used as the preset scene for the third intersection;

[0150] Mapping is performed based on the split scenes, and the constructed abstract space includes both interaction and intersection scenes, with interaction occurring before intersection, but including multiple intersections, and serving as the fourth intersection preset scene;

[0151] The first intersection preset scene, the second intersection preset scene, the third intersection preset scene and the fourth intersection preset scene all include multiple game trajectories.

[0152] In one embodiment of the present application, the abstract space only includes interaction scenes, and the generating module 730 is further used to configure the preset agent embedding strategy as a car-following strategy according to the interaction scenes only, and the car-following strategy at least includes a car-following model;

[0153] Based on the vehicle-following strategy, a corresponding traffic flow is generated.

[0154] In one embodiment of the present application, the abstract space includes a scene of intersection in front and interaction in the back, and the generation module 730 is further used to configure the preset agent embedding strategy as a following strategy and a one-to-many crossing strategy according to the scene of intersection in front and interaction in the back, and the following strategy at least includes a following model;

[0155] Based on the vehicle following strategy and the one-to-many traveling strategy, a corresponding traffic flow is generated.

[0156] In one embodiment of the present application, the abstract space includes a scenario in which the interaction comes first and the intersection comes later and only one intersection is included. The generation module 730 is further used to configure the preset agent embedding strategy as a car-following strategy according to the scenario in which the interaction comes first and the intersection comes later and only one intersection is included. The car-following strategy at least includes a car-following model.

[0157] Based on the vehicle-following strategy, a corresponding traffic flow is generated.

[0158] In one embodiment of the present application, the abstract space includes a scenario in which the interaction is first, the intersection is later, and the intersection is included twice. The generation module 730 is further used to configure the preset agent embedding strategy as a following strategy and a one-to-many merging strategy according to the scenario in which the interaction is first, the intersection is later, and the intersection is included twice. The following strategy at least includes a following model.

[0159] A corresponding traffic flow is generated according to the vehicle following strategy and the one-to-many merging strategy.

[0160] In one embodiment of the present application, it also includes: a data diversity module for controlling the intersection scene, which is used to control the traffic flow by driving along a preset path and randomly generating the traffic flow in the process of calling the preset intelligent agent embedding strategy according to the abstract space and combining to generate the intersection scene under the abstract space.

[0161] It can be understood that the above-mentioned intersection scene generation device can implement each step of the intersection scene generation method provided in the aforementioned embodiment, and the relevant explanations about the intersection scene generation method are applicable to the intersection scene generation device, which will not be repeated here.

[0162] Figure 8 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 8 At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. Of course, the electronic device may also include hardware required for other services.

[0163] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0164] The memory is used to store the program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.

[0165] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a road intersection scene generation device at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:

[0166] Divide and split the target scene to obtain the split scene;

[0167] Mapping is performed based on the split scenes to construct an abstract space;

[0168] The preset intelligent agent embedding strategy is called according to the abstract space, and the intersection scene under the abstract space is generated by combination, and the preset intelligent agent embedding strategy is effective in the intersection scene.

[0169] The above application Figure 6The method performed by the intersection scene generation device disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0170] The electronic device may also perform Figure 6 The method executed by the intersection scene generation device in the Figure 6 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.

[0171] The present application also provides a computer-readable storage medium, which stores one or more programs, wherein the one or more programs include instructions, which, when executed by an electronic device including multiple application programs, enable the electronic device to execute Figure 6 The method performed by the intersection scene generation device in the illustrated embodiment is specifically used to perform:

[0172] Divide and split the target scene to obtain the split scene;

[0173] Mapping is performed based on the split scenes to construct an abstract space;

[0174] The preset intelligent agent embedding strategy is called according to the abstract space, and the intersection scene under the abstract space is generated by combination, and the preset intelligent agent embedding strategy is effective in the intersection scene.

[0175] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0176] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0177] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0178] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0179] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0180] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0181] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0182] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0183] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0184] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for generating a road intersection scene, wherein: The method comprises: Divide and split the target scene to obtain the split scene; Mapping is performed based on the split scenes to construct an abstract space; The preset intelligent agent embedding strategy is called according to the abstract space, and the intersection scene under the abstract space is generated by combination, and the preset intelligent agent embedding strategy is effective in the intersection scene.

2. The method of claim 1, wherein: The step of dividing and splitting the target scene to obtain the split scenes includes: Dividing the target scene into a real intersection scene in the world space; Based on the driving conditions of vehicles in the real intersection scene, two types of basic intersection scenes are divided, including one-to-many crossing scenes and one-to-many merging scenes. The one-to-many crossing scenes refer to that the vehicle and other traffic participants will eventually be separated in the lateral space, and the one-to-many merging scenes refer to that the vehicle and other traffic participants will eventually merge in the lateral space.

3. The method of claim 2, wherein: The dividing and splitting of the target scene to obtain the split scenes also includes: Setting a one-to-many walking interaction strategy based on the one-to-many walking scenario; The ego vehicle executes a deceleration and parking strategy before reaching the conflict area, during which it checks whether there is a gap; If there is a gap, a passing strategy is executed, wherein the passing strategy includes giving way to vehicles on the right and simultaneously grabbing vehicles on the left; and / or, Setting a one-to-many import interaction strategy based on the one-to-many import scenario; The ego vehicle executes a deceleration and parking strategy before reaching the conflict area, during which it checks whether there is a gap; If there is a gap, a merging strategy is executed, which includes following the vehicle in front and simultaneously overtaking the vehicle behind.

4. The method of claim 1, wherein: The mapping is performed according to the split scenes to construct an abstract space, including: The trajectory of the ego vehicle is abstracted as a straight line and the interaction between the ego vehicle and other intelligent agents is abstracted as an interactive game or an intersection game; Mapping is performed based on the split scenes, and the constructed abstract space only contains the interactive scenes, which are used as the preset scenes for the first intersection; Mapping is performed based on the split scenes, and the constructed abstract space includes both interaction and intersection scenes, with intersection first and interaction later, but only one intersection is included and used as the preset scene for the second intersection; Mapping is performed based on the split scenes, and the constructed abstract space includes both interaction and intersection scenes, with interaction occurring before intersection, but only one intersection is included and is used as the preset scene for the third intersection; Mapping is performed based on the split scenes, and the constructed abstract space includes both interaction and intersection scenes, with interaction occurring before intersection, but including multiple intersections, and serving as the fourth intersection preset scene; The first intersection preset scene, the second intersection preset scene, the third intersection preset scene and the fourth intersection preset scene all include multiple game trajectories.

5. The method of claim 1, wherein: The abstract space only includes the interaction scene, the calling of the preset agent embedding strategy according to the abstract space, the combination generating the intersection scene under the abstract space, and the preset agent embedding strategy taking effect in the intersection scene include: According to the interactive scenario, the preset agent embedding strategy is configured as a car-following strategy, wherein the car-following strategy at least includes a car-following model; Based on the vehicle-following strategy, a corresponding traffic flow is generated.

6. The method of claim 1, wherein: The abstract space includes a scene with intersection in front and interaction in the back, and the preset agent embedding strategy is called according to the abstract space to combine and generate an intersection scene under the abstract space, and the preset agent embedding strategy is effective in the intersection scene, including: According to the scenario of the intersection first and the interaction later, the preset agent embedding strategy is configured as a car-following strategy and a one-to-many crossing strategy, and the car-following strategy at least includes a car-following model; Based on the vehicle following strategy and the one-to-many traveling strategy, a corresponding traffic flow is generated.

7. The method of claim 1, wherein: The abstract space includes a scene in which interaction comes first and intersection comes later and only includes one intersection. The preset agent embedding strategy is called according to the abstract space to combine and generate an intersection scene under the abstract space. The preset agent embedding strategy is effective in the intersection scene, including: According to the scenario in which the interaction occurs first and the intersection occurs later and the scenario only includes one intersection, the preset agent embedding strategy is configured as a car-following strategy, and the car-following strategy at least includes a car-following model; Based on the vehicle-following strategy, a corresponding traffic flow is generated.

8. The method of claim 1, wherein: The abstract space includes a scene in which interaction comes first and intersection comes later and includes two intersections. The preset agent embedding strategy is called according to the abstract space to combine and generate an intersection scene under the abstract space. The preset agent embedding strategy is effective in the intersection scene, including: According to the scenario in which the interaction comes first and the intersection comes later and includes two intersections, the preset agent embedding strategy is configured as a following strategy and a one-to-many merging strategy, and the following strategy at least includes a following model; A corresponding traffic flow is generated according to the vehicle following strategy and the one-to-many merging strategy.

9. The method of claim 1, wherein: The method further comprises: a step of controlling the data diversity of the intersection scene: In the process of calling the preset intelligent agent embedding strategy according to the abstract space and combining to generate the intersection scene in the abstract space, the traffic flow is obtained by controlling the driving along the preset path and randomly generating it.

10. A device for generating a road intersection scene, wherein: The device comprises: A scene division and splitting module is used to divide and split the target scene to obtain the split scene; A mapping module is used to map the split scenes and construct an abstract space; A generation module is used to call a preset intelligent agent embedding strategy according to the abstract space, and combine and generate an intersection scene under the abstract space, wherein the preset intelligent agent embedding strategy takes effect in the intersection scene.

11. An electronic device, comprising: processor; as well as A memory arranged to store computer executable instructions, which when executed cause the processor to perform the method of any one of claims 1 to 7.

12. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of application programs, causes the electronic device to execute any one of the methods of claims 1 to 7.

Images