Automatic driving simulation scene generalization method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses an automatic driving simulation scene generalization method and device, electronic equipment and a storage medium, and the method comprises the steps: adding a random variable for an initial simulation scene case based on a random sampling algorithm, and obtaining a plurality of derivative simulation scene cases, the initial simulation scene use case is a real high-risk scene use case or an automatically defined scene use case; for the plurality of derivative simulation scene cases, performing scene verification processing on the derivative simulation scene cases based on compliance verification conditions of at least one dimension to obtain derivative scene compliance attributes; based on the derivative scene compliance attribute of each derivative simulation scene case, a target simulation scene case set is determined, and a generalization scene case set is stored in a target format to form a scene library with high coverage rate and high-risk scene discovery capability, so that the generation efficiency of the automatic driving scene is improved, the risk mode generalization capability of the generated scene is realized, and the generation efficiency of the automatic driving scene is improved. And a more comprehensive and efficient test verification basis is provided for the automatic driving system.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a method, device, electronic device, and storage medium for generalizing autonomous driving simulation scenarios. Background Art

[0002] The practical application of autonomous driving technology requires rigorous and comprehensive testing and verification. This is due to the inherently high-risk nature of vehicles—algorithm failures can lead to significant economic losses and even endanger lives. Current testing methods are primarily categorized into two types: real-world road testing and simulation platform testing. While real-world testing ensures environmental realism, it is constrained by high costs and policy restrictions, making it difficult to cover the vast state space of autonomous driving scenarios. Simulation testing, on the other hand, is essential for large-scale verification due to its cost advantages and controllable scenarios. Existing high-fidelity simulation platforms have already implemented the construction of digital twin environments driven by physics engines, providing the foundation for testing.

[0003] Currently, simulation scenario generation technology faces two core challenges: scenario generalization and coverage of high-risk scenarios. Regarding generalization, existing methods often rely on rule engines to pre-set a limited number of scenarios. For high-risk scenario processing, traditional methods rely primarily on replaying real accident data.

[0004] However, the above-mentioned existing technologies suffer from the following drawbacks: First, rule-driven scenario generation suffers from a combinatorial explosion. For example, to cover all parameter combinations for a simple three-vehicle interaction scenario, millions of scenarios would need to be generated, making manual design completely impractical. Second, methods based on real data struggle to generalize new hazard patterns beyond those in historical data. Consequently, existing autonomous driving scenario generation methods suffer from low generation efficiency and a lack of generalization of hazard patterns. Summary of the Invention

[0005] The present invention provides a method, device, electronic device and storage medium for generalizing autonomous driving simulation scenarios to form a scenario library with both high coverage and high-risk scenario detection capabilities, improve the generation efficiency of autonomous driving scenarios, and achieve the ability to generalize dangerous patterns in the generated scenarios.

[0006] In a first aspect, an embodiment of the present invention provides a method for generalizing an autonomous driving simulation scenario, the method comprising:

[0007] Adding random variables to the initial simulation scenario use case based on a random sampling algorithm to obtain multiple derived simulation scenario use cases; wherein the initial simulation scenario use case is a real high-risk scenario use case or an automatically defined scenario use case;

[0008] For the multiple derived simulation scenario use cases, performing scenario verification processing on the derived simulation scenario use cases based on the compliance verification condition of at least one dimension to obtain a derived scenario compliance attribute corresponding to each of the derived simulation scenario use cases;

[0009] Based on the derived scenario compliance attribute of each of the derived simulation scenario use cases, a target simulation scenario use case set is determined, and the generalized scenario use case set is stored in a target format.

[0010] In a second aspect, an embodiment of the present invention further provides an autonomous driving simulation scenario generalization device, the device comprising:

[0011] A random variable introduction module is used to add random variables to the initial simulation scenario use case based on a random sampling algorithm to obtain multiple derived simulation scenario use cases; wherein the initial simulation scenario use case is a real high-risk scenario use case or an automatically defined scenario use case;

[0012] A derived scenario verification module, configured to perform scenario verification processing on the multiple derived simulation scenario use cases based on compliance verification conditions of at least one dimension, and obtain a derived scenario compliance attribute corresponding to each of the derived simulation scenario use cases;

[0013] A simulation scenario storage module is used to determine a target simulation scenario use case set based on the derived scenario compliance attributes of each derived simulation scenario use case, and store the generalized scenario use case set in a target format.

[0014] In a third aspect, an embodiment of the present invention further provides an electronic device, the electronic device comprising:

[0015] one or more processors;

[0016] a storage device for storing one or more programs,

[0017] When one or more programs are executed by one or more processors, the one or more processors implement the generalization method of autonomous driving simulation scenarios as described in any of the embodiments of the present invention.

[0018] In a fourth aspect, an embodiment of the present invention further provides a storage medium comprising computer-executable instructions, which, when executed by a computer processor, are used to execute a method for generalizing an autonomous driving simulation scenario as described in any one of the embodiments of the present invention.

[0019] The technical solution of an embodiment of the present invention obtains multiple derived simulation scenario use cases by adding random variables to the initial simulation scenario use case based on a random sampling algorithm, wherein the initial simulation scenario use case is a real high-risk scenario use case or an automatically defined scenario use case, and then, for the multiple derived simulation scenario use cases, the derived simulation scenario use cases are subjected to scenario verification processing based on compliance verification conditions of at least one dimension to obtain derived scenario compliance attributes corresponding to each of the derived simulation scenario use cases, thereby determining the target simulation scenario use case set based on the derived scenario compliance attributes of each derived simulation scenario use case, and storing the generalized scenario use case set in a target format. The technical solution of this embodiment injects multi-dimensional random variables into the initial simulation scenario use case through a random sampling algorithm to achieve automated and batch scenario derivation, effectively solving the combinatorial explosion problem faced by the rule-driven method. At the same time, it intelligently screens the derived scenarios based on multiple compliance verifications, ensuring the diversity and compliance of the generated scenarios, and retaining and expanding the conflict characteristics of the original high-risk scenarios through hazard validity verification, overcoming the defect that traditional real data methods are difficult to generalize new hazard patterns. Ultimately, a scenario library with high coverage and high-risk scenario detection capabilities is formed, which improves the generation efficiency of autonomous driving scenarios, realizes the hazard pattern generalization capability of generated scenarios, and provides a more comprehensive and efficient testing and verification basis for autonomous driving systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings introduced here only illustrate some of the embodiments to be described by the present invention, and are not exhaustive. A person skilled in the art can derive other drawings based on these drawings without inventive effort.

[0021] Figure 1 Schematic diagram of the system structure for the autonomous driving system simulation platform test;

[0022] Figure 2 A schematic diagram of a flow chart of a method for generalizing an autonomous driving simulation scenario provided by an embodiment of the present invention;

[0023] Figure 3 A schematic flow chart of another method for generalizing an autonomous driving simulation scenario provided by an embodiment of the present invention;

[0024] Figure 4 A schematic diagram of the structure of an autonomous driving simulation scenario generalization device provided by an embodiment of the present invention;

[0025] Figure 5 The present invention provides a schematic structural diagram of an electronic device. DETAILED DESCRIPTION

[0026] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It will be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all structures.

[0027] It should be noted that similar reference numerals and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined or explained in subsequent figures. Furthermore, in the description of the present invention, the terms "first," "second," etc. are used only to distinguish descriptions and should not be understood to indicate or imply relative importance. The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0028] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned. They should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0029] The acquisition, storage, use, and processing of data in this application's technical solution comply with relevant national laws and regulations.

[0030] Below, the present application is further described in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0031] Before introducing this technical solution, we can first provide an example of an application scenario. This technical solution can be applied to scenarios where a small number of existing autonomous driving test scenario use cases are generalized to obtain more usable derivative simulation scenario use cases.

[0032] The technical solutions provided by the embodiments of this invention focus on improving the scenario generalization and efficiency of autonomous driving system simulation platform testing. Since simulation platform testing is indispensable in the production of autonomous driving systems, and simulation scenario generation methods are the core modules of simulation platform testing, improving their scenario generalization and efficiency is one of the most fundamental research directions in the field of autonomous driving.

[0033] Next, the system architecture of the autonomous driving system simulation platform test involved in the embodiment of the present invention can be exemplified. Figure 1 The system structure diagram for the autonomous driving system simulation platform test is as follows: Figure 1As shown in Figure 1, a simulation platform test for an autonomous driving system generally consists of three modules: a simulation scenario generation tool, a simulation platform, and the autonomous driving system. After the simulation scenario generation tool generates a simulation scenario using a specific method, it interacts with the simulation platform. The simulation platform then physically models and renders scene elements based on scene attribute parameters or scene actions, and transmits the simulation results (primarily signals from the host vehicle's sensors) to the autonomous driving system. The autonomous driving system processes the sensor signals through modules such as perception, prediction, planning, and control, and then returns control signals for the host vehicle to the simulation platform. The simulation platform updates the scene changes based on these control signals and renders them in real time, then returns the updated scene state to the simulation scenario generation tool. The simulation scenario generation tool can adjust the scene actions in real time based on the updated scene state, and again transmits the updated scene actions to the simulation platform for physical modeling and rendering, repeating this cycle.

[0034] like Figure 1 As shown in the simulation scenario generation tool in , the technical solution provided by the embodiment of the present invention can automatically expand the existing simulation scenarios and generate a large number of similar simulation scenarios by introducing a certain degree of randomness. It includes two parts. The first part is a custom simulation scenario based on experience or manual design, and the second part is a real high-risk scenario. After collecting these two parts of the scenario, random variables are introduced, and the scenarios are subjected to a large amount of generalization processing to generate a large number of derivative scenarios. Subsequently, the derived scenarios are intelligently screened based on multiple compliance verifications, which not only ensures the diversity and compliance of the generated scenarios, but also retains and expands the conflict characteristics of the original high-risk scenarios through hazard validity verification, overcoming the defect that the traditional real data method is difficult to generalize new hazard patterns, and finally forming a scenario library with both high coverage and high-risk scenario discovery capabilities.

[0035] Figure 2 This is a flow chart of a method for generalizing an autonomous driving simulation scenario provided by an embodiment of the present invention. This embodiment can be applied to scenarios where more available derivative simulation scenario use cases are obtained by generalizing a small number of existing autonomous driving test scenario use cases. The method can be executed by an autonomous driving simulation scenario generalization device, which can be implemented in the form of software and / or hardware. The hardware can be an electronic device, such as a mobile terminal, PC or server.

[0036] like Figure 2 As shown in FIG, the generalization method of the autonomous driving simulation scenario includes:

[0037] S110. Add random variables to the initial simulation scenario use case based on a random sampling algorithm to obtain multiple derived simulation scenario use cases.

[0038] Initial simulation scenario use cases are either real high-risk scenarios or automatically defined scenarios. Initial simulation scenario use cases refer to benchmark scenarios with specific conditions and parameters used in autonomous driving simulation testing. These can be accurately reproduced data models based on high-risk driving scenarios captured in the real world, such as complex traffic conflicts, extreme weather conditions, or unexpected accidents. Alternatively, they can be virtual scenarios manually customized based on experience or rules, such as vehicle interactions under specific traffic regulations or extreme edge cases. These initial simulation scenario use cases serve as the raw input for scenario generalization, providing a basic template and constraint framework for the subsequent generation of derived simulation scenario use cases through random sampling. Derived simulation scenario use cases are new test scenarios generated by perturbing the initial simulation scenario use cases with random variables. These use cases retain the core logic of the original scenario while introducing controllable randomness, thereby expanding test coverage and evaluating the robustness of the autonomous driving system under different boundary conditions. A random sampling algorithm, a method for extracting random values ​​from a given parameter space using a probability distribution or rule constraints, is used to dynamically generate derived scenarios in this simulation scenario generalization. It systematically expands the diversity of test scenarios by injecting random perturbations that conform to a preset distribution or logical rules into specific variables in the initial scenario. The core goal of this algorithm is to efficiently cover possible edge cases and long-tail scenarios while ensuring the rationality of the generated scenarios, thereby improving the generalization testing capabilities of autonomous driving systems. Random variables are variable parameters that are dynamically adjusted based on probability distributions or rule constraints during the generalization process of autonomous driving simulation scenarios. They are used to introduce controllable changes based on the initial scenario.

[0039] Specifically, during the specific application process, multiple initial simulation scenario use cases can be imported in batches from the autonomous driving test database, simulation platform or the enterprise's internal scenario management system as benchmark templates. Then, for the key parameters in the scenario, such as vehicle position, speed, pedestrian behavior, weather conditions, etc., the variable parameter range and its probability distribution are defined; then, a random sampling algorithm is used to extract random values ​​from these distributions, replace or superimpose them on the corresponding parameters of the initial simulation scenario use cases, and generate multiple derivative simulation scenario use cases with reasonable randomness.

[0040] Optionally, random variables include parameter information from at least one of the following dimensions: active vehicle dynamic parameters, ambient vehicle behavior characteristics, environmental condition parameters, and vulnerable road user behavior. Active vehicle dynamic parameters refer to the motion state variables of the autonomous vehicle under test (i.e., the active vehicle), including but not limited to dynamic indicators such as speed, acceleration, steering angle, and yaw rate. These parameters are used to simulate the active vehicle's behavioral changes under different driving strategies, such as sudden acceleration and emergency obstacle avoidance. Ambient vehicle behavior characteristics refer to the driving behavior parameters of other traffic participants in the scene, such as surrounding cars and trucks. These parameters include following distance, lane change frequency, expected path planning, and braking response time. These parameters are used to model complex traffic flow interactions, such as vehicle insertion and following vehicles in congestion. Environmental condition parameters describe the physical state of the simulated scene, including light intensity, weather, road adhesion coefficient, and visibility. These parameters are used to test the autonomous driving system's perception and decision-making capabilities in extreme or changing environments. Vulnerable road user behavior specifically refers to the motion characteristics of vulnerable traffic participants, such as pedestrians, bicycles, and electric scooters, such as walking speed, probability of sudden crossing, and trajectory randomness. These parameters are used to verify the system's ability to respond to long-tail scenarios.

[0041] On this basis, the random variables are the dynamic parameters of the main vehicle, and random variables are added to the initial simulation scenario use case based on the random sampling algorithm to obtain multiple derived simulation scenario use cases, including: randomly sampling within a first preset speed range based on the random sampling algorithm to determine at least one main vehicle initial speed; randomly sampling within the acceleration safety threshold range based on the random sampling algorithm to determine at least one main vehicle initial acceleration; randomly sampling within the deflection range allowed by the scene road based on the random sampling algorithm to determine at least one driving heading angle; based on at least one main vehicle initial speed, at least one main vehicle initial acceleration and / or at least one driving heading angle, determine multiple main vehicle parameter derived scenario use cases corresponding to the initial simulation scenario use case.

[0042] Among them, the main vehicle parameter derived scenario use case refers to a series of new test scenarios with differentiated motion characteristics generated by applying random perturbations that meet the preset range and safety constraints to the main vehicle dynamic parameters in the initial simulation scenario.

[0043] In this embodiment, the constraint range of the main vehicle dynamic parameters can be set based on the road type and traffic rules of the initial simulation scenario use case. Specifically, the constraint range can be: the initial speed of the main vehicle is within a reasonable range (i.e., a first preset speed range, such as 30-60 km / h on urban roads and 60-120 km / h on highways) through uniform distribution or normal distribution random sampling; the initial acceleration of the main vehicle is within the vehicle dynamics safety threshold (i.e., the safety threshold range, such as -4 m / s 2 Up to 2m / s 2) is sampled using a truncated Gaussian distribution to avoid extreme values; the driving heading angle is sampled randomly based on lane curvature and traffic regulations (i.e., the permissible deflection range of the scenario road, such as ±10°). Based on this, this step is specifically implemented by independently generating multiple sets of initial velocity, initial acceleration, and heading angle combinations using stratified random sampling or probability distribution sampling. Each set of parameters is then injected into the main vehicle model of the initial simulation scenario use case, generating multiple main vehicle parameter-derived scenario use cases with varying dynamic behaviors.

[0044] Optionally, the random variable is the behavior characteristic of the environmental vehicle, and the random variable is added to the initial simulation scenario use case based on the random sampling algorithm to obtain multiple derived simulation scenario use cases, including: random sampling within a preset range of the number of environmental vehicles based on the random sampling algorithm to determine the number of environmental vehicles; for each environmental vehicle, random sampling within a preset spatial range of the preset position of the main vehicle based on the random sampling algorithm to determine at least one initial position information corresponding to each environmental vehicle; random sampling within an offset threshold range relative to the main vehicle speed based on the random sampling algorithm to determine at least one initial speed corresponding to each environmental vehicle; random sampling within the acceleration range allowed by traffic regulations based on the random sampling algorithm to determine at least one acceleration value corresponding to each environmental vehicle; random sampling within a preset lane change threshold range based on the random sampling algorithm to determine at least one lane change frequency corresponding to each environmental vehicle; based on the environmental vehicle number information, at least one initial position information corresponding to each environmental vehicle, at least one initial speed, at least one acceleration value and at least one lane change frequency, determine multiple environmental vehicle derived scenario use cases corresponding to the initial simulation scenario use case.

[0045] Among them, the environmental vehicle derived scenario use case refers to a series of new test scenarios generated by performing multi-dimensional random adjustments to the behavioral characteristics of the surrounding vehicles, i.e., the environmental vehicles, in the initial simulation scenario use case.

[0046] In this embodiment, the adjustable parameter dimensions and their constraint ranges of the environmental vehicle behavior can be predefined, which may include: the number of vehicles (e.g., 1-5 vehicles), the initial position (based on the main vehicle position, set within a 50-meter range), the speed offset relative to the main vehicle speed (±20 km / h), the acceleration (0-2 m / s 2 ) and lane change frequency (0-1 times / minute). On this basis, stratified random sampling or constrained random algorithm is then used to generate parameter combinations that conform to the preset distribution for each dimension in turn. For example, first randomly determine that the scene contains 3 vehicles, and then generate the initial position (30 meters to the left front of the main vehicle's initial position), speed (main vehicle speed + 15km / h), acceleration (1.2m / s 2) and lane change probability (0.5 times / minute); finally, the randomized environmental vehicle behaviors are injected into the initial simulation scenario use case through the parameter combination engine. Finally, dozens to hundreds of derivative scenario use cases reflecting different traffic flow dynamic characteristics are output, that is, multiple environmental vehicle derivative scenario use cases.

[0047] Optionally, the random variable is an environmental condition parameter, and random variables are added to the initial simulation scenario use case based on a random sampling algorithm to obtain multiple derived simulation scenario use cases, including: segmented random sampling within a preset illumination range based on a random sampling algorithm to obtain at least one lighting condition information; determining sunny, rainy or foggy days according to a preset probability distribution, determining at least one weather type information, and associating random disturbances based on the weather type information to determine the road friction coefficient corresponding to different weather type information; based on at least one lighting condition information, and / or at least one weather type information and the corresponding road friction coefficient, determining multiple environmental condition derived scenario use cases corresponding to the initial simulation scenario use case.

[0048] Among them, the environmental condition-derived scenario use case refers to a series of diversified test scenarios generated by randomly perturbing the environmental parameters in the initial simulation scene, including light intensity, weather type, and the associated road friction coefficient.

[0049] In this embodiment, adjustable dimensions and constraint rules for environmental parameters can be established. Specifically, lighting conditions are sampled randomly in segments (such as daytime [30,000-100,000 lux], dusk [1,000-3,000 lux], and nighttime [5-30 lux]), and weather types are sampled polynomially according to preset probabilities, such as 60% sunny days, 30% rainy days, and 10% foggy days. Physical perturbation rules are associated, such as automatically triggering uniform sampling of the road friction coefficient between 0.4 and 0.6 on rainy days, and superimposing a Gaussian perturbation of visibility of 50-200 meters on foggy days. Subsequently, a compliant parameter group of lighting-weather-friction coefficient is generated through a combined sampling algorithm, and the parameters are injected into the environmental model of the initial scene. The light source angle, precipitation particle effect, and vehicle dynamics parameters are automatically adjusted, and finally multiple environmental condition-derived scenario use cases covering day and night alternation, extreme weather, and complex environmental effects are output.

[0050] Optionally, the random variable is the behavior of vulnerable road users, and random variables are added to the initial simulation scenario use case based on a random sampling algorithm to obtain multiple derived simulation scenario use cases, including: uniformly sampling within a second preset speed range based on a random sampling algorithm, and setting the pedestrian trajectory offset to obey the Brownian motion model for random sampling to obtain at least one pedestrian motion behavior information; randomly sampling according to the trigonometric function law to determine at least one non-motor vehicle steering angle; based on at least one pedestrian motion behavior information, and / or at least one non-motor vehicle steering angle, determine multiple user behavior derived scenario use cases corresponding to the initial simulation scenario use case.

[0051] The user behavior-derived scenario use cases are a series of diverse test scenarios generated by randomly modeling the behavioral characteristics of vulnerable road users, such as pedestrians and non-motorized vehicles, in the initial simulation scenario. While maintaining the host vehicle parameters and the underlying road environment, these use cases employ an algorithm to parametrically perturb pedestrian speed and trajectory, as well as non-motorized vehicle steering behavior. This creates a collection of test instances that retains the original scenario framework while also reflecting the behavioral uncertainty of vulnerable road users.

[0052] In this embodiment, randomization rules for behavioral parameters can be defined for vulnerable road users. Specifically, pedestrian speeds are uniformly sampled within a safe range (e.g., 0.5-2 m / s), and the motion trajectory is generated using a Brownian motion model with random offsets (e.g., lateral perturbations of 0-0.3 meters per step) to simulate the uncertainty of real walking. Non-motorized vehicle steering behavior uses trigonometric functions to generate periodic steering angles (±10°-±30°) to reproduce the swaying characteristics of cycling. The parameters are then injected into the initial simulation scenario use case through a combined sampling algorithm, for example, generating an interactive scenario of "pedestrians changing direction at a speed of 1.2 m / s with random direction changes + bicycles swinging sinusoidally at 20° amplitude." Ultimately, a set of user behavior-derived scenario use cases that reflect the randomness of vulnerable road user behavior and conform to biomechanical constraints are output, which are used to test the autonomous driving system's ability to respond to sudden behaviors of vulnerable traffic participants.

[0053] It should be noted that multiple main vehicle parameter derived scenario use cases, multiple environmental vehicle derived scenario use cases, multiple environmental condition derived scenario use cases, and multiple user behavior derived scenario use cases can also be randomly combined to generate more derived simulation scenario use cases.

[0054] S120. For multiple derived simulation scenario use cases, perform scenario verification processing on the derived simulation scenario use cases based on compliance verification conditions of at least one dimension to obtain derived scenario compliance attributes corresponding to each of the derived simulation scenario use cases.

[0055] Compliance verification conditions are a set of multi-dimensional validation rules used to evaluate the legality, rationality, and effectiveness of derived simulation scenario use cases. These conditions automatically verify derived scenarios using pre-set mathematical models, rule engines, or physical simulators, eliminating illegal or unrealistic scenarios such as "vehicle acceleration exceeding physical limits" or "pedestrians penetrating obstacles," ensuring that the generated test cases are both diverse and conform to real-world operating principles.

[0056] In this embodiment, at least one dimension of compliance verification conditions includes: parameter rationality verification conditions, traffic rule legality verification conditions, risk effectiveness verification conditions, and environmental consistency verification conditions. Parameter rationality verification conditions refer to the physical feasibility verification of all dynamic parameters in the scene, such as speed, acceleration, steering angle, etc., to ensure that they comply with vehicle dynamics constraints, such as the maximum acceleration does not exceed 5m / s 2 , biomechanical limitations, such as pedestrian speeds of 0-3m / s, and environmental physics laws, such as a friction coefficient greater than 0.3 in rainy weather. Traffic rule legality verification conditions verify the scenario's compliance with traffic regulations, including lane compliance, signal logic such as stopping at red lights, priority rules such as yielding to pedestrians, and prohibited behaviors such as driving against traffic or changing lanes on a solid line. Hazard validity verification conditions quantitatively assess the scenario's risk level, requiring the derived scenario to retain the core conflict characteristics of the initial high-risk scenario, such as a collision time less than 3 seconds, to avoid diluting dangerous scenarios into common scenarios due to excessive randomization. Environmental consistency verification conditions verify the logical relevance of each element in the scenario. For example, rainy weather requires simultaneous triggering of slippery road conditions, reduced visibility, and increased sensor noise to ensure that the coupling relationships between environmental parameters such as weather, lighting, and road conditions conform to real-world laws. These conditions, through multi-level verification, collectively guarantee the usability and test value of the derived scenarios.

[0057] The derived scenario compliance attribute refers to the structured evaluation results output after multi-dimensional verification of the derived simulation scenario use case through compliance verification conditions. It is used to quantitatively represent the legality and effectiveness of a scenario. This attribute can be represented by a binary flag (such as scenario compliance attribute / scenario non-compliance attribute) or a graded score (such as 90% compliance), and ultimately serves as the basis for scenario screening to ensure that the target scenario set meets the requirements of diversity, security, and authenticity.

[0058] Specifically, for these derivative simulation scenario use cases, the verification process of each derivative simulation scenario use case is consistent. In order to clearly introduce the present technical solution, one of the derivative simulation scenario use cases may be used as an example for introduction. Each derivative simulation scenario use case may be automatically verified in multiple dimensions through a preset compliance verification engine. For example, each derivative simulation scenario use case may be verified for physical rationality, verification of traffic rule legality, effectiveness of hazard level, and environmental consistency. The verification process uses a joint calculation of a rule engine, a physical simulator, and a risk model, and finally outputs a structured derivative scenario compliance attribute. For example, for a certain derivative simulation scenario use case, if the verification results corresponding to each dimension are all verification pass results, then the derivative scenario compliance attribute corresponding to the derivative simulation scenario use case is a scenario compliance attribute; if there is at least one dimension corresponding to a verification failure result, then the derivative scenario compliance attribute corresponding to the derivative simulation scenario use case is a scenario non-compliance attribute.

[0059] S130 . Determine a target simulation scenario use case set based on the derived scenario compliance attributes of each derived simulation scenario use case, and store the generalized scenario use case set in a target format.

[0060] Among them, the target simulation scenario use case set refers to a collection of high-quality derivative scenarios retained after compliance verification screening. These scenarios not only retain the core risk characteristics of the initial high-risk scenarios, but also cover a variety of boundary conditions through randomized expansion. Each use case in this set meets the verification requirements of physical rationality, traffic law legality, hazard effectiveness and environmental consistency. The target format refers to the standardized data storage specification adopted by the target simulation scenario use case set after compliance verification, which is usually a common scene description format in the field of autonomous driving, such as JSON, OpenSCENARIO, ASAMOpenDRIVE or XML and other structured formats. This format must meet the requirements of machine-readable, extensible, and compatible with mainstream simulation platforms. It can fully encode the road topology, traffic participant behavior, environmental parameters and time series of the scene, and support the rapid loading, parameter modification and batch testing of subsequent scenarios to ensure that the generalized scenario library can be directly integrated into the autonomous driving simulation test tool chain.

[0061] In this embodiment, a set of derived simulation scenario use cases whose derived scenario compliance attributes are scenario compliance attributes can be identified as the target simulation scenario use case set. Based on this, the scenarios that pass the screening can be subjected to standardized encoding conversion, serializing elements such as road structure, dynamic object trajectories, and environmental parameters according to the syntax and data structure of the target format. The final output is a set of scenario files that can be directly parsed by the simulation platform. A metadata database is also established to record the compliance attributes, risk characteristics, and parameter distribution of each scenario, forming a structured, traceable, and rapidly searchable autonomous driving test scenario library.

[0062] It can be understood that by storing the generalized scenario use case set, the recovery function from the historical simulation scenario use case file to the simulation platform interface class is realized, so that the simulation platform test can retest the failure scenarios that have occurred or the scenarios that have been tested.

[0063] The technical solution of an embodiment of the present invention obtains multiple derived simulation scenario use cases by adding random variables to the initial simulation scenario use case based on a random sampling algorithm, wherein the initial simulation scenario use case is a real high-risk scenario use case or an automatically defined scenario use case, and then, for the multiple derived simulation scenario use cases, the derived simulation scenario use cases are subjected to scenario verification processing based on compliance verification conditions of at least one dimension to obtain derived scenario compliance attributes corresponding to each of the derived simulation scenario use cases, thereby determining the target simulation scenario use case set based on the derived scenario compliance attributes of each derived simulation scenario use case, and storing the generalized scenario use case set in a target format. The technical solution of this embodiment injects multi-dimensional random variables into the initial simulation scenario use case through a random sampling algorithm to achieve automated and batch scenario derivation, effectively solving the combinatorial explosion problem faced by the rule-driven method. At the same time, it intelligently screens the derived scenarios based on multiple compliance verifications, ensuring the diversity and compliance of the generated scenarios, and retaining and expanding the conflict characteristics of the original high-risk scenarios through hazard validity verification, overcoming the defect that traditional real data methods are difficult to generalize new hazard patterns. Ultimately, a scenario library with high coverage and high-risk scenario detection capabilities is formed, which improves the generation efficiency of autonomous driving scenarios, realizes the hazard pattern generalization capability of generated scenarios, and provides a more comprehensive and efficient testing and verification basis for autonomous driving systems.

[0064] Example 2

[0065] Figure 3 This is a schematic diagram of a generalization method for autonomous driving simulation scenarios provided by an embodiment of the present invention. Based on the previous embodiment, S120 is further refined. For its specific implementation, please refer to the technical solution of this embodiment. Technical terms that are identical or corresponding to those in the previous embodiment are not repeated here.

[0066] like Figure 3 As shown, the method specifically includes the following steps:

[0067] S210. Add random variables to the initial simulation scenario use case based on a random sampling algorithm to obtain multiple derived simulation scenario use cases.

[0068] Among them, the initial simulation scenario use case is a real high-risk scenario use case or an automatically defined scenario use case.

[0069] S220. Based on the parameter rationality verification conditions, perform a physical feasibility check on the random variables in the derived simulation scenario use case to determine the parameter rationality properties.

[0070] The parameter plausibility attribute refers to a quantitative assessment of the physical feasibility of random variables in derived simulation scenario use cases, such as vehicle speed, acceleration, steering angle, and other dynamic parameters. This assessment determines whether these parameters conform to real-world physical laws and system constraints. This attribute can be represented by a binary flag (e.g., verification passed / failed) or a confidence score (e.g., a range of 0-1).

[0071] In this embodiment, the specific implementation of this step is as follows: the random variables in the derived scene are systematically checked through the preset physical rule engine. First, key dynamic parameters such as the maximum acceleration of the main vehicle, the speed of pedestrians, and the steering angular velocity of non-motor vehicles are extracted. Then, based on the vehicle dynamics model, such as the acceleration threshold of ±3m / s 2 , biokinematic constraints, such as pedestrian cadence of 0.5-2Hz and environmental physical limits, such as centrifugal force not exceeding tire grip, build a judgment rule library, and use a combination of real-time numerical analysis and boundary checking to output structured parameter rationality properties.

[0072] S230. Based on the traffic rule legitimacy verification conditions, the legitimacy of the derived simulation scenario use case is detected frame by frame by a preset traffic rule engine to determine the legal attributes of the traffic rule.

[0073] The pre-built traffic rule engine refers to a pre-built automated rule verification system with a built-in knowledge base of traffic regulations, such as signal light logic, lane keeping requirements, and priority rules. It determines legality by analyzing the behavior of all traffic participants in the scene frame by frame, such as lane change timing, stopping distance, and turn signal use. Traffic rule legality attributes refer to the compliance assessment results generated by the pre-built traffic rule engine through frame-by-frame analysis of derived simulation scenario use cases. These results are used to determine whether the behavior of all traffic participants in the scene complies with traffic regulations in the target area.

[0074] In this embodiment, derived scenarios can be analyzed sequentially using a pre-set traffic rules engine's built-in knowledge base of traffic signal logic, lane management rules, right-of-way priorities, and other regulations. Specifically, the behavior of all traffic participants is detected frame by frame, including whether the vehicle obeys traffic lights (such as stop at red lights and go at green lights), whether it complies with lane marking constraints (such as prohibiting lane changes on solid lines), and whether it violates yielding rules (such as failing to yield to pedestrians or priority vehicles). At the same time, context-related judgments are made based on the scene's geographic features, such as intersection type and speed limit zones, ultimately outputting structured traffic rule legality attributes. Scene frames with continuous or serious violations can also be automatically marked, ensuring that derived scenarios strictly comply with the rigid constraints of traffic regulations while expanding diversity.

[0075] S240. Conduct dynamic interactive tests on the derived simulation scenario use cases and calculate safety indicators. Determine the hazard effectiveness attributes based on the hazard effectiveness verification conditions and safety indicators.

[0076] Dynamic interaction testing involves using a simulation platform to simulate the time-series behavior and conduct conflict analysis of traffic participants in derived scenarios. This testing runs the entire scenario's time series, calculating and recording key safety indicators in real time. It also monitors interactions between multiple agents, such as changes in following distance, lane change conflicts, and pedestrian avoidance. This allows for a quantitative assessment of the scenario's riskiness and the autonomous driving system's response capabilities, providing data support for risk assessment validation. Safety indicators are a series of parameters used to quantitatively assess the riskiness of derived simulation scenarios through dynamic interaction testing. Examples include conflict time indicators, such as time to collision and time to cross; spatial distance indicators, such as minimum safe separation distance and braking distance margin; and behavioral risk indicators, such as frequency of sudden acceleration and deceleration and lane deviation. These indicators are dynamically generated through mathematical modeling and real-time simulation data, objectively characterizing the riskiness of interactions between traffic participants in the scenario and providing a data-based basis for verifying whether the derived scenarios retain the initial high-risk characteristics.

[0077] In this embodiment, the complete time sequence of the derived scenario is run through a high-precision simulation engine, dynamic interaction data between traffic participants is collected in real time, and safety indicators such as collision time, crossing time, and deceleration conflict rate are calculated based on a preset algorithm; these safety indicators are then graded and evaluated through hazard effectiveness verification conditions, and finally a structured hazard effectiveness attribute is output.

[0078] S250. Based on the environmental consistency verification conditions, perform a physical rationality check on the random variable combination in the derived simulation scenario use case to determine the environmental consistency attribute.

[0079] Among them, the environmental consistency attribute refers to the compliance assessment result generated by performing physical logical verification on the combination relationship of various environmental parameters in the derived simulation scenario use case, which is used to determine whether the coupling relationship between environmental elements conforms to the physical laws of the real world.

[0080] In this example, a systematic logical check is performed on random environmental variable combinations in derived scenarios by constructing a rule base for coupling environmental parameters, such as "Rainy day → Road friction coefficient 0.4-0.6 + Visibility 50-200 meters + Sensor noise increased by 20dB." Key parameters are first extracted, and their physical relevance is verified. For example, whether snowy days trigger icing on low-temperature roads or whether foggy days match the diffraction halo effect. The physics engine then simulates parameter interactions, such as the effect of heavy rain on LiDAR point cloud attenuation, ultimately outputting structured environmental consistency attributes.

[0081] S260. Based on parameter rationality attributes, traffic rule legality attributes, hazard validity attributes, and environmental consistency attributes, determine the derived scenario compliance attributes corresponding to the derived simulation scenario use case.

[0082] In this embodiment, a multi-dimensional decision engine can be used to comprehensively assess the compliance of derived scenarios. First, it integrates four types of verification results: parameter rationality, traffic rule legality, hazard effectiveness, and environmental consistency. Then, it performs quantitative scoring based on pre-set weighting rules. Finally, it generates a structured derivative scenario compliance attribute.

[0083] S270. Determine a target simulation scenario use case set based on the derived scenario compliance attributes of each derived simulation scenario use case, and store the generalized scenario use case set in a target format.

[0084] The technical solution of the embodiment of the present invention can firstly perform a physical feasibility check on the random variables in the derived simulation scenario use case based on the parameter rationality verification condition to determine the parameter rationality property, and then based on the traffic rule legality verification condition, detect the legality of the derived simulation scenario use case frame by frame through the preset traffic rule engine to determine the traffic rule legality property, and further, conduct a dynamic interactive test on the derived simulation scenario use case and calculate the safety index, and determine the hazard validity property based on the hazard validity verification condition and the safety index, and then, based on the environmental consistency verification condition, perform a physical rationality check on the random variable combination in the derived simulation scenario use case to determine the environmental consistency property, thereby determining the derived scenario compliance property corresponding to the derived simulation scenario use case based on the parameter rationality property, the traffic rule legality property, the hazard validity property and the environmental consistency property. The technical solution of the embodiment of the present invention significantly improves the reliability of generating generalized scenarios through a four-layer progressive verification mechanism.

[0085] Example 3

[0086] Figure 4 This is a structural diagram of an autonomous driving simulation scenario generalization device provided by an embodiment of the present invention, which includes: a random variable introduction module 310, a derived scenario verification module 320, and a simulation scenario storage module 330.

[0087] The random variable introduction module 310 is used to add random variables to the initial simulation scenario use case based on a random sampling algorithm to obtain multiple derived simulation scenario use cases; wherein the initial simulation scenario use case is a real high-risk scenario use case or an automatically defined scenario use case;

[0088] A derived scenario verification module 320 is configured to perform scenario verification processing on the multiple derived simulation scenario use cases based on the compliance verification condition of at least one dimension to obtain a derived scenario compliance attribute corresponding to each of the derived simulation scenario use cases;

[0089] The simulation scenario storage module 330 is used to determine a target simulation scenario use case set based on the derived scenario compliance attribute of each of the derived simulation scenario use cases, and store the generalized scenario use case set in a target format.

[0090] The technical solution of an embodiment of the present invention obtains multiple derived simulation scenario use cases by adding random variables to the initial simulation scenario use case based on a random sampling algorithm, wherein the initial simulation scenario use case is a real high-risk scenario use case or an automatically defined scenario use case, and then, for the multiple derived simulation scenario use cases, the derived simulation scenario use cases are subjected to scenario verification processing based on compliance verification conditions of at least one dimension to obtain derived scenario compliance attributes corresponding to each of the derived simulation scenario use cases, thereby determining the target simulation scenario use case set based on the derived scenario compliance attributes of each derived simulation scenario use case, and storing the generalized scenario use case set in a target format. The technical solution of this embodiment injects multi-dimensional random variables into the initial simulation scenario use case through a random sampling algorithm to achieve automated and batch scenario derivation, effectively solving the combinatorial explosion problem faced by the rule-driven method. At the same time, it intelligently screens the derived scenarios based on multiple compliance verifications, ensuring the diversity and compliance of the generated scenarios, and retaining and expanding the conflict characteristics of the original high-risk scenarios through hazard validity verification, overcoming the defect that traditional real data methods are difficult to generalize new hazard patterns. Ultimately, a scenario library with high coverage and high-risk scenario detection capabilities is formed, which improves the generation efficiency of autonomous driving scenarios, realizes the hazard pattern generalization capability of generated scenarios, and provides a more comprehensive and efficient testing and verification basis for autonomous driving systems.

[0091] Based on the above device, optionally, the random variables include parameter information of at least one dimension of the following: dynamic parameters of the main vehicle, behavior characteristics of environmental vehicles, environmental condition parameters, and behavior of vulnerable road users.

[0092] Based on the above-mentioned device, optionally, the random variable is a dynamic parameter of the main vehicle, and the random variable introduction module is specifically used to determine at least one initial speed of the main vehicle by random sampling within a first preset speed range based on a random sampling algorithm; determine at least one initial acceleration of the main vehicle by random sampling within an acceleration safety threshold range based on a random sampling algorithm; determine at least one driving heading angle by random sampling within a deflection range allowed by the scene road based on a random sampling algorithm; and determine multiple main vehicle parameter-derived scenario use cases corresponding to the initial simulation scenario use case based on the at least one initial speed of the main vehicle, the at least one initial acceleration of the main vehicle and / or the at least one driving heading angle.

[0093] Based on the above-mentioned device, optionally, the random variable is the environmental vehicle behavior characteristic, and the random variable introduction module is specifically used to randomly sample within a preset environmental vehicle number range based on a random sampling algorithm to determine the environmental vehicle number information; for each environmental vehicle, randomly sample within a preset spatial range of a preset position of the main vehicle based on a random sampling algorithm to determine at least one initial position information corresponding to each of the environmental vehicles; randomly sample within an offset threshold range relative to the main vehicle speed based on a random sampling algorithm to determine at least one initial speed corresponding to each of the environmental vehicles; randomly sample within an acceleration range allowed by traffic regulations based on a random sampling algorithm to determine at least one acceleration value corresponding to each of the environmental vehicles; randomly sample within a preset lane change threshold range based on a random sampling algorithm to determine at least one lane change frequency corresponding to each of the environmental vehicles; based on the environmental vehicle number information, the at least one initial position information corresponding to each of the environmental vehicles, the at least one initial speed, the at least one acceleration value and the at least one lane change frequency, determine multiple environmental vehicle derived scenario use cases corresponding to the initial simulation scenario use case.

[0094] Based on the above-mentioned device, optionally, the random variable is an environmental condition parameter, and the random variable introduction module is specifically used to perform segmented random sampling within a preset illumination range based on a random sampling algorithm to obtain at least one lighting condition information; determine sunny, rainy or foggy days according to a preset probability distribution, determine at least one weather type information, and associate random disturbances with the weather type information to determine the road friction coefficient corresponding to different weather type information; based on the at least one lighting condition information, and / or the at least one weather type information and the corresponding road friction coefficient, determine multiple environmental condition derived scenario use cases corresponding to the initial simulation scenario use case.

[0095] Based on the above-mentioned device, optionally, the random variable is the behavior of the vulnerable road user, and the random variable introduction module is specifically used to uniformly sample within a second preset speed range based on a random sampling algorithm, and set the pedestrian trajectory offset to obey the Brownian motion model for random sampling to obtain at least one pedestrian motion behavior information; randomly sample according to the trigonometric function law to determine at least one non-motor vehicle steering angle; based on the at least one pedestrian motion behavior information, and / or the at least one non-motor vehicle steering angle, determine multiple user behavior derived scenario use cases corresponding to the initial simulation scenario use case.

[0096] Based on the above device, optionally, the compliance verification conditions based on at least one dimension include: parameter rationality verification conditions, traffic rule legality verification conditions, hazard validity verification conditions and environmental consistency verification conditions.

[0097] Based on the above device, optionally, the derived scenario verification module 320 includes:

[0098] A parameter rationality verification unit, configured to perform a physical feasibility check on the random variables in the derived simulation scenario use case based on parameter rationality verification conditions to determine parameter rationality properties;

[0099] A traffic rule legality verification unit is used to detect the legality of the derived simulation scenario use case frame by frame through a preset traffic rule engine based on the traffic rule legality verification conditions to determine the legality attributes of the traffic rule;

[0100] a risk degree validity verification unit, configured to perform a dynamic interactive test on the derived simulation scenario use case and calculate a safety index, and determine a risk degree validity attribute based on the risk degree validity verification condition and the safety index;

[0101] An environment consistency verification unit, configured to perform a physical rationality check on the random variable combination in the derived simulation scenario use case based on the environment consistency verification condition, and determine the environment consistency attribute;

[0102] A compliance attribute determination unit is used to determine the derivative scenario compliance attribute corresponding to the derived simulation scenario use case based on the parameter rationality attribute, traffic rule legality attribute, hazard validity attribute and environment consistency attribute.

[0103] The autonomous driving simulation scene generalization device provided in an embodiment of the present invention can execute the autonomous driving simulation scene generalization method provided in any embodiment of the present invention, and has the corresponding functional model and beneficial effects of the execution method.

[0104] It is worth noting that the various units and models included in the above system are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the embodiments of the present invention.

[0105] Example 4

[0106] Figure 5 The present invention provides a schematic structural diagram of an electronic device. Figure 5 A block diagram of an exemplary electronic device 40 suitable for implementing exemplary embodiments of the present invention is shown. Figure 5 The electronic device 40 shown is only an example and should not limit the functionality and scope of use of the embodiments of the present invention.

[0107] like Figure 5 As shown, electronic device 40 is a general-purpose computing device. Components of electronic device 40 may include, but are not limited to, one or more processors or processing units 401, system memory 402, and a bus 403 connecting various system components (including system memory 402 and processing unit 401).

[0108] Bus 403 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0109] The electronic device 40 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the electronic device 40, including volatile and non-volatile media, removable and non-removable media.

[0110] System memory 402 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 404 and / or cache memory 405. Electronic device 40 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 406 may be used to read and write non-removable, non-volatile magnetic media ( Figure 5 Not shown, often called a "hard drive"). Although Figure 5Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 403 via one or more data medium interfaces. Memory 402 may include at least one program product having a set (e.g., at least one) of program models configured to perform the functions of various embodiments of the present invention.

[0111] A program / utility 408 having a set (at least one) of program models 407 may be stored, for example, in memory 402. Such program models 407 include, but are not limited to, an operating system, one or more application programs, other program models, and program data, each of which, or some combination thereof, may include an implementation of a network environment. Program models 407 generally implement the functions and / or methods of the embodiments described herein.

[0112] The electronic device 40 may also communicate with one or more external devices 409 (e.g., keyboard, pointing device, display 810, etc.), and may also communicate with one or more devices that enable a user to interact with the electronic device 40, and / or any device that enables the electronic device 40 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed through an input / output (I / O) interface 411. Furthermore, the electronic device 40 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 412. As shown, the network adapter 412 communicates with other models of the electronic device 40 via the bus 403. It should be understood that although Figure 5 Not shown, other hardware and / or software models may be used in conjunction with the electronic device 40, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0113] The processing unit 401 executes various functional applications and page processing by running the programs stored in the system memory 402, such as implementing the autonomous driving simulation scenario generalization method provided in an embodiment of the present invention.

[0114] Example 5

[0115] An embodiment of the present invention further provides a storage medium containing computer-executable instructions. When executed by a computer processor, the computer-executable instructions are used to perform a method for generalizing an autonomous driving simulation scenario. The method includes:

[0116] Adding random variables to the initial simulation scenario use case based on a random sampling algorithm to obtain multiple derived simulation scenario use cases; wherein the initial simulation scenario use case is a real high-risk scenario use case or an automatically defined scenario use case;

[0117] For the multiple derived simulation scenario use cases, performing scenario verification processing on the derived simulation scenario use cases based on the compliance verification condition of at least one dimension to obtain a derived scenario compliance attribute corresponding to each of the derived simulation scenario use cases;

[0118] Based on the derived scenario compliance attribute of each of the derived simulation scenario use cases, a target simulation scenario use case set is determined, and the generalized scenario use case set is stored in a target format.

[0119] The computer storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device.

[0120] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0121] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0122] Computer program code for performing the operations of embodiments of the present invention can be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0123] Note that the above are only preferred embodiments of the present invention and the technical principles employed. Those skilled in the art will appreciate that the present invention is not limited to the specific embodiments herein, and that various obvious changes, readjustments, and substitutions are possible for those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments and may include many other equivalent embodiments without departing from the scope of the present invention. The scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for generalizing autonomous driving simulation scenarios, characterized in that: include: Adding random variables to the initial simulation scenario use case based on a random sampling algorithm to obtain multiple derived simulation scenario use cases; wherein the initial simulation scenario use case is a real high-risk scenario use case or an automatically defined scenario use case; For the multiple derived simulation scenario use cases, performing scenario verification processing on the derived simulation scenario use cases based on the compliance verification condition of at least one dimension to obtain a derived scenario compliance attribute corresponding to each of the derived simulation scenario use cases; Based on the derived scenario compliance attribute of each of the derived simulation scenario use cases, a target simulation scenario use case set is determined, and the generalized scenario use case set is stored in a target format.

2. The method according to claim 1, characterized in that The random variables include parameter information of at least one dimension of the following: dynamic parameters of the host vehicle, behavioral characteristics of the surrounding vehicles, environmental condition parameters, and behavior of vulnerable road users.

3. The method according to claim 2, characterized in that The random variables are dynamic parameters of the main vehicle, and the random variables are added to the initial simulation scenario use case based on the random sampling algorithm to obtain multiple derived simulation scenario use cases, including: Determine at least one initial speed of the host vehicle by randomly sampling within a first preset speed range based on a random sampling algorithm; Determine at least one initial acceleration of the host vehicle by randomly sampling within an acceleration safety threshold based on a random sampling algorithm; Random sampling is performed within the allowed deflection range of the scene road based on a random sampling algorithm to determine at least one driving heading angle; Based on the at least one main vehicle initial speed, the at least one main vehicle initial acceleration and / or the at least one driving heading angle, a plurality of main vehicle parameter derived scenario use cases corresponding to the initial simulation scenario use case are determined.

4. The method according to claim 2, characterized in that The random variables are environmental vehicle behavior characteristics. The random sampling algorithm is used to add random variables to the initial simulation scenario use case to obtain multiple derived simulation scenario use cases, including: Based on the random sampling algorithm, random sampling is performed within a preset range of the number of environmental vehicles to determine the number of environmental vehicles; For each environment vehicle, randomly sampling within a preset spatial range of a preset position of the host vehicle based on a random sampling algorithm to determine at least one initial position information corresponding to each of the environment vehicles; Determine at least one initial speed corresponding to each of the environment vehicles by randomly sampling within a threshold range of an offset relative to a speed of the host vehicle based on a random sampling algorithm; Randomly sampling within the acceleration range permitted by traffic regulations based on a random sampling algorithm, and determining at least one acceleration value corresponding to each of the environment vehicles; Determine at least one lane change frequency corresponding to each of the environment vehicles by random sampling within a preset lane change threshold range based on a random sampling algorithm; Based on the environmental vehicle quantity information, the at least one initial position information corresponding to each of the environmental vehicles, the at least one initial speed, the at least one acceleration value, and the at least one lane change frequency, multiple environmental vehicle derived scenario use cases corresponding to the initial simulation scenario use case are determined.

5. The method according to claim 2, characterized in that The random variables are environmental condition parameters. The random sampling algorithm is used to add random variables to the initial simulation scenario use case to obtain multiple derived simulation scenario use cases, including: Based on a random sampling algorithm, random sampling is performed in sections within a preset illumination range to obtain at least one piece of lighting condition information; Determine sunny, rainy, or foggy weather according to a preset probability distribution, determine at least one weather type information, and associate random disturbances with the weather type information to determine a road friction coefficient corresponding to the different weather type information; Based on the at least one lighting condition information, and / or the at least one weather type information and the corresponding road friction coefficient, multiple environmental condition derived scenario use cases corresponding to the initial simulation scenario use case are determined.

6. The method according to claim 2, characterized in that The random variable is the vulnerable road user behavior. The random sampling algorithm is used to add random variables to the initial simulation scenario use case to obtain multiple derived simulation scenario use cases, including: uniformly sampling within a second preset speed range based on a random sampling algorithm, and setting the pedestrian trajectory offset to obey the Brownian motion model for random sampling to obtain at least one pedestrian motion behavior information; Random sampling according to trigonometric function rules to determine at least one non-motor vehicle steering angle; Based on the at least one pedestrian motion behavior information and / or the at least one non-motor vehicle steering angle, multiple user behavior derived scenario use cases corresponding to the initial simulation scenario use case are determined.

7. The method according to claim 1, characterized in that The compliance verification conditions of the at least one dimension include: parameter rationality verification conditions, traffic rule legality verification conditions, hazard validity verification conditions and environmental consistency verification conditions.

8. The method according to claim 7, characterized in that The compliance verification condition based on at least one dimension performs scenario verification processing on the derived simulation scenario use case to obtain a derived scenario compliance attribute corresponding to each of the derived simulation scenario use cases, including: Based on the parameter rationality verification conditions, the physical feasibility check is performed on the random variables in the derived simulation scenario use case to determine the parameter rationality properties; Based on the traffic rule legitimacy verification conditions, the legitimacy of the derived simulation scenario use case is detected frame by frame by a preset traffic rule engine to determine the legal attributes of the traffic rule; Performing a dynamic interactive test on the derived simulation scenario use case and calculating a safety index, and determining a risk effectiveness attribute based on the risk effectiveness verification condition and the safety index; Based on the environmental consistency verification conditions, a physical rationality check is performed on the random variable combination in the derived simulation scenario use case to determine the environmental consistency attribute; Based on the parameter rationality attribute, traffic rule legality attribute, hazard validity attribute and environment consistency attribute, the derived scenario compliance attribute corresponding to the derived simulation scenario use case is determined.

9. An autonomous driving simulation scenario generalization device, characterized in that: include: A random variable introduction module is used to add random variables to the initial simulation scenario use case based on a random sampling algorithm to obtain multiple derived simulation scenario use cases; wherein the initial simulation scenario use case is a real high-risk scenario use case or an automatically defined scenario use case; A derived scenario verification module, configured to perform scenario verification processing on the multiple derived simulation scenario use cases based on compliance verification conditions of at least one dimension, and obtain a derived scenario compliance attribute corresponding to each of the derived simulation scenario use cases; A simulation scenario storage module is used to determine a target simulation scenario use case set based on the derived scenario compliance attributes of each of the derived simulation scenario use cases, and store the generalized scenario use case set in a target format.

10. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the autonomous driving simulation scenario generalization method described in any one of claims 1 to 8.

Images