Adversarial simulation scene generation method and device, electronic equipment and storage medium

Abstract

The present disclosure provides a method and device for generating an adversarial simulation scenario, electronic equipment and a storage medium, relating to the technical field of artificial intelligence, and in particular to the technical field of autonomous driving. The implementation scheme is: based on the initial state information of a first object and environment information, determining the after-driving behavior of the first object by using a first control model, wherein the environment information includes the state information of at least one second object; based on the initial state information, the environment information and the after-driving behavior, generating a candidate simulation scenario; determining the adversarial degree of the candidate simulation scenario based on the response behavior of the at least one second object to the after-driving behavior; and in response to the adversarial degree being greater than a threshold, determining the candidate simulation scenario as an adversarial simulation scenario.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of artificial intelligence, in particular to the technical field of automatic driving, and more particularly to a method and apparatus for generating an adversarial simulation scenario, an electronic device, a computer readable storage medium, and a computer program product. BACKGROUND

[0002] Automatic driving technology involves environment perception, behavior decision, trajectory planning, and motion control, etc. A vehicle with automatic driving function can automatically run without the operation of a driver or with only a small amount of operation of the driver, by relying on the collaborative work of sensors, visual computing systems, and positioning systems.

[0003] An algorithm model for realizing automatic driving is deployed in an automatic driving vehicle. Before the automatic driving vehicle is put into production or application, the algorithm model needs to be tested to verify its effectiveness or accuracy.

[0004] The methods described in this section can not necessarily be prior art methods. Unless otherwise indicated, it should not be assumed that any of the methods described in this section are considered prior art merely because of their inclusion in this section. Similarly, issues mentioned in this section should not be assumed to have been admitted to be prior art in any jurisdiction unless otherwise indicated. SUMMARY

[0005] The present disclosure provides a method and apparatus for generating an adversarial simulation scenario, an electronic device, a computer readable storage medium, and a computer program product.

[0006] According to an aspect of the present disclosure, a method for generating an adversarial simulation scenario is provided, including: determining a post-behavior of a first object based on initial state information of the first object and environment information, wherein the environment information includes respective state information of at least one second object; generating a candidate simulation scenario based on the initial state information, the environment information, and the post-behavior; determining an adversarial degree of the candidate simulation scenario based on response behaviors of the at least one second object to the post-behavior; and determining the candidate simulation scenario as an adversarial simulation scenario in response to the adversarial degree being greater than a threshold value.

[0007] According to an aspect of the present disclosure, there is provided a device for generating an adversarial simulation scenario, comprising: a first determining module configured to determine a post-action behavior of a first object based on initial state information of the first object and environment information, wherein the environment information comprises state information of at least one second object; a generating module configured to generate a candidate simulation scenario based on the initial state information, the environment information and the post-action behavior; a second determining module configured to determine an adversarial degree of the candidate simulation scenario based on response behaviors of the at least one second object to the post-action behavior; and a third determining module configured to determine the candidate simulation scenario as an adversarial simulation scenario in response to the adversarial degree being greater than a threshold value.

[0008] According to an aspect of the present disclosure, there is provided an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory storing instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above method.

[0009] According to an aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the above method.

[0010] According to an aspect of the present disclosure, there is provided a computer program product comprising computer program instructions for implementing the above method when executed by a processor.

[0011] According to one or more embodiments of the present disclosure, the generation efficiency and authenticity of an adversarial simulation scenario can be improved.

[0012] It should be understood that the contents described in this section are not intended to identify key or important features of the embodiments of the present disclosure, nor to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0013] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments and together with the description serve to explain exemplary implementations of the application. The illustrated embodiments are exemplary only and not limiting of the scope of the application. In all the drawings, like reference numerals refer to like parts throughout the several views.

[0014] Figure 1 A flowchart of a method for generating an adversarial simulation scenario according to some embodiments of the present disclosure is shown;

[0015] Figures 2A-2DA schematic diagram of a candidate simulation scenario is shown according to some embodiments of the present disclosure.

[0016] Figure 3 A schematic diagram of a generation process of an adversarial simulation scenario is shown according to some embodiments of the present disclosure.

[0017] Figure 4 A structural block diagram of a generation device of an adversarial simulation scenario is shown according to some embodiments of the present disclosure; and

[0018] Figure 5 A structural block diagram of an exemplary electronic device that can be used to implement embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0019] Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, which include various details of the embodiments of the present disclosure to assist in understanding them. These should be considered in their context only as illustrative. Thus, those of ordinary skill in the art will recognize various changes and modifications of the embodiments described herein, without departing from the scope of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

[0020] In the present disclosure, the terms "first", "second", and the like are used to describe various elements only and do not intend to limit the positional relationship, the time relationship, or the importance of the elements, and such terms are only used to distinguish one element from another. In some examples, the first element and the second element can refer to the same instance of the element, and in some cases, based on the context of the description, they can also refer to different instances.

[0021] The terms used in the description of various examples described in the present disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the number of elements is specifically limited, the element can be one or more than one, unless the context clearly indicates otherwise. In addition, the term "and / or" used in the present disclosure encompasses any one of the listed items and all possible combinations thereof.

[0022] In the technical solutions of the present disclosure, the acquisition, storage, and application of user personal information involved comply with relevant laws and regulations and do not violate public order and good customs.

[0023] Before an autonomous vehicle is put into production or application, it needs to be tested to verify whether the algorithm model deployed therein can enable the autonomous vehicle to run safely and smoothly.

[0024] To improve the efficiency and safety of testing, a large number of tests for autonomous vehicles are performed in a simulation environment. First, a simulation scene (i.e., a simulated traffic scene) is constructed, and then the autonomous vehicle is tested using the constructed simulation scene.

[0025] There are a large number of agents (such as motor vehicles, bicycles, pedestrians, etc.) in a real traffic scene. The autonomous vehicle itself is also an agent. An agent refers to an entity with autonomous behavior capability, which can respond to changes in the external environment and make appropriate behavior. When constructing a simulation scene, an agent needs to be constructed and the behavior of the agent needs to be simulated.

[0026] In most traffic scenes, the behavior of the agent is normal and safe. For example, the agent usually moves according to traffic rules and avoids colliding with other agents. In a small number of traffic scenes, the behavior of the agent is abnormal and dangerous. In this case, the agent can be in confrontation with other agents. For example, a motor vehicle agent (due to vehicle failure or driver error) can suddenly rush to other agents for a row, or collide with other agents. Hereinafter, the abnormal behavior of the agent is referred to as "confrontational behavior", the abnormal degree of the abnormal behavior of the agent is referred to as "confrontational degree", and the traffic scene in which the agents are in confrontation due to the abnormal behavior of the agent is referred to as "confrontational scene".

[0027] To ensure the driving safety of the autonomous vehicle in the confrontational scene, a confrontational simulation scene needs to be constructed and the autonomous vehicle needs to be tested. In related technologies, the confrontational simulation scene is usually generated by manually setting the confrontational behavior of the agent. This scheme is low in efficiency, and the confrontational behavior set by the human is fixed and mechanical, which cannot simulate the interaction of different agents in a real scene, resulting in low authenticity, rationality and application value of the generated confrontational simulation scene.

[0028] To solve the above problems, embodiments of the present disclosure provide a method for generating a confrontational simulation scene, which can improve the generation efficiency and authenticity of the confrontational simulation scene.

[0029] Embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0030] Figure 1 A flowchart of a method 100 for generating a confrontational simulation scene according to an embodiment of the present disclosure is shown. As shown in Figure 1 The method 100 includes steps S110-S140.

[0031] In step S110, a post-behavior of the first object is determined based on initial state information of the first object and environment information, by using a first control model. The environment information includes respective state information of the at least one second object.

[0032] In step S120, a candidate simulation scenario is generated based on the initial state information, the environment information and the post-behavior.

[0033] In step S130, a confrontation degree of the candidate simulation scenario is determined based on response behaviors of the at least one second object to the post-behavior.

[0034] In step S140, in response to the confrontation degree being greater than a threshold, the candidate simulation scenario is determined as a confrontation simulation scenario.

[0035] According to an embodiment of the present disclosure, the post-behavior of the first object is determined by using the first control model, and the state information of the second object is considered when determining the post-behavior, which improves the intelligence of the first object, so that the first object can interact with the second object, thereby improving the authenticity and rationality of the candidate simulation scenario.

[0036] The confrontation simulation scenario is filtered from the candidate simulation scenario based on the response behaviors of the second object to the post-behavior, which realizes the automatic generation of the confrontation simulation scenario and improves the generation efficiency and authenticity.

[0037] The following describes each step of the method 100 in detail.

[0038] In step S110, a post-behavior of the first object is determined based on initial state information of the first object and environment information, by using a first control model.

[0039] The first object and the second object may, for example, be motor vehicles, bicycles, pedestrians, etc. The second object is an object different from the first object. It can be understood that the first object and the second object are both agents in a traffic environment.

[0040] For the first object, each second object is an obstacle thereof. For any second object, the first object and other second objects are obstacles thereof.

[0041] According to some embodiments, one of the at least one second object is an automatic driving vehicle to be tested. Accordingly, the first object is an obstacle of the automatic driving vehicle. According to an embodiment of the present disclosure, by simulating the behavior of the first object, a confrontation simulation scenario for testing the automatic driving vehicle can be generated.

[0042] According to some embodiments, an interface for configuring an obstacle in a simulation scenario can be set in a simulation test platform of an autonomous vehicle. The first object can be an obstacle for testing the autonomous vehicle selected or newly created by a user through the interface. Further, the user can configure initial state information and environment information of the first object through the interface.

[0043] The initial state information indicates an initial state of the first object. According to some embodiments, the initial state information of the first object can include at least one of an initial position, an initial velocity, or an initial orientation. In other embodiments, the initial state information can further include information such as acceleration, aggressiveness (conservative, moderate, aggressive), size, type (motor vehicle, bicycle, pedestrian, etc.). By flexibly setting different initial state information, different post-behaviors of the first object can be generated, thereby generating different candidate simulation scenarios and improving the diversity and coverage of the candidate simulation scenarios.

[0044] According to some embodiments, the content included in the initial state information can be determined according to an atomic behavior of the first object. The atomic behavior refers to a fixed behavior that does not need to interact with other agents, such as driving along the road, following the track, changing lanes, following the target, and keeping still. Different initial state information can be set for different atomic behaviors. For example, for the "driving along the road" behavior, the initial state information can include an initial position (the starting point of driving), an initial velocity, and an initial orientation. For the "keeping still" behavior, the initial state information can only include an initial position.

[0045] In embodiments of the present disclosure, the environment information includes state information of each of at least one second object. Similar to the initial state information of the first object, the state information of the second object can include information such as position, velocity, orientation, acceleration, aggressiveness, size, and type. According to some embodiments, the environment information can further include at least one of traffic light information and road network information. The road network information indicates the position of the road. The road network information can also be referred to as map information or road network topology information. By flexibly setting different environment information, different post-behaviors of the first object can be generated, thereby generating different candidate simulation scenarios and improving the diversity and coverage of the candidate simulation scenarios.

[0046] In embodiments of the present disclosure, the first control model is used to control the behavior of the first object. According to some embodiments, the first control model can be determined based on the behavior type of the first object. Accordingly, the method 100 can further include: obtaining the behavior type of the first object; and determining the first control model for controlling the first object based on the behavior type.

[0047] In embodiments of the present disclosure, the behavior type of the first object refers to the type of interaction behavior of the first object with other intelligent agents (i.e., the second object). The behavior type includes, for example, cutting, following, merging, straight-ahead meeting left turn (i.e., the first object straight-ahead meets the second object left turn), left turn meeting straight-ahead (i.e., the first object left turn meets the second object straight-ahead), left / right turn cutting small bend (i.e., the first object turns with a smaller radius), etc. According to some embodiments, the behavior type of the first object configured by the user can be obtained through a preset interface.

[0048] Based on the behavior type of the first object, a first control model for controlling the first object can be determined.

[0049] The first control model may, for example, be a neural network model, such as an LSTM (Long Short-Term Memory), a Transformer, a CNN (Convolutional neural network), etc.

[0050] According to some embodiments, the control model corresponding to each behavior type can be preset to obtain a correspondence between the behavior type and the control model. It should be noted that the behavior type and the control model can have a one-to-one, one-to-many, or many-to-many relationship. One behavior type can correspond to one or more control models, and one control model can correspond to one or more behavior types. For example, each behavior type can correspond to a control model that is only applicable to the behavior type and a general control model that is applicable to all behavior types.

[0051] Accordingly, the first control model for controlling the first object can be determined based on the preset correspondence between the behavior type and the control model. In the case where the behavior type corresponds to multiple control models, the first control model can be any one of the multiple control models.

[0052] According to some embodiments, step S110 can include inputting the initial state information and the environment information into the first control model to obtain target state information of the first object output by the first control model, and determining the post-behavior based on the target state information. According to this embodiment, the post-behavior of the first object is determined based on the target state information output by the first control model, and by changing the content of the target state information output by the model, flexible behavior control of the first object can be achieved. The target state information may, for example, include a target speed, a target position, a target acceleration, etc.

[0053] According to some embodiments, the target state information can include a target speed. Accordingly, determining the post-behavior based on the target state information can include obtaining a reference trajectory of the first object, and determining the post-behavior as moving along the reference trajectory at the target speed.

[0054] According to the above embodiment, by modifying the initial state information, the first object can be made to travel at different speeds along the reference trajectory, thereby changing the order in which the first object and the second object reach the interaction point (i.e., the intersection of the trajectories of the two), making different post-behaviors, and realizing the generalization of the candidate simulation scenario.

[0055] Based on the initial state information of the first object, the environment information and the post-behavior, the candidate simulation scenario can be generated.

[0056] Figure 2A 、 2B The schematic diagrams of the candidate simulation scenarios S1 and S2 generated based on different initial state information of the first object 211 are respectively shown.

[0057] In Figure 2A 、 2B , the behavior type of the first object 211 is left turn encountering straight running, that is, the first object 211 turns left while the second object 212 opposite to the first object 211 runs straight. Figure 2A 、 2B The dashed line in

[0058] Based on the initial state information info1, the target speed of the first object 211 at subsequent time points can be determined by using the first control model, so as to control the first object 211 to travel along the reference trajectory at the corresponding target speed, and generate the candidate simulation scenario S1. As shown in Figure 2A , in the candidate simulation scenario S1, the first object 211 reaches the interaction point A (the intersection of the trajectories of the two) later than the second object 212, that is, the first object 211 makes the post-behavior of yielding to the second object 212.

[0059] By modifying the initial state information info1, the initial state information info2 is obtained. Based on the initial state information info2, the target speed of the first object 211 at subsequent time points can be determined by using the first control model, so as to control the first object to travel along the reference trajectory at the corresponding target speed, and generate the candidate simulation scenario S2. As shown in Figure 2B , in the candidate simulation scenario S2, the first object 211 reaches the interaction point A (the intersection of the trajectories of the two) earlier than the second object 212, that is, the first object 211 makes the post-behavior of cutting in ahead of the second object 212.

[0060] Based on the traffic rules, in the scenario of left turn encountering straight running, the straight running vehicle has the right of way, and the left turning vehicle should yield to the straight running vehicle. As shown in Figure 2B the candidate simulation scenario S2, the first object 211 makes an abnormal behavior (cutting in ahead) that does not conform to the traffic rules, so the candidate simulation scenario belongs to the adversarial simulation scenario.

[0061] According to some embodiments, the target state information can include a target position and a target speed. Accordingly, determining the post-behavior based on the target state information includes determining the post-behavior as moving to the target position at the target speed.

[0062] According to the above embodiments, by modifying the initial state information, the position and speed of the first object can be changed, thereby changing the post-behavior of the first object, making the first object produce different interactions with the second object, and realizing the generalization of the candidate simulation scene.

[0063] Figure 2C 、 2D The schematic diagrams of the candidate simulation scenes S3 and S4 generated based on different initial state information of the first object 221 are respectively shown.

[0064] In Figure 2C 、 2D , the behavior type of the first object 221 is left turn encountering straight running, that is, the first object 221 turns left while the second object 223 in the opposite direction runs straight. Moreover, there is a second object 222 turning left in front of the first object 221.

[0065] Based on the initial state information info3, the target position and target speed of the first object 221 at subsequent time points can be determined by using the first control model, so as to control the first object 221 to move to the corresponding target position at the corresponding target speed in different time periods, forming the driving trajectory of the first object 221 (as shown by the dashed line in Figure 2C , and generating the candidate simulation scene S3. As shown in Figure 2C , in the candidate simulation scene S3, the first object 221 makes the post-behavior of following the second object 222 in front to turn left, and the behavior mode is relatively conservative.

[0066] By modifying the initial state information info3, the initial state information info4 is obtained. Based on the initial state information info4, the target position and target speed of the first object 221 at subsequent time points can be determined by using the first control model, so as to control the first object 221 to move to the corresponding target position at the corresponding target speed in different time periods, forming the driving trajectory of the first object 221 (as shown by the dashed line in Figure 2D , and generating the candidate simulation scene S4. As shown in Figure 2D , in the candidate simulation scene S4, the first object 221 makes the overtaking behavior of turning left to cut a small bend, and completes the left turn ahead of the second object 222, and the behavior mode is relatively aggressive.

[0067] In order to ensure traffic safety, vehicles are usually required to run in order at intersections. In Figure 2DThe candidate simulation scenario S4 shown in the figure is that the first object 221 makes an unsafe abnormal behavior (overtaking), and therefore the candidate simulation scenario belongs to the adversarial simulation scenario.

[0068] According to some embodiments, the generating the candidate simulation scenario based on the initial state information, the environment information and the post-behavior (step S120) comprises: determining whether the first object is located in a road when performing the post-behavior based on road network information. In response to the first object being located in the road when performing the post-behavior, the candidate simulation scenario is generated based on the initial state information, the environment information and the post-behavior.

[0069] It can be understood that in a real traffic scene, an object (agent) usually travels in a road. However, in a simulation scenario, the post-behavior obtained by simulation may be located in a non-road environment such as a lawn or a building area. The post-behavior located in the non-road environment is not credible and usually does not have an impact on the behaviors of other objects in the road. Therefore, according to the above-mentioned embodiments, if the first object is located in the road, the post-behavior is relatively high in authenticity, and the candidate simulation scenario can be generated accordingly. If the first object is not located in the road, the post-behavior is relatively low in authenticity, and the first object not in the road is difficult to affect the behaviors of other objects, and therefore the post-behavior is discarded and the candidate simulation scenario is not generated. In this way, the authenticity and rationality of the candidate simulation scenario can be improved.

[0070] In step S130, the adversarial degree of the candidate simulation scenario is determined based on the response behavior of the at least one second object to the post-behavior. The adversarial degree is used to represent the abnormal degree of the behaviors of the objects in the candidate simulation scenario.

[0071] According to some embodiments, each of the at least one second object corresponds to a second control model, and the response behavior of the second object to the post-behavior is determined by using the second control model.

[0072] According to the above-mentioned embodiments, the behaviors of the second objects are controlled by the second control models, which improves the intelligence of the second objects and makes the interaction between the first object and the second objects more authentic and reasonable.

[0073] According to some embodiments, one of the at least one second object is an autonomous vehicle. Since the adversarial simulation scenario is used to test the autonomous vehicle, the behavior of the autonomous vehicle is particularly important for determining the adversarial degree of the simulation scenario. Accordingly, in step S130, the adversarial degree of the candidate simulation scenario can be determined based on the response behavior of the autonomous vehicle to the post-behavior.

[0074] According to the above embodiment, the adversarial degree of the candidate simulation scene is determined based on the response behavior of the autonomous vehicle, which can improve the accuracy of the determined simulation degree. Further, the adversarial degree of the candidate simulation scene can be determined based on and only based on the response behavior of the autonomous vehicle, thereby improving the computational efficiency and thus improving the efficiency of scene screening.

[0075] According to some embodiments, the adversarial degree of the candidate simulation scene can be determined based on at least one of: a minimum distance between any of the at least one second object and the first object; or a speed of the first object when the first object is closest to any of the second object. Wherein the adversarial degree is negatively correlated with the above minimum distance and positively correlated with the above speed. According to the above embodiment, the closer the distance between the first object and the second object, the greater the speed, the higher the abnormality of the behavior of the first object, and the greater the challenge to the second object, so the corresponding candidate simulation scene is determined as an adversarial simulation scene. Thereby the authenticity and rationality of the adversarial scene can be improved.

[0076] According to some embodiments, in the case where there are multiple second objects, for each second object, the minimum distance between the second object and the first object and / or the speed of the first object when the first object is closest to the second object can be determined, and the adversarial degree corresponding to the second object is determined based on the determined minimum distance and / or speed. Subsequently, the average of the adversarial degrees of the multiple second objects can be calculated, and the average is determined as the adversarial degree of the candidate simulation scene.

[0077] According to some embodiments, in the case where at least one of the at least one second object is an autonomous vehicle, the minimum distance between the autonomous vehicle and the first object and / or the speed of the first object when the first object is closest to the autonomous vehicle can be determined. Based on the determined minimum distance and / or speed, the adversarial degree of the candidate simulation scene can be determined. According to this embodiment, the closer the distance between the first object and the autonomous vehicle, the greater the speed, the higher the abnormality of the behavior of the first object, and the greater the challenge to the autonomous vehicle, so the corresponding candidate simulation scene is determined as an adversarial simulation scene. Thereby the authenticity and rationality of the adversarial scene can be improved.

[0078] According to some embodiments, the at least one second object further includes a third object different from the above autonomous vehicle. Accordingly, the adversarial degree of the candidate simulation scene can be determined based on the response behavior of the autonomous vehicle to the rear behavior of the first object and the behavior of the third object.

[0079] It should be noted that in the case where at least one of the at least one second object is an autonomous vehicle, the objects in the at least one second object other than the autonomous vehicle are third objects. It should be understood that the third object itself can also be an autonomous vehicle.

[0080] It should be understood that the rearward behavior of the first object will not only affect the behavior of the autonomous vehicle, but also affect the behavior of the third object. The behavior of the first object can affect the behavior of the third object, causing the third object to be in confrontation with the autonomous vehicle. According to the above-mentioned embodiments, the degree of confrontation of the scene is determined based on the response behavior of the autonomous vehicle to the first object and the third object, which can achieve accurate evaluation of the degree of confrontation.

[0081] According to some embodiments, an evaluation index (such as the number of collisions, the collision rate, etc.) for evaluating the degree of confrontation can be calculated based on the response behavior of the at least one second object to the rearward behavior of the first object, and the degree of confrontation of the candidate simulation scene is determined based on the value of the evaluation index.

[0082] In step S140, the candidate simulation scene is determined as a confrontation simulation scene in response to the degree of confrontation being greater than the threshold value.

[0083] As described above, the confrontation simulation scene can be used to test the autonomous vehicle. Accordingly, the method 100 can further include simulating the test of the autonomous vehicle using the confrontation simulation scene. Thereby, the response capability of the autonomous vehicle to the confrontation scene can be evaluated, so that the autonomous vehicle can be improved and optimized accordingly.

[0084] According to some embodiments, the autonomous vehicle to be tested can be one of the at least one second object described above. According to other embodiments, the autonomous vehicle to be tested can also be an autonomous vehicle different from the at least one second object described above.

[0085] Figure 3 A schematic diagram of a process of generating a confrontation simulation scene according to some embodiments of the present disclosure is shown. As Figure 3 shown, the process includes steps S371-S375.

[0086] In step S371, a following behavior is selected as an atomic behavior of the first object from a set of preset atomic behaviors 310. The behavior parameters (such as initial position, initial speed, initial direction, initial acceleration, etc.) of the following behavior are configured, and the configured behavior parameters are taken as initial state information 320 of the first object.

[0087] In step S372, the initial state information 320 is generalized to generate a set of initial state information 330. As Figure 3 shown, the set of initial state information includes initial state information 331, 332, 333, 334, 335, etc.

[0088] According to some embodiments, the value range of each item of information can be determined based on the initial state information 320, and a random value is taken within the value range to generate new initial state information. For example, the initial state information 320 includes initial position coordinates (x0, y0) and initial velocity v0. The value range of the initial x coordinate can be determined as [x0-△x, x0+△x], the value range of the initial y coordinate can be determined as [y0-△y, y0+△y], and the value range of the initial velocity can be determined as [v0-△v, v0+△v]. By taking a random value within the above value range, new initial state information 331-335 can be generated to realize the generalization of the initial state information.

[0089] It should be understood that although Figure 3 is not shown in FIG. 3, in steps S371 and S372, the environment information of the first object can also be configured and generalized.

[0090] In step S373, for each initial state information 331, 332, 333, 334, 335, … in the initial state information set 330, an intelligent control model for processing the initial state information can be obtained from the control model set 340. As Figure 3 shown, the control model set 340 includes an intelligent control model 341 for a specific behavior type and a general intelligent control model 342 applicable to all behavior types.

[0091] In step S374, the intelligent control model obtained in step S373 is used to determine the post-behavior of the first object, and simulation scenarios 351, 352, 353, 354, 355, … corresponding to the initial state information 331, 332, 333, 334, 335, … can be generated. These simulation scenarios form a candidate simulation set 350.

[0092] In step S375, the confrontation degree of each simulation scenario in the candidate simulation scenario set 350 is calculated respectively, and the simulation scenarios with a confrontation degree higher than a threshold value are taken as confrontation simulation scenarios to obtain a confrontation simulation scenario set 360. As Figure 3 shown, the simulation scenarios 351, 354, and 355 are confrontation simulation scenarios.

[0093] According to embodiments of the present disclosure, a generation apparatus of confrontation simulation scenarios is also provided. Figure 4 A structural block diagram of the generation apparatus 400 of confrontation simulation scenarios according to embodiments of the present disclosure is shown. As Figure 4 shown, the apparatus 400 includes a first determination module 410, a generation module 420, a second determination module 430, and a third determination module 440.

[0094] The first determining module 410 is configured to determine, based on initial state information of the first object and environment information, a post-behavior of the first object by using a first control model. The environment information includes respective state information of at least one second object.

[0095] The generating module 420 is configured to generate a candidate simulation scenario based on the initial state information, the environment information, and the post-behavior.

[0096] The second determining module 430 is configured to determine, based on a response behavior of the at least one second object to the post-behavior, an adversarial degree of the candidate simulation scenario.

[0097] The third determining module 440 is configured to determine, in response to the adversarial degree being greater than a threshold, the candidate simulation scenario as an adversarial simulation scenario.

[0098] According to an embodiment of the present disclosure, the first control model is used to determine the post-behavior of the first object, and the state information of the second object is considered when determining the post-behavior, which improves the intelligence of the first object and enables the first object to interact with the second object, thereby improving the authenticity and rationality of the candidate simulation scenario.

[0099] The adversarial simulation scenario is screened from the candidate simulation scenario based on the response behavior of the second object to the post-behavior, which realizes the automatic generation of the adversarial simulation scenario and improves the generation efficiency and authenticity.

[0100] According to some embodiments, the first determining module 410 includes: an output unit configured to input the initial state information and the environment information into the first control model to obtain target state information of the first object output by the first control model; and a determining unit configured to determine the post-behavior based on the target state information.

[0101] According to some embodiments, the target state information includes a target speed, and the determining unit includes: an obtaining sub-unit configured to obtain a reference trajectory of the first object; and a first determining sub-unit configured to determine the post-behavior as moving along the reference trajectory at the target speed.

[0102] According to some embodiments, the target state information includes a target position and a target speed, and the determining unit includes: a second determining sub-unit configured to determine the post-behavior as moving to the target position at the target speed.

[0103] According to some embodiments, the generating module 420 comprises: a judging unit configured to judge whether the first object is located in a road when performing the rear behavior based on road network information; and a generating unit configured to generate a candidate simulation scenario based on the initial state information, the environment information and the rear behavior in response to the first object being located in the road when performing the rear behavior.

[0104] According to some embodiments, each of the at least one second object corresponds to a second control model, and the response behavior of the second object to the rear behavior is determined by using the second control model.

[0105] According to some embodiments, the at least one second object comprises an autonomous vehicle, and the second determining module 430 is further configured to determine the confrontation degree of the candidate simulation scenario based on the response behavior of the autonomous vehicle to the rear behavior.

[0106] According to some embodiments, the second determining module 430 is further configured to determine the confrontation degree of the candidate simulation scenario based on at least one of: a minimum distance between any second object in the at least one second object and the first object; or a speed of the first object when the first object is closest to the second object, wherein the confrontation degree is negatively correlated with the minimum distance and positively correlated with the speed.

[0107] According to some embodiments, the at least one second object further comprises a third object different from the autonomous vehicle, and the second determining module 430 is further configured to determine the confrontation degree of the candidate simulation scenario based on the response behavior of the autonomous vehicle to the rear behavior and the behavior of the third object.

[0108] According to some embodiments, the apparatus 400 further comprises a testing module configured to perform simulation testing on an autonomous vehicle by using the confrontation simulation scenario.

[0109] According to some embodiments, the initial state information comprises at least one of: an initial position, an initial speed or an initial orientation.

[0110] According to some embodiments, the environment information further comprises at least one of: traffic signal information or road network information.

[0111] According to some embodiments, the apparatus 400 further comprises: an obtaining module configured to obtain a behavior type of the first object; and a fourth determining module configured to determine the first control model for controlling the first object based on the behavior type.

[0112] It should be understood that, Figure 4The various modules or units of the apparatus 400 shown in FIG. 4 can correspond to the various modules or units described above with reference to the method 100. Figure 1 The various steps in the method 100 described above correspond to the various modules or units of the apparatus 400. Thus, the operations, features and advantages described above for the method 100 apply equally to the apparatus 400 and its included modules and units. For the sake of brevity, certain operations, features and advantages are not described again here.

[0113] Although specific functions are discussed above with reference to particular modules, it should be noted that the functions of the various modules discussed herein can be split among multiple modules and / or at least some functions of multiple modules can be combined into a single module. For example, the second determining module 430 and the third determining module 440 described above can be combined into a single module in some embodiments.

[0114] It should also be understood that various techniques described herein can be described in the general context of software hardware elements or program modules. The various software -based aspects of the technology described above can be implemented with software directions that are executed using one or more processing units. The software instructions can be stored in one or more computer-readable memory devices, such as the computer-readable storage medium 430 described above. Figure 4 The various modules described above can be implemented in hardware or in hardware combined with software and / or firmware. For example, the modules can be implemented as computer program code / instructions configured to be executed in one or more processors and stored in a computer-readable storage medium. Alternatively, the modules can be implemented as hardware logic / circuitry. For example, in some embodiments, one or more of the modules 410-440 can be implemented together in a System on Chip (SoC). The SoC can include an integrated circuit chip (which includes one or more of a processor (e.g., a Central Processing Unit (CPU), a microcontroller, a microprocessor, a Digital Signal Processor (DSP), etc.), a memory, one or more communication interfaces, and / or other circuitry), and can optionally execute received program code and / or include embedded firmware to perform functions.

[0115] According to an embodiment of the disclosure, an electronic device is also provided, comprising: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the above-mentioned method for generating an adversarial simulation scenario.

[0116] According to an embodiment of the disclosure, a non-transitory computer-readable storage medium storing computer instructions is also provided, and the computer instructions are used to enable the computer to perform the above-mentioned method for generating an adversarial simulation scenario.

[0117] According to an embodiment of the present disclosure, a computer program product is also provided, comprising computer program instructions which, when executed by a processor, implement the above-mentioned method for generating an adversarial simulation scenario.

[0118] Reference Figure 5 A block diagram of an electronic device 500, which can be used as the server or the client of the present disclosure, will now be described, which is an example of a hardware device that can be applied to aspects of the present disclosure. The electronic device is intended to represent a wide variety of digital electronic computing devices, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computing devices. The electronic device can also represent a variety of mobile devices, such as personal digital assistants, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the present disclosure described and / or claimed in this document.

[0119] As Figure 5 shown, the electronic device 500 includes a computing unit 501 that can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded into a random access memory (RAM) 503 from a storage unit 508. In the RAM 503, various programs and data required for the operation of the electronic device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0120] A plurality of components in the electronic device 500 are connected to the I / O interface 505, including: an input unit 506, an output unit 507, a storage unit 508, and a communication unit 509. The input unit 506 can be any type of device that can input information to the electronic device 500, and can receive inputted digital or character information, and generate key signal inputs related to user settings and / or function controls of the electronic device, and can include, but is not limited to, a mouse, a keyboard, a touch screen, a track pad, a track ball, a joystick, a microphone, and / or a remote controller. The output unit 507 can be any type of device that can present information, and can include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 508 can include, but is not limited to, a magnetic disk, an optical disk. The communication unit 509 allows the electronic device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks, and can include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth device, an 802.11 device, a Wi-Fi device, a WiMAX device, a cellular communication device, and / or the like.

[0121] The computing unit 501 can be various general and / or special purpose processing components having processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 501 performs various methods and processes described above, such as the method 100. For example, in some embodiments, the method 100 can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded onto the RAM 503 and executed by the computing unit 501, one or more steps of the method 100 described above can be performed. Alternatively, in other embodiments, the computing unit 501 can be configured to perform the method 100 by any other appropriate means, such as by means of firmware.

[0122] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a complex programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0123] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, or entirely on a remote machine or server.

[0124] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0125] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0126] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0127] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0128] It should be understood that the various forms of flow shown above can be re-ordered, added to, or deleted from without departing from the spirit of the present disclosure. For example, the steps recited in the present disclosure can be performed in parallel, in series, or in a different order, as long as the desired results of the technology disclosed in the present disclosure are achieved, which is not limited herein.

[0129] While embodiments or examples of this disclosure have been described with reference to the figures, it will be understood that the methods, systems, and devices described above are merely exemplary embodiments or examples, and the scope of this disclosure is not limited to these embodiments or examples, but only by the claims and their equivalents. Various elements of the embodiments or examples can be omitted or substituted by equivalents thereof. Furthermore, the steps can be performed in a different order than described in the disclosure. Further, various elements of the embodiments or examples can be combined in various ways. It is important that as technology evolves, many of the elements described herein can be substituted by equivalents which become known after this disclosure.

Claims

1. A method for generating an adversarial simulation scenario, comprising: determining, based on initial state information of a first object and environment information, a post-behavior of the first object by using a first control model, wherein the environment information comprises respective state information of at least one second object, one of the at least one second object being an autonomous vehicle to be tested, and the first object being an obstacle of the autonomous vehicle; generating a candidate simulation scenario based on the initial state information, the environment information, and the post-behavior; determining an adversarial degree of the candidate simulation scenario based on response behaviors of the at least one second object to the post-behavior, comprising: determining the adversarial degree based on at least one of a minimum distance between the autonomous vehicle and the first object or a speed of the first object when the first object is closest to the autonomous vehicle, wherein the adversarial degree is negatively correlated with the minimum distance and positively correlated with the speed; and in response to the adversarial degree being greater than a threshold, determining the candidate simulation scenario as an adversarial simulation scenario.

2. The method of claim 1, wherein, The determining, based on initial state information of a first object and environment information, a post-behavior of the first object by using a first control model comprises: inputting the initial state information and the environment information into the first control model to obtain target state information of the first object output by the first control model; and determining the post-behavior based on the target state information.

3. The method of claim 2, wherein, The target state information comprises a target speed, and the determining the post-behavior based on the target state information comprises: obtaining a reference trajectory of the first object; and determining the post-behavior as moving along the reference trajectory at the target speed.

4. The method of claim 2, wherein, The target state information comprises a target position and a target speed, and the determining the post-behavior based on the target state information comprises: determining the post-behavior as moving to the target position at the target speed.

5. The method of claim 1, wherein, The generating a candidate simulation scenario based on the initial state information, the environment information, and the post-behavior comprises: judging whether the first object is located in a road when performing the post-behavior based on road network information; and in response to the first object being located in the road when performing the post-behavior, generating the candidate simulation scenario based on the initial state information, the environment information, and the post-behavior.

6. The method of claim 1, wherein, Each of the at least one second object corresponds to a second control model, and the response behavior of the second object to the post-behavior is determined by using the second control model.

7. The method of claim 1, wherein, The at least one second object further comprises a third object different from the autonomous vehicle, and the determining the adversarial degree of the candidate simulation scenario based on the response behavior of the autonomous vehicle to the post-behavior comprises: determining the adversarial degree of the candidate simulation scenario based on a response behavior of the autonomous vehicle to the post-behavior and a behavior of the third object.

8. The method of any one of claims 1-7, further comprising: performing simulation testing on the autonomous vehicle by using the adversarial simulation scenario.

9. The method of any one of claims 1-7, wherein, The initial state information comprises at least one of an initial position, an initial speed, or an initial orientation.

10. The method of any one of claims 1-7, wherein, The environment information further comprises at least one of traffic signal information or road network information.

11. The method of any one of claims 1-7, further comprising: obtaining a behavior type of the first object; and determining the first control model for controlling the first object based on the behavior type.

12. An apparatus for generating an adversarial simulation scenario, comprising: a first determining module configured to determine, based on initial state information of a first object and environment information, a post-behavior of the first object by using a first control model, wherein the environment information comprises respective state information of at least one second object, one of the at least one second object is an autonomous vehicle to be tested, and the first object is an obstacle of the autonomous vehicle; a generating module configured to generate a candidate simulation scenario based on the initial state information, the environment information, and the post-behavior; a second determining module configured to determine an adversarial degree of the candidate simulation scenario based on response behaviors of the at least one second object to the post-behavior, the second determining module being further configured to: determine the adversarial degree based on at least one of a minimum distance between the autonomous vehicle and the first object or a speed of the first object when the first object is closest to the autonomous vehicle, wherein the adversarial degree is negatively correlated with the minimum distance and positively correlated with the speed; and a third determining module configured to determine the candidate simulation scenario as an adversarial simulation scenario in response to the adversarial degree being greater than a threshold. The first determining module comprises:

13. The apparatus of claim 12, wherein, an output unit configured to input the initial state information and the environment information into the first control model to obtain target state information of the first object output by the first control model; and a determining unit configured to determine the post-behavior based on the target state information. The target state information comprises a target speed, and the determining unit comprises:

14. The apparatus of claim 13, wherein, an obtaining sub-unit configured to obtain a reference trajectory of the first object; and a first determining sub-unit configured to determine the post-behavior as moving along the reference trajectory at the target speed. The target state information comprises a target position and a target speed, and the determining unit comprises:

15. The apparatus of claim 13, wherein, a second determining sub-unit configured to determine the post-behavior as moving to the target position at the target speed. The generating module comprises:

16. The apparatus of claim 12, wherein, a judging unit configured to judge, based on road network information, whether the first object is located in a road when performing the post-behavior; and a generating unit configured to generate, in response to the first object being located in the road when performing the post-behavior, a candidate simulation scenario based on the initial state information, the environment information, and the post-behavior. Each of the at least one second object corresponds to a second control model, and the response behavior of the second object to the post-behavior is determined by using the second control model.

17. The apparatus of claim 12, wherein, ​ 18. The apparatus of claim 12, wherein, The at least one second object further includes a third object different from the autonomous vehicle, and the second determining module is further configured to: determine, based on a response behavior of the autonomous vehicle to the rearward behavior and a behavior of the third object, a confrontation degree of the candidate simulation scenario.

19. The apparatus of any one of claims 12-18, further comprising: a testing module configured to perform simulation testing on an autonomous vehicle using the adversarial simulation scenario.

20. The apparatus of any one of claims 12-18, wherein, The initial state information includes at least one of an initial position, an initial speed, or an initial orientation.

21. The apparatus of any one of claims 12-18, wherein, The environment information further includes at least one of traffic signal information or road network information.

22. The apparatus of any one of claims 12-18, further comprising: an obtaining module configured to obtain a behavior type of the first object; and a fourth determining module configured to determine, based on the behavior type, the first control model for controlling the first object.

23. An electronic device, comprising: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-11. The computer instructions are used to enable a computer to perform the method of any one of claims 1-11.

24. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer program instructions, when executed by a processor, implement the method of any one of claims 1-11.

25. A computer program product comprising computer program instructions, wherein, ​

Images