Simulation test method and device for automatic driving algorithm, computer equipment and medium

Abstract

The invention provides a simulation test method and device for an automatic driving algorithm, computer equipment and a medium. The simulation test method comprises the following steps: constructing a simulation test scene; inputting the virtual sensor data into an automatic driving algorithm to obtain a virtual control instruction; when a driving traffic participant corresponding to the driving vehicle is associated with a preset behavior model, simulating a first dynamic response of the driving vehicle to the virtual control instruction in the simulation test scene; determining first simulation test data based on the first dynamic response and a corresponding preset reference response; when the driving traffic participant corresponding to the driving vehicle is associated with a non-preset behavior model, simulating a second dynamic response of the driving vehicle to the virtual control instruction in the simulation test scene; determining second simulation test data based on the second dynamic response and a corresponding preset reference response; and determining target simulation test data based on the first simulation test data and the second simulation test data. Therefore, the simulation test efficiency of the automatic driving algorithm is effectively improved.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the technical field of simulation testing, and more particularly, to a simulation testing method, apparatus, computer device, and medium applicable to an autonomous driving algorithm. Background Art

[0002] With the rapid development of artificial intelligence and computer vision, the development of autonomous driving systems has undergone a transformation from traditional modular methods to end-to-end deep learning methods. To ensure the stability and safety of end-to-end autonomous driving algorithms, large-scale simulation testing and verification of autonomous driving algorithms are required.

[0003] In related technologies, the simulation testing of autonomous driving algorithms is usually implemented by a data-driven simulation method. For example, a large amount of driving data in the real world is collected in advance, and the massive data is manually preprocessed, such as cleaning, screening, labeling, and removing noise. Finally, a data test model is constructed by training and testing the processed data to perform simulation testing on the autonomous driving algorithm.

[0004] However, with the existing method, the algorithm testing efficiency is low. Summary of the Invention

[0005] Embodiments described herein provide a simulation testing method, apparatus, computer device, and medium for an autonomous driving algorithm, which overcome the above problems.

[0006] In a first aspect, according to the content of the present disclosure, there is provided a simulation testing method for an autonomous driving algorithm, including:

[0007] Constructing a simulation testing scenario, the simulation testing scenario being constructed based on a three-dimensional model of a virtual scenario target, the virtual scenario target including: dynamic traffic participants, static traffic participants, and environmental participation factors;

[0008] Obtaining virtual sensor data preset for a driving vehicle in the simulation testing scenario; and inputting the virtual sensor data into the autonomous driving algorithm to obtain a virtual control command output by the autonomous driving algorithm for the driving vehicle;

[0009] When associating a preset behavior model with a driving traffic participant corresponding to the driving vehicle, simulating a first dynamic response of the driving vehicle to the virtual control command in the simulation testing scenario; and determining first simulation testing data of the autonomous driving algorithm based on the first dynamic response and a corresponding preset reference response;

[0010] When associating a non - preset behavior model with a driving traffic participant corresponding to the moving vehicle, simulate a second dynamic response of the moving vehicle to the virtual control instruction in the simulation test scenario; and determine second simulation test data of the autonomous driving algorithm based on the second dynamic response and a corresponding preset reference response;

[0011] Based on the first simulation test data and the second simulation test data, determine the target simulation test data of the autonomous driving algorithm.

[0012] In a second aspect, according to the content of the present disclosure, a simulation test device for an autonomous driving algorithm is provided, including:

[0013] A construction module for constructing a simulation test scenario, which is constructed based on a three - dimensional model of a virtual scenario target, and the virtual scenario target includes: dynamic traffic participants, static traffic participants, and environmental participation factors;

[0014] An acquisition module for acquiring pre - set virtual sensor data corresponding to a moving vehicle in the simulation test scenario;

[0015] A first determination module for inputting the virtual sensor data into the autonomous driving algorithm to obtain a virtual control instruction for the moving vehicle output by the autonomous driving algorithm;

[0016] A first simulation module for, when associating a preset behavior model with a driving traffic participant corresponding to the moving vehicle, simulating a first dynamic response of the moving vehicle to the virtual control instruction in the simulation test scenario; and determining first simulation test data of the autonomous driving algorithm based on the first dynamic response and a corresponding preset reference response;

[0017] A second simulation module for, when associating a non - preset behavior model with a driving traffic participant corresponding to the moving vehicle, simulating a second dynamic response of the moving vehicle to the virtual control instruction in the simulation test scenario; and determining second simulation test data of the autonomous driving algorithm based on the second dynamic response and a corresponding preset reference response;

[0018] A second determination module for determining the target simulation test data of the autonomous driving algorithm based on the first simulation test data and the second simulation test data.

[0019] In a third aspect, a computer device is provided, including a memory and a processor, where a computer program is stored in the memory, and when the processor executes the computer program, the steps of the simulation test method for an autonomous driving algorithm in any one of the above embodiments are implemented.

[0020] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the simulation test method of the autonomous driving algorithm in any of the above embodiments are implemented.

[0021] The simulation test method of the autonomous driving algorithm provided by the embodiments of the present application constructs a simulation test scenario, which is constructed based on the three-dimensional model of the virtual scenario target. The virtual scenario target includes: dynamic traffic participants, static traffic participants, and environmental participation factors; obtaining the virtual sensor data preset for the driving vehicle in the simulation test scenario; and inputting the virtual sensor data into the autonomous driving algorithm to obtain the virtual control instruction of the driving vehicle output by the autonomous driving algorithm; when the driving traffic participant corresponding to the driving vehicle is associated with a preset behavior model, simulating the first dynamic response of the driving vehicle to the virtual control instruction in the simulation test scenario; and determining the first simulation test data of the autonomous driving algorithm based on the first dynamic response and the corresponding preset reference response; when the driving traffic participant corresponding to the driving vehicle is associated with a non-preset behavior model, simulating the second dynamic response of the driving vehicle to the virtual control instruction in the simulation test scenario; and determining the second simulation test data of the autonomous driving algorithm based on the second dynamic response and the corresponding preset reference response; determining the target simulation test data of the autonomous driving algorithm based on the first simulation test data and the second simulation test data. In this way, by simulating the response of the autonomous driving algorithm for the driving vehicle in the pre-constructed simulation test scenario, and respectively simulating the dynamic responses of different driving behaviors to the virtual control instruction, the simulation test efficiency of the autonomous driving algorithm is effectively improved.

[0022] The above description is only an overview of the technical solutions of the embodiments of the present application. In order to be able to understand the technical means of the embodiments of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the embodiments of the present application more obvious and understandable, the specific embodiments of the present application are hereinafter specifically exemplified. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments will be briefly described below. It should be understood that the following described drawings only relate to some embodiments of the present disclosure and do not limit the present disclosure, where:

[0024] Figure 1 is a flowchart of a simulation test method of an autonomous driving algorithm provided by the present disclosure.

[0025] Figure 2 is a structural diagram of a simulation test device of an autonomous driving algorithm provided by the present disclosure.

[0026] Figure 3It is a schematic structural diagram of a computer device provided by the present disclosure.

[0027] It should be noted that the elements in the drawings are schematic and not drawn to scale. Detailed implementation manners

[0028] In order to make the objectives, technical solutions, and advantages of the embodiments of the present disclosure clearer, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings. Apparently, the described embodiments are some but not all of the embodiments of the present disclosure. All other embodiments obtained by those skilled in the art without creative efforts based on the described embodiments of the present disclosure also belong to the scope of protection of the present disclosure.

[0029] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which the subject matter of the present disclosure belongs. It will be further understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the specification and the relevant art, and will not be interpreted in an idealized or overly formal form unless expressly defined otherwise herein. As used herein, the statement of connecting or coupling two or more parts together shall mean that these parts are directly joined together or joined through one or more intermediate components.

[0030] Referring to "embodiments" in this document means that the specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase "embodiments" appearing in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0031] The term "and / or" in this document is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: the existence of A, the simultaneous existence of A and B, and the existence of B. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after. Terms such as "first" and "second" are only used to distinguish one component (or a part of the component) from another component (or another part of the component).

[0032] In the description of the present application, unless otherwise specified, the meaning of "a plurality" refers to two or more (including two). Similarly, "a plurality of groups" refers to two or more groups (including two groups).

[0033] To enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0034] Figure 1 It is a schematic flowchart of a simulation test method for an autonomous driving algorithm provided by an embodiment of the present disclosure. In this embodiment, a simulation engine is established to schedule and run simulation test tasks. For task management: a task management system is established to allow testers to conveniently create, edit, and manage simulation test tasks. Each test task can be configured with different scenario parameters according to specific requirements, such as selecting a specific simulation test scenario (including randomly generated scenarios or artificially created extreme scenarios), setting the test duration, selecting different weather and lighting conditions, etc. Testers can also specify the initial state and goals of the autonomous driving algorithm in the task, such as the starting position of the vehicle, the driving route planning, and the criteria for judging the completion of the task. According to the task configuration information, the simulation engine is responsible for coordinating the work of each subsystem to ensure that the test task proceeds smoothly according to the predetermined plan. For example, at the start of the task, the selected scenario model and related resources are correctly loaded, the autonomous driving algorithm is initialized, and various parameters of the simulation environment are set.

[0035] For task execution: during the task execution process, the simulation engine monitors the running state of the test task in real time. This includes: tracking key information such as the position, speed, and driving trajectory of the vehicle in the simulation scenario, and at the same time monitoring the output instructions of the autonomous driving algorithm, such as the steering wheel angle, acceleration, or braking signals. By comparing with the set task goals and expected behaviors, it is judged whether the execution of the algorithm meets the requirements. For example, if the vehicle deviates from the predetermined driving route or fails to complete the task within the specified time, the system will record the relevant abnormal situations. A visual task playback interface is provided, and testers can observe the running situation of the vehicle in the simulation scenario in real time, as well as the algorithm's response to various scenario changes. This interface can also display real-time performance metrics, such as frame rate, latency, resource occupancy rate, etc., to help testers evaluate the running efficiency and stability of the simulation test. If problems occur during the test, such as system crashes, algorithm anomalies, etc., the monitoring interface will issue an alarm in a timely manner and provide detailed error information to facilitate testers to quickly locate and solve the problems.

[0036] As Figure 1 shown, the specific process of the simulation test method for the autonomous driving algorithm includes:

[0037] S110, construct a simulation test scenario.

[0038] Among them, the simulation test scenario is constructed based on the three-dimensional model of the virtual scenario target, and the virtual scenario target includes: dynamic traffic participants, static traffic participants, and environmental participation factors.

[0039] Dynamic traffic participants may include, but are not limited to: pedestrians, vehicles, animals, etc. Static traffic participants may include, but are not limited to: traffic lights, traffic signs, etc. Environmental participation factors may include, but are not limited to: cloud cover, wind force, air visibility, rainfall, solar radiation, etc.

[0040] The simulation test scenario can be a random traffic flow and simulation test scenario generated by calling the assets (i.e., the 3D models of each virtual scene target) reconstructed by AI (Artificial Intelligence). By collecting a large amount of data on real traffic scenarios, including traffic flow data under different types of roads (urban roads, highways, rural roads, etc.), different time periods (peak hours, off-peak hours, etc.), and different weather conditions (sunny days, rainy days, snowy days, foggy days, etc.), such as vehicle speed, vehicle trajectory, traffic flow, traffic participant type and behavior, etc. Preprocess this data, including data cleaning to remove outliers and incorrect data, data normalization to map data in different ranges to the same interval, and data splitting to divide the data into a training set, a validation set, and a test set, for example, divided according to a ratio of 7:2:1. Use this data to train an AI model, for example, use deep learning algorithms such as Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), etc. to train a traffic flow prediction model, which can learn the spatio-temporal patterns and rules of traffic flow.

[0041] For example, a network containing 3 layers of LSTM units is adopted, with 128 neurons in each layer. The input data is traffic flow data within 10 seconds, and the output is the traffic flow prediction value for 5 seconds. During the training process, loss functions such as mean squared error or cross-entropy are used to measure the difference between the predicted value and the true value. The weights and bias parameters of the network are continuously adjusted through the backpropagation algorithm and optimizers (such as Adam, Adagrad, etc.) to minimize the loss function. After 100 - 500 iterations, until the performance of the model on the validation set reaches a certain metric, such as the MSE is less than a certain threshold or the accuracy reaches a certain level, the training of the model is completed.

[0042] When generating a simulation test scenario, the AI model can randomly generate traffic flow data according to preset scenario parameters (such as road type, weather conditions, time, etc.). For example, according to the selected urban road scenario and peak hour parameters, the model generates corresponding information such as vehicle flow, vehicle speed distribution, vehicle driving trajectories, etc. And by using the pre-constructed basic scenario model and assets (such as road models, vehicle models, pedestrian models, etc.), traffic participants (vehicles, pedestrians, etc.) are instantiated in the simulation environment according to the generated traffic flow data, and their initial states (position, speed, direction, etc.) are set, so as to generate a random traffic flow and a complete simulation test scenario. At the same time, it is determined whether the generated scenario conforms to traffic rules and physical laws, such as maintaining a reasonable safety distance between vehicles and obeying traffic signal rules.

[0043] The simulation test scenario can be constructed through a scenario editor. A powerful and easy-to-use scenario editor tool is developed in advance. The editor provides a visual interface that allows testers to intuitively operate and edit scenario elements. For example, various types of obstacles (such as suddenly appearing vehicles, pedestrians, falling objects, etc.) can be added to the scenario, special road conditions (such as road damage, water accumulation, icing, etc.) can be set, weather conditions can be adjusted to extreme states (such as heavy rain pouring, heavy snow falling, thick fog spreading, etc.), and abnormal behaviors of traffic participants (such as vehicles driving in reverse, pedestrians suddenly rushing into the road, etc.) can be defined. Testers use the scenario editor to manually create specific simulation test scenarios according to test requirements and extreme dangerous situations that may actually be encountered; during the creation process, the attributes and behavior parameters of each element can be precisely set to ensure the authenticity and effectiveness of the scenario; after creation, these scenarios are saved in the scenario library for subsequent testing at any time. The scenario library should have a classification management function to facilitate the organization and retrieval of different types of extreme scenarios.

[0044] In addition, it can also be classified through a scenario classification system according to factors such as the characteristics, complexity, and danger level of the scenario. For example, it can be divided into conventional driving scenarios (such as normal urban road driving, highway cruising, etc.), complex scenarios (such as traffic jams, construction sections, roundabouts, etc.), and extreme scenarios (such as driving in bad weather, emergency avoidance scenarios, traffic accident scenes, etc.). At the same time, each major category can be further subdivided into smaller categories. For example, under the extreme scenario category, it can be divided into bad weather categories (heavy rain, heavy snow, thick fog, etc.), sudden danger categories (vehicle breakdown, pedestrian intrusion, road collapse, etc.), etc.

[0045] Establish a scenario management module responsible for managing all simulation test scenarios. The scenario management module has functions such as scenario addition, deletion, modification, and query. Testers can quickly retrieve the required test scenarios based on information such as scenario name, category, keywords, etc. At the same time, the scenario management module can also record relevant information such as the usage history and test results of the scenarios, so as to evaluate the effectiveness and importance of the scenarios and provide a reference basis for the optimization and update of subsequent scenarios.

[0046] S120. Obtain the virtual sensor data preset for the moving vehicle in the simulation test scenario; and input the virtual sensor data into the autonomous driving algorithm to obtain the virtual control command for the moving vehicle output by the autonomous driving algorithm.

[0047] Among them, the virtual sensor data includes image data captured by a camera, lidar point cloud data, IMU data, etc. By integrating the autonomous driving algorithm to be tested into the simulation test platform, it is ensured that the algorithm can use the virtual sensor data provided by the platform as input, and perform calculations and decisions according to the neural network model inside the algorithm, and output the virtual control command of the vehicle. The virtual control command includes steering wheel angle, acceleration or braking signal, etc.

[0048] For example, perform preprocessing on the virtual sensor data provided by the platform, such as resizing the image and normalizing the point cloud data, to make it meet the input requirements of the neural network; input the preprocessed data into the neural network, and after processing by structures such as multiple convolutional layers, pooling layers, and fully connected layers, extract the features of the data and perform classification or regression prediction. In an autonomous driving algorithm based on a convolutional neural network, the input image data extracts the features of objects such as roads, vehicles, and pedestrians through the convolutional layer, and the pooling layer performs downsampling to reduce the data volume. The fully connected layer maps the features to the final control command space. The neural network outputs the virtual control command for the vehicle according to the learned patterns and weights. During the decision-making process, the virtual control command is generated through a preset decision-making strategy. The preset decision-making strategy can be a probability-based decision, such as selecting the one with the highest probability among multiple possible driving directions or operations, or the preset decision-making strategy can be a threshold-based decision, such as triggering a braking command when the detected distance to the obstacle ahead is less than a certain safety threshold.

[0049] Run the end-to-end autonomous driving algorithm in the reconstructed 3D scene model to simulate the autonomous driving process of the vehicle in various scenarios and working conditions. Different test scenario sequences can be set, including normal driving scenarios (such as urban road cruising, highway driving, etc.), complex scenarios (such as traffic congestion, sudden appearance of obstacles, etc.), and extreme scenarios (such as driving under bad weather conditions, emergency avoidance, etc.), so as to comprehensively test the performance of the algorithm in different situations.

[0050] S130. When associating a preset behavior model with a driving traffic participant corresponding to a moving vehicle, simulate a first dynamic response of the moving vehicle to a virtual control instruction in a simulation test scenario; and determine first simulation test data of the autonomous driving algorithm based on the first dynamic response and the corresponding preset reference response.

[0051] Among them, the driving traffic participant is the driver of the moving vehicle, and the behavior model can be used to identify the driving style of the driver. The preset behavior model can be used to describe that the driving style of the driving traffic participant is a conservative driving style. Correspondingly, the non-preset behavior model can be used to describe that the driving style of the driving traffic participant is an aggressive driving style. Vehicles with an aggressive driving style may overtake, accelerate, and brake suddenly more frequently, while vehicles with a conservative driving style pay more attention to maintaining a safe distance and a stable driving speed.

[0052] In this embodiment, by introducing the advanced physics engine UE5 (Unreal Engine 5), the physical behaviors of vehicles and traffic participants are simulated based on Newton's laws of motion. When simulating the movement of a vehicle, calculate the driving force of the vehicle according to physical properties such as the mass and inertia tensor of the vehicle, as well as parameters such as the torque-speed curve of the engine and the transmission ratio of the transmission system. Combine the friction coefficient model between the tire and the ground (this coefficient will change dynamically according to factors such as road surface type and weather conditions) to calculate the friction force of the vehicle on different road surfaces. By solving the dynamic equation of the vehicle (such as the application of Newton's second law in vehicle motion), obtain the changes in the motion states of the vehicle such as acceleration, speed, and displacement.

[0053] In some embodiments, the virtual control instruction is an acceleration instruction. Simulating a first dynamic response of the moving vehicle to the virtual control instruction in a simulation test scenario includes:

[0054] Determine a first acceleration of the moving vehicle in the simulation test scenario in response to the virtual control instruction based on the vehicle attribute data and the vehicle driving state corresponding to the driving traffic participant associated with the preset behavior model; control the moving vehicle to perform an acceleration action in the simulation test scenario based on the first acceleration to obtain the first dynamic response.

[0055] Among them, the vehicle attribute data includes: engine power, vehicle weight. The vehicle driving state corresponding to the driving traffic participant associated with the preset behavior model can be, for example, the vehicle speed. When simulating acceleration, according to the engine power curve and the current vehicle speed, use the power formula, power = torque × angular velocity, to calculate the torque output by the engine; combine the efficiency and transmission ratio of the transmission system to obtain the driving torque on the wheels; according to the friction force between the wheels and the ground and the mass of the vehicle, use Newton's second law, F = ma, to calculate the actual acceleration of the vehicle, where F is the resultant force, m is the vehicle mass, and a is the acceleration.

[0056] For example, for a 1500kg car, the engine output torque is 200N·m at a certain speed, the transmission system efficiency is 0.9, the gear ratio is 4, the wheel radius is 0.3m, and the road friction coefficient is 0.8. The driving force on the wheel is calculated to be 5400N. According to the friction formula F_friction = μmg, the friction force is calculated to be 3528N, where μ is the friction coefficient and g is the acceleration due to gravity. The net force on the vehicle is 1872N, and the calculated acceleration is approximately 1.25m / s. 2 .

[0057] When the virtual control instruction is an acceleration instruction, the moving vehicle is controlled to perform an acceleration action in the simulation test scenario based on the first acceleration, and the obtained first dynamic response is the simulated acceleration of the moving vehicle in the simulation test scenario.

[0058] Accordingly, the preset reference response may be a preset acceleration, and the first simulation test data of the autonomous driving algorithm may be determined based on the difference between the simulated acceleration and the preset acceleration. The smaller the difference between the simulated acceleration and the preset acceleration, the larger the first simulation test data of the autonomous driving algorithm; and the larger the difference between the simulated acceleration and the preset acceleration, the smaller the first simulation test data of the autonomous driving algorithm.

[0059] The virtual control command is a steering command. The first dynamic response of the moving vehicle to the virtual control command is simulated in the simulation test scenario, including:

[0060] Based on the friction of the vehicle structure data relative to the ground and the vehicle driving state corresponding to the driving traffic participant in the associated preset behavior model, a first steering trajectory of the moving vehicle in response to the virtual control command in the simulation test scenario is determined; the moving vehicle is controlled to travel according to the first steering trajectory in the simulation test scenario to obtain a first dynamic response.

[0061] In the steering simulation, the mechanical structure and steering assist characteristics of the vehicle's steering system are taken into consideration. Based on the steering wheel angle input and the vehicle's speed, the vehicle's steering radius and steering angular velocity are calculated to determine the steering trajectory and ensure that the vehicle's steering trajectory conforms to actual physical laws.

[0062] When the virtual control instruction is a steering instruction, the moving vehicle is controlled to move along a first steering trajectory in the simulation test scenario, and the obtained first dynamic response is the simulated steering trajectory of the moving vehicle in the simulation test scenario.

[0063] Correspondingly, the preset reference response can be a preset steering trajectory. Based on the deviation between the simulated steering trajectory and the preset steering trajectory (such as the sum of the error values of each point on the curve), the first simulation test data of the autonomous driving algorithm can be determined. The smaller the deviation between the simulated steering trajectory and the preset steering trajectory, the larger the first simulation test data of the autonomous driving algorithm; the larger the deviation between the simulated steering trajectory and the preset steering trajectory, the smaller the first simulation test data of the autonomous driving algorithm.

[0064] S140. When associating a non-preset behavior model with a driving traffic participant corresponding to a driving vehicle, simulate the second dynamic response of the driving vehicle to the virtual control instruction in the simulation test scenario; and determine the second simulation test data of the autonomous driving algorithm based on the second dynamic response and the corresponding preset reference response.

[0065] In some embodiments, the virtual control instruction is an acceleration instruction. Simulating the second dynamic response of the driving vehicle to the virtual control instruction in the simulation test scenario includes:

[0066] Based on the vehicle attribute data, the preset behavior control data, and the vehicle driving state of the driving traffic participant associated with the non-preset behavior model, determine the second acceleration of the driving vehicle in the simulation test scenario in response to the virtual control instruction; control the driving vehicle to perform an acceleration action in the simulation test scenario based on the second acceleration to obtain the second dynamic response.

[0067] Among them, the preset behavior control data can be a time conversion coefficient between the preset behavior model and the non-preset behavior model, and this conversion coefficient is greater than 1.

[0068] For example, when the first acceleration value is x1 for the driving traffic participant associated with the preset behavior model, the second acceleration of the driving vehicle in the simulation test scenario in response to the virtual control instruction can be 1.3x1.

[0069] When the virtual control instruction is an acceleration instruction, controlling the driving vehicle to perform an acceleration action in the simulation test scenario based on the second acceleration, the obtained second dynamic response is the simulated acceleration of the driving vehicle in the simulation test scenario.

[0070] Correspondingly, the preset reference response can be a preset acceleration. Based on the difference between the simulated acceleration and the preset acceleration, the second simulation test data of the autonomous driving algorithm can be determined. The smaller the difference between the simulated acceleration and the preset acceleration, the larger the second simulation test data of the autonomous driving algorithm; the larger the difference between the simulated acceleration and the preset acceleration, the smaller the second simulation test data of the autonomous driving algorithm.

[0071] In some embodiments, the virtual control instruction is a steering instruction. Simulating the second dynamic response of the driving vehicle to the virtual control instruction in the simulation test scenario includes:

[0072] Based on the frictional force of the vehicle structure data relative to the ground, the preset behavior control data, and the vehicle driving state corresponding to the driving traffic participant associated with the non-preset behavior model, determine the second steering trajectory of the driving vehicle in response to the virtual control instruction in the simulation test scenario; control the driving vehicle to drive along the second steering trajectory in the simulation test scenario to obtain the second dynamic response.

[0073] Among them, after determining the initial steering trajectory of the driving vehicle in response to the virtual control instruction in the simulation test scenario through the frictional force of the vehicle structure data relative to the ground and the vehicle driving state corresponding to the driving traffic participant associated with the non-preset behavior model, the preset driving control data can be used to adjust the initial steering trajectory, such as increasing the curve radian corresponding to the trajectory, to obtain the second steering trajectory.

[0074] When the virtual control instruction is a steering instruction, control the driving vehicle to drive along the first steering trajectory in the simulation test scenario, and the second dynamic response obtained is the simulated steering trajectory of the driving vehicle in the simulation test scenario.

[0075] Correspondingly, the preset reference response can be a preset steering trajectory. The second simulation test data of the autonomous driving algorithm can be determined according to the deviation between the simulated steering trajectory and the preset steering trajectory (such as the sum of the error values of each point in the curve). The smaller the deviation between the simulated steering trajectory and the preset steering trajectory, the larger the second simulation test data of the autonomous driving algorithm, and the larger the deviation between the simulated steering trajectory and the preset steering trajectory, the smaller the second simulation test data of the autonomous driving algorithm.

[0076] S150. Determine the target simulation test data of the autonomous driving algorithm based on the first simulation test data and the second simulation test data.

[0077] Among them, different / same weights can be assigned to the first simulation test data and the second simulation test data respectively, and then weighted summation is performed to obtain the target simulation test data of the autonomous driving algorithm.

[0078] In addition, a series of performance evaluation indicators can also be set to measure the performance of the end-to-end autonomous driving algorithm in the simulation test. The safety indicators include whether the vehicle can avoid collisions, maintain a safe distance from obstacles, and comply with traffic rules (such as not running red lights, not speeding, etc.); the comfort indicators focus on the smoothness of the vehicle driving (such as whether the changes in acceleration and deceleration are gentle), and the riding experience (such as whether to avoid sudden braking and sharp turning); the accuracy indicators evaluate the closeness of the control instructions output by the algorithm to the ideal control instructions (such as the error of the steering wheel angle, the accuracy of speed control, etc.); the efficiency indicators consider the time and energy consumption required for the vehicle to complete specific tasks (such as driving from the starting point to the end point), etc.

[0079] In this embodiment, a simulation test scenario is constructed. The simulation test scenario is constructed based on the three-dimensional model of the virtual scenario target, and the virtual scenario target includes: dynamic traffic participants, static traffic participants, and environmental participation factors. The virtual sensor data preset for the driving vehicle in the simulation test scenario is obtained, and the virtual sensor data is input into the autonomous driving algorithm to obtain the virtual control command for the driving vehicle output by the autonomous driving algorithm. When a preset behavior model is associated with the driving traffic participant corresponding to the driving vehicle, the first dynamic response of the driving vehicle to the virtual control command is simulated in the simulation test scenario, and the first simulation test data of the autonomous driving algorithm is determined based on the first dynamic response and the corresponding preset reference response. When a non-preset behavior model is associated with the driving traffic participant corresponding to the driving vehicle, the second dynamic response of the driving vehicle to the virtual control command is simulated in the simulation test scenario, and the second simulation test data of the autonomous driving algorithm is determined based on the second dynamic response and the corresponding preset reference response. The target simulation test data of the autonomous driving algorithm is determined based on the first simulation test data and the second simulation test data. In this way, by simulating the response of the autonomous driving algorithm for the driving vehicle in the pre-constructed simulation test scenario and respectively simulating the dynamic responses of different driving behaviors to the virtual control command, the simulation test efficiency of the autonomous driving algorithm is effectively improved.

[0080] In some embodiments, it further includes:

[0081] If a collision occurs between the driving vehicle and a preset virtual target when the driving vehicle is driving in the simulation test scenario, the collision information between the driving vehicle and the preset virtual target is determined through a collision detection mechanism, and the third dynamic response of the driving vehicle in the simulation test scenario after the collision is determined based on the collision information and the material attributes corresponding to the preset virtual target.

[0082] Among them, when a collision occurs, the collision force and the motion state after the collision can be accurately calculated according to factors such as the material, speed, and mass of the colliding object. The motion state after the collision includes: the rebound, rollover, or damage degree of the vehicle, etc. The collision information can be, for example, the collision force and the running state of the vehicle after the collision.

[0083] Specifically, the colliding object pair, the collision point, and the collision normal direction can be determined through a collision detection algorithm, and the collision force is calculated using an elastic collision or inelastic collision model according to the material attributes (such as elastic modulus, Poisson's ratio, etc.) of the colliding object. For elastic collisions, the velocity and motion direction after the collision are calculated according to the laws of conservation of momentum and conservation of energy. For inelastic collisions, considering the energy loss during the collision process, the velocity after the collision is adjusted by setting a restitution coefficient.

[0084] For example, in a scenario where two vehicles collide (i.e., the colliding objects are vehicles), one vehicle has a mass of 1200 kg and a speed of 10 m / s, the other vehicle has a mass of 1500 kg and a speed of 8 m / s, and the collision angle is 30°. Assuming that both vehicles are in a partially elastic collision with a coefficient of restitution of 0.6. By calculating the momentum and energy of the two vehicles in the collision direction and combining the coefficient of restitution, the speeds and directions of motion of the two vehicles after the collision can be obtained. At the same time, the degree of damage to the vehicles (i.e., the third dynamic response) can be calculated based on the collision force and the structural strength of the vehicles. The degree of damage includes factors such as the deformation of the vehicle body and the damage of components.

[0085] In some embodiments, it further includes:

[0086] Based on the historical operating state of a moving vehicle in a simulation test scenario, a temporary traffic participant target is generated in the simulation test scenario; based on the temporary traffic participant target, an emergency event simulation is performed on the moving vehicle to obtain a fourth dynamic response of the moving vehicle in the simulation test scenario when responding to the emergency event.

[0087] Among them, the temporary traffic participant target can be a dynamic traffic participant, a static traffic participant, or an environmental participation factor. For example, if there is no vehicle / pedestrian congestion in the historical operating state of the moving vehicle in the simulation test scenario, the generated temporary traffic participant target can be a dynamic traffic participant, such as a vehicle or a pedestrian; if there is no road condition such as a road surface collapse in the historical operating state of the moving vehicle in the simulation test scenario, the generated temporary traffic participant target can be a static traffic participant, such as a road surface collapse; if there is no rainy day in the historical operating state of the moving vehicle in the simulation test scenario, the generated temporary traffic participant target can be an environmental participation factor, such as heavy rain weather.

[0088] The fourth dynamic response can be whether the moving vehicle takes correct measures to respond to the emergency event, such as stopping, avoiding, decelerating, etc.

[0089] Thus, during the testing process of the autonomous driving algorithm, by temporarily adding emergency event situations, it is convenient to test the processing ability of the autonomous driving algorithm for emergency events, ensuring that the autonomous driving algorithm in the real environment can provide safety guarantees for driving.

[0090] This embodiment accurately restores the physical phenomena and vehicle dynamics responses in the real scenario by simulating extreme situations through an accurate physical engine and behavior model, realizing the accurate evaluation of the performance of the end-to-end autonomous driving algorithm in extreme situations, providing a reliable basis for algorithm optimization, and enhancing the reliability of the system in high-risk scenarios.

[0091] In some embodiments, it further includes:

[0092] Update the environmental participation factor in the simulation test scenario according to the virtual driving duration of the moving vehicle in the simulation test scenario, so as to reconstruct the environment of the simulation test scenario.

[0093] For example, according to different weather conditions, such as rainfall, snowfall, foggy days, etc., adjust the physical characteristics and visual effects of the scenario. In the rainy day scenario, simulate the falling of raindrops and the accumulation of water on the road surface, increase the slipperiness of the road surface, affect the braking distance and handling performance of the vehicle; at the same time, reduce the visibility, affecting the detection range and accuracy of the sensor for the surrounding environment. On a snowy day, simulate the falling of snowflakes and the covering of the road surface by snow accumulation, change the road friction, making it more difficult for the vehicle to drive, and further reducing the visibility. The foggy day scenario simulates the scattering and attenuation of light in the fog by increasing the fog concentration, making distant objects blurred, seriously affecting visual perception, so as to comprehensively test the adaptability and robustness of the autonomous driving algorithm under bad weather conditions.

[0094] Adjust the light and shadow effects of the scenario according to the changes in the light angle and intensity at different times of the day. For example, in the early morning and evening, the sun angle is relatively low, the light is soft and the color is warm, which will produce longer shadows; while at noon, the sun shines directly, the light is strong and the shadows are shorter. The night scene creates a real night driving environment through street lamp lighting and vehicle lighting effects. The change of light not only affects the visual effect, but also has an impact on the performance of the sensor. For example, in low light conditions, the image sensor may generate more noise, affecting the accuracy of the image recognition algorithm. By simulating these light changes, the perception and decision-making capabilities of the autonomous driving algorithm in different light environments can be tested.

[0095] Update the road traffic flow in the simulation test scenario according to the road traffic data of the moving vehicle in the simulation test scenario, so as to reconstruct the traffic flow of the simulation test scenario.

[0096] Among them, reconstruct the traffic flow according to the road traffic data of the moving vehicle in the simulation test scenario. For example, when the road traffic data shows that the current section belongs to the peak section, increase the traffic flow of vehicles and pedestrians in the current section of the simulation test scenario.

[0097] Thus, by reconstructing the environment of the simulation test scenario and reconstructing the traffic flow of the simulation test scenario, it is convenient to make the simulation test scenario closer to the real scenario, and further ensure the safe application of the autonomous driving algorithm in the real scenario.

[0098] In addition, according to traffic rules and road conditions, the driving trajectory of a moving vehicle can be reasonably generated. For example, at an intersection, an appropriate turning or straight-ahead route can be selected based on the signal light status and traffic flow. Specifically, the topological structure of the current road (such as the number of lanes, lane width, intersection type, etc.) and traffic sign information (such as speed limit signs, no-overtaking signs, etc.) are obtained; a traffic flow model based on deep learning is used to predict the motion states of surrounding vehicles and the changing trend of traffic flow, and the driving trajectory of the vehicle is generated.

[0099] For example, in an intersection scenario, if the traffic signal is green and the road ahead is clear, the moving vehicle can maintain a straight drive or turn with a certain curvature according to a preset speed plan (such as according to speed limit and safety distance requirements). If the vehicle ahead decelerates or stops, the vehicle can adjust its own speed according to a safety distance model (such as a safety distance model based on time interval or distance) to ensure a safe distance from the vehicle ahead. In an overtaking scenario, the vehicle can judge whether the overtaking conditions are met (such as the vehicle ahead is moving slowly and there are no obstacles in the adjacent lane) based on the speeds and positions of surrounding vehicles. If the conditions are met, a driving trajectory is generated according to a certain overtaking strategy (such as accelerating, changing lanes, maintaining a safe distance and then changing back to the original lane after overtaking).

[0100] In some embodiments, it further includes:

[0101] Generating test feedback information corresponding to the autonomous driving algorithm based on the target simulation test data of the autonomous driving algorithm; optimizing the autonomous driving algorithm based on the test feedback information.

[0102] Among them, during the simulation test process, the virtual sensor data input by the algorithm, the vehicle control commands output, and the state information of the vehicle in the simulation scenario are recorded in real time. The state information includes position, speed, acceleration, etc. At the same time, relevant parameters of the simulation test scenario are recorded, such as weather conditions, traffic flow, road conditions, etc., so as to analyze the relationship between the algorithm performance and different scenario factors later.

[0103] The recorded data is analyzed to calculate the values of various performance evaluation indicators (i.e., test feedback information). By comparing the performance of the algorithm in different scenarios, the advantages and disadvantages of the algorithm are found. For example, analyze the safety performance of the algorithm under different weather conditions to determine in which weather the algorithm is prone to problems; study the decision-making accuracy of the algorithm in complex traffic scenarios to judge the algorithm's ability to handle multi-vehicle interactions and emergencies.

[0104] When comparing the performance of an algorithm in different scenarios to identify its advantages and disadvantages, take the security performance analysis as an example. Under different weather conditions (such as sunny, rainy, and foggy days), the autonomous driving algorithm is run for multiple simulation tests; indicators such as the minimum distance between the vehicle and obstacles, whether a collision occurs, and the number of traffic rule violations are recorded for each test. For the sunny scenario, assume that after 100 tests, the vehicle maintains an average safe distance of 5 meters from obstacles, no collision occurs, and the number of traffic rule violations is 0; in the rainy scenario, after the same number of tests, the average distance between the vehicle and obstacles decreases to 3 meters, the number of collisions is 5 times, and the number of traffic rule violations is 2 times; in the foggy scenario, the average distance between the vehicle and obstacles is only 2 meters, the number of collisions increases to 10 times, and the number of traffic rule violations is 5 times. Through comparative analysis, the differences in the security performance of the algorithm under different weather conditions can be obtained, and it can be determined that the algorithm is prone to problems in adverse weather (such as foggy days) and needs further optimization and improvement. Similar methods can also be used to analyze other performance indicators such as comfort, accuracy, and efficiency. By collecting and statistically analyzing relevant data in different scenarios, such as the acceleration and deceleration change curves of the vehicle, the error distribution of control commands, the time and energy consumption to complete tasks, etc., the performance of the algorithm can be comprehensively evaluated, providing targeted directions and suggestions for the improvement of the algorithm.

[0105] Figure 2 The following is a schematic structural diagram of a simulation test device for an autonomous driving algorithm provided in this embodiment. The simulation test device for the autonomous driving algorithm may include: a construction module 210, an acquisition module 220, a first determination module 230, a first simulation module 240, a second simulation module 250, and a second determination module 260.

[0106] The construction module 210 is used to construct a simulation test scenario, which is constructed based on a three-dimensional model of a virtual scenario target. The virtual scenario target includes: dynamic traffic participants, static traffic participants, and environmental participation factors.

[0107] The acquisition module 220 is used to acquire the virtual sensor data preset for the driving vehicle in the simulation test scenario.

[0108] The first determination module 230 is used to input the virtual sensor data into the autonomous driving algorithm to obtain the virtual control command for the driving vehicle output by the autonomous driving algorithm.

[0109] The first simulation module 240 is used to simulate the first dynamic response of the driving vehicle to the virtual control command in the simulation test scenario when associating a preset behavior model with the driving traffic participant corresponding to the driving vehicle; and determine the first simulation test data of the autonomous driving algorithm based on the first dynamic response and the corresponding preset reference response.

[0110] A second simulation module 250, configured to simulate a second dynamic response of a traveling vehicle to a virtual control instruction in a simulation test scenario when a driving traffic participant corresponding to the traveling vehicle is associated with a non - preset behavior model; and determine second simulation test data of the autonomous driving algorithm based on the second dynamic response and a corresponding preset reference response.

[0111] A second determination module 260, configured to determine target simulation test data of the autonomous driving algorithm based on the first simulation test data and the second simulation test data.

[0112] In this embodiment, optionally, the first simulation module 240 is specifically configured to:

[0113] Determine a first acceleration of the traveling vehicle in the simulation test scenario in response to the virtual control instruction based on vehicle attribute data and the vehicle driving state of the driving traffic participant associated with the preset behavior model; control the traveling vehicle to perform an acceleration action in the simulation test scenario based on the first acceleration to obtain a first dynamic response; determine a first steering trajectory of the traveling vehicle in the simulation test scenario in response to the virtual control instruction based on the friction force of the vehicle structure data relative to the ground and the vehicle driving state of the driving traffic participant associated with the preset behavior model; control the traveling vehicle to travel along the first steering trajectory in the simulation test scenario to obtain a first dynamic response.

[0114] In this embodiment, optionally, the second simulation module 250 is specifically configured to:

[0115] Determine a second acceleration of the traveling vehicle in the simulation test scenario in response to the virtual control instruction based on vehicle attribute data, preset behavior control data, and the vehicle driving state of the driving traffic participant associated with the non - preset behavior model; control the traveling vehicle to perform an acceleration action in the simulation test scenario based on the second acceleration to obtain a second dynamic response; determine a second steering trajectory of the traveling vehicle in the simulation test scenario in response to the virtual control instruction based on the friction force of the vehicle structure data relative to the ground, the preset behavior control data, and the vehicle driving state of the driving traffic participant associated with the non - preset behavior model; control the traveling vehicle to travel along the second steering trajectory in the simulation test scenario to obtain a second dynamic response.

[0116] In this embodiment, optionally, it further includes: a third determination module.

[0117] The third determination module is configured to, if a collision occurs between the traveling vehicle and a preset virtual target when the traveling vehicle is traveling in the simulation test scenario, determine collision information between the traveling vehicle and the preset virtual target through a collision detection mechanism; and determine a corresponding third dynamic response of the traveling vehicle in the simulation test scenario after the collision based on the collision information and the material attributes of the preset virtual target.

[0118] In this embodiment, optionally, it further includes: a generation module and a fourth determination module.

[0119] The generation module is configured to generate a temporary traffic participant target in the simulation test scenario based on the historical operating state of the moving vehicle in the simulation test scenario.

[0120] The fourth determination module is configured to simulate an emergency for the moving vehicle based on the temporary traffic participant target, and obtain a fourth dynamic response of the moving vehicle to the emergency in the simulation test scenario.

[0121] In this embodiment, optionally, it further includes: a reconstruction module.

[0122] The reconstruction module is configured to update the environmental participation factor in the simulation test scenario according to the virtual driving duration of the moving vehicle in the simulation test scenario, so as to perform environmental reconstruction on the simulation test scenario; update the road traffic flow in the simulation test scenario according to the driving road data of the moving vehicle in the simulation test scenario, so as to perform traffic flow reconstruction on the simulation test scenario.

[0123] In this embodiment, optionally, it further includes: an optimization module.

[0124] The generation module is further configured to generate test feedback information corresponding to the autonomous driving algorithm based on the target simulation test data of the autonomous driving algorithm.

[0125] The optimization module is configured to perform optimization processing on the autonomous driving algorithm based on the test feedback information.

[0126] The simulation test device for the autonomous driving algorithm provided by the present disclosure can execute the above method embodiments, and its specific implementation principle and technical effects can be referred to the above method embodiments, which will not be elaborated herein by the present disclosure.

[0127] The embodiment of the present application also provides a computer device. Specifically, please refer to Figure 3 , Figure 3 which is the basic structural block diagram of the computer device in this embodiment.

[0128] The computer device includes a memory 310 and a processor 320 that are communicatively connected to each other via a system bus. It should be noted that only the computer device with the memory 310 and the processor 320 is shown in the figure, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Among them, those skilled in the art of the present technology can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0129] The computer device can be a desktop computer, a notebook, a palm computer, a cloud server and other computing devices. The computer device can interact with the user through a keyboard, a mouse, a remote control, a touchpad or a voice control device and other means.

[0130] The memory 310 includes at least one type of readable storage medium, and the readable storage medium includes non-volatile memory or volatile memory, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory, etc.), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disc, etc. The RAM may include static RAM or dynamic RAM. In some embodiments, the memory 310 may be an internal storage unit of a computer device, such as the hard disk or memory of the computer device. In other embodiments, the memory 310 may also be an external storage device of the computer device, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device. Of course, the memory 310 may also include both the internal storage unit and the external storage device of the computer device. In this embodiment, the memory 310 is generally used to store the operating system and various application software installed on the computer device, such as the program code of the above method. In addition, the memory 310 may also be used to temporarily store various data that have been output or will be output.

[0131] The processor 320 is generally used to execute the overall operations of the computer device. In this embodiment, the memory 310 is used to store program code or instructions, and the program code includes computer operation instructions. The processor 320 is used to execute the program code or instructions stored in the memory 310 or process data, such as running the program code of the above method.

[0132] In this article, the bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. This bus system can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience in representation, only a thick line is used in the figure to represent it, but it does not mean that there is only one bus or one type of bus.

[0133] Another embodiment of the present application further provides a computer-readable medium, which can be a computer-readable signal medium or a computer-readable medium. The processor in the computer reads the computer-readable program code stored in the computer-readable medium, so that the processor can execute the functional actions specified in each step or the combination of steps in the above method; and generate a device that implements the functional actions specified in each block or the combination of blocks in the block diagram.

[0134] The computer-readable medium includes but is not limited to electronic, magnetic, optical, electromagnetic, infrared memories or semiconductor systems, devices or apparatuses, or any suitable combination of the foregoing. The memory is used to store program codes or instructions, and the program codes include computer operation instructions. The processor is used to execute the program codes or instructions of the above method stored in the memory.

[0135] For the definitions of the memory and the processor, reference can be made to the description of the foregoing computer device embodiments, and details are not described herein again.

[0136] In several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of the devices or units can be in an electrical, mechanical or other form.

[0137] In each embodiment of the present application, each functional unit or module can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0138] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods in various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0139] In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The "including" described in this application does not exclude the existence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the existence of a plurality of such elements. This application can be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In the claims listing several units of a device, several of these units of the device can be embodied by the same hardware item. The use of "first", "second", and "third", etc. does not denote any order and these words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

[0140] The above embodiments are only used to illustrate the technical solution of this application, rather than to limit it; although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of this application.

Claims

1. A simulation test method for an autonomous driving algorithm, characterized in that Including: Construct a simulation test scenario, which is constructed based on the three-dimensional model of the virtual scenario target. The virtual scenario target includes: dynamic traffic participants, static traffic participants, and environmental participation factors; Obtain the virtual sensor data preset for the traveling vehicle in the simulation test scenario; and input the virtual sensor data into the autonomous driving algorithm to obtain the virtual control instruction for the traveling vehicle output by the autonomous driving algorithm; When associating a preset behavior model with the driving traffic participant corresponding to the traveling vehicle, simulate the first dynamic response of the traveling vehicle to the virtual control instruction in the simulation test scenario; and determine the first simulation test data of the autonomous driving algorithm based on the first dynamic response and the corresponding preset reference response; When associating a non-preset behavior model with the driving traffic participant corresponding to the traveling vehicle, simulate the second dynamic response of the traveling vehicle to the virtual control instruction in the simulation test scenario; and determine the second simulation test data of the autonomous driving algorithm based on the second dynamic response and the corresponding preset reference response; Based on the first simulation test data and the second simulation test data, determine the target simulation test data of the autonomous driving algorithm.

2. The method according to claim 1, characterized in that, The virtual control instruction is an acceleration instruction. The simulation of the first dynamic response of the traveling vehicle to the virtual control instruction in the simulation test scenario includes: Based on the vehicle attribute data and the vehicle driving state corresponding to the driving traffic participant associated with the preset behavior model, determine the first acceleration of the traveling vehicle in the simulation test scenario in response to the virtual control instruction; Based on the first acceleration, control the traveling vehicle to perform an acceleration action in the simulation test scenario to obtain the first dynamic response; The simulation of the second dynamic response of the traveling vehicle to the virtual control instruction in the simulation test scenario includes: Based on the vehicle attribute data, the preset behavior control data, and the vehicle driving state corresponding to the driving traffic participant associated with the non-preset behavior model, determine the second acceleration of the traveling vehicle in the simulation test scenario in response to the virtual control instruction; Based on the second acceleration, control the traveling vehicle to perform an acceleration action in the simulation test scenario to obtain the second dynamic response.

3. The method according to claim 1, characterized in that, The virtual control instruction is a steering instruction. The simulation of the first dynamic response of the traveling vehicle to the virtual control instruction in the simulation test scenario includes: Based on the friction force of the vehicle structure data relative to the ground and the vehicle driving state corresponding to the driving traffic participant associated with the preset behavior model, determine the first steering trajectory of the traveling vehicle in the simulation test scenario in response to the virtual control instruction; Control the traveling vehicle to travel along the first steering trajectory in the simulation test scenario to obtain the first dynamic response; The simulation of the second dynamic response of the traveling vehicle to the virtual control instruction in the simulation test scenario includes: Determine a second steering trajectory of the traveling vehicle in the simulation test scenario in response to the virtual control instruction based on the frictional force of the vehicle structure data relative to the ground, the preset behavior control data, and the vehicle driving state corresponding to the driving traffic participant associated with the non-preset behavior model; Control the traveling vehicle to travel along the second steering trajectory in the simulation test scenario to obtain the second dynamic response.

4. The method according to claim 1, wherein Further included are: If the traveling vehicle collides with a preset virtual target during traveling in the simulation test scenario, determine the collision information of the traveling vehicle and the preset virtual target through a collision detection mechanism; Based on the collision information and the material attributes corresponding to the preset virtual target, determine the third dynamic response corresponding to the traveling vehicle in the simulation test scenario after the collision.

5. The method according to claim 1, characterized in that, Further included are: Generate a temporary traffic participation target in the simulation test scenario based on the historical running state of the traveling vehicle in the simulation test scenario; Perform an emergency simulation on the traveling vehicle based on the temporary traffic participation target to obtain a fourth dynamic response of the traveling vehicle in the simulation test scenario in response to the emergency.

6. The method according to claim 1, wherein Further included are: Update the environmental participation factor in the simulation test scenario according to the virtual driving duration of the traveling vehicle in the simulation test scenario to reconstruct the environment of the simulation test scenario; Update the road traffic flow in the simulation test scenario according to the road traffic data of the traveling vehicle in the simulation test scenario to reconstruct the traffic flow of the simulation test scenario.

7. The method according to claim 1, wherein Further included are: Generate test feedback information corresponding to the autonomous driving algorithm based on the target simulation test data of the autonomous driving algorithm; Perform an optimization process on the autonomous driving algorithm based on the test feedback information.

8. A simulation test device for an autonomous driving algorithm, characterized in that, Including: A construction module for constructing a simulation test scenario, which is constructed based on a three-dimensional model of a virtual scenario target, and the virtual scenario target includes: dynamic traffic participants, static traffic participants, and environmental participation factors; An acquisition module for acquiring the preset virtual sensor data corresponding to the traveling vehicle in the simulation test scenario; A first determination module for inputting the virtual sensor data into the autonomous driving algorithm to obtain a virtual control instruction output by the autonomous driving algorithm for the traveling vehicle; A first simulation module for simulating a first dynamic response of the traveling vehicle to the virtual control instruction in the simulation test scenario when the driving traffic participant corresponding to the traveling vehicle is associated with a preset behavior model; and determining the first simulation test data of the autonomous driving algorithm based on the first dynamic response and the corresponding preset reference response; A second simulation module for simulating a second dynamic response of the traveling vehicle to the virtual control instruction in the simulation test scenario when the driving traffic participant corresponding to the traveling vehicle is associated with a non-preset behavior model; and determining the second simulation test data of the autonomous driving algorithm based on the second dynamic response and the corresponding preset reference response; A second determination module, configured to determine target simulation test data of the autonomous driving algorithm based on the first simulation test data and the second simulation test data.

9. A computer device, characterized in that, It includes a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, it implements the simulation test method of the autonomous driving algorithm according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the simulation test method of the autonomous driving algorithm according to any one of claims 1 to 7.

Images