Test method, device and electronic equipment of vehicle auxiliary driving system

Abstract

This invention discloses a testing method, apparatus, and electronic device for a vehicle assisted driving system. The method includes: acquiring a preset traffic operation dataset, wherein the preset traffic operation dataset contains a sequence of driving condition parameters for multiple vehicles, the driving condition parameter sequence being driving condition parameters for multiple vehicles in multiple consecutive time frames while driving on the road; determining at least two target parameter types from multiple initial parameter types corresponding to the driving condition parameters, according to their importance in testing the assisted driving system installed on the multiple vehicles; extracting information from the preset traffic operation dataset based on the at least two target parameter types to obtain multiple target test scenarios; and performing simulation testing on the assisted driving system based on the multiple target test scenarios to obtain simulation test results for the assisted driving system. This invention solves the technical problem of low testing efficiency in related technologies for testing assisted driving systems.

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and more specifically, to a test method, apparatus, and electronic device for a vehicle driver assistance system. Background Technology

[0002] Currently, testing of vehicle driver assistance systems relies on a combination of real-world road data and simulation scenarios. However, typical scenarios in actual traffic are sparse and unevenly distributed, resulting in insufficient test coverage. Many related methods employ manually designed or randomly generated scenarios, leading to simulation scenarios that lack engineering representativeness. The parametric modeling approach introduces a large number of low-correlation variables, increasing computational redundancy and obscuring the influencing factors of driver assistance system decisions, resulting in low testing efficiency for driver assistance systems in related technologies.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a testing method, apparatus, and electronic device for a vehicle driver assistance system, to at least solve the technical problem of low testing efficiency in related technologies for testing driver assistance systems.

[0005] According to one aspect of the present invention, a testing method for a vehicle assisted driving system is provided, comprising: acquiring a preset traffic operation dataset, wherein the preset traffic operation dataset contains a sequence of driving condition parameters for multiple vehicles, the sequence of driving condition parameters being driving condition parameters of multiple vehicles in multiple consecutive time frames while driving on a road; determining at least two target parameter types from multiple initial parameter types corresponding to the driving condition parameters, according to the importance of their impact on testing the assisted driving system installed on the multiple vehicles; extracting information from the preset traffic operation dataset based on the at least two target parameter types to obtain multiple target test scenarios; and performing simulation testing on the assisted driving system based on the multiple target test scenarios to obtain simulation test results of the assisted driving system.

[0006] In this embodiment of the invention, from multiple initial parameter types corresponding to driving condition parameters, at least two target parameter types are determined according to the importance of their impact on testing the assisted driving systems installed on multiple vehicles. This includes: using an ensemble learning model to determine the importance scores corresponding to multiple initial parameter types according to the importance of their impact on testing the assisted driving systems; and determining at least two target parameter types based on the importance scores corresponding to multiple initial parameter types.

[0007] In this embodiment of the invention, determining at least two target parameter types based on the importance scores corresponding to multiple initial parameter types includes: sorting the multiple initial parameter types in descending order based on the importance scores corresponding to multiple initial parameter types to obtain an importance ranking queue; and selecting the first preset number of parameter types in the importance ranking queue to obtain at least two target parameter types.

[0008] In this embodiment of the invention, information is extracted from a preset traffic operation dataset based on at least two target parameter types to obtain multiple target test scenarios, including: constructing a joint probability density function corresponding to at least two target parameter types, wherein the joint probability density function is used to characterize the cooperative distribution characteristics of at least two target parameter types in real traffic operation scenarios; and extracting test scenarios from the preset traffic operation dataset based on the joint probability density function to obtain multiple target test scenarios.

[0009] In this embodiment of the invention, test scenarios are extracted from a preset traffic operation dataset based on a joint probability density function to obtain multiple target test scenarios, including: extracting test scenarios from the preset traffic operation dataset based on a state acceptance mechanism sampling algorithm and a joint probability density function to obtain at least one first test scenario; extracting test scenarios from the preset traffic operation dataset based on a weighted augmentation sampling algorithm and a joint probability density function to obtain at least one second test scenario; and integrating at least one first test scenario and at least one second test scenario to obtain multiple target test scenarios.

[0010] In this embodiment of the invention, based on the state acceptance mechanism sampling algorithm and the joint probability density function, information test scenarios are extracted from a preset traffic operation dataset to obtain at least one first test scenario, including: constructing a scenario rationality evaluation function corresponding to the state acceptance mechanism sampling algorithm based on the joint probability density function; and extracting test scenarios from the preset traffic operation dataset based on the state acceptance mechanism sampling algorithm and the scenario rationality evaluation function to obtain at least one first test scenario.

[0011] In this embodiment of the invention, test scenarios are extracted from a preset traffic operation dataset based on a weighted augmented sampling algorithm and a joint probability density function to obtain at least one second test scenario. This includes: constructing importance sampling weights corresponding to the weighted augmented sampling algorithm based on the joint probability density function; and extracting test scenarios from the preset traffic operation dataset based on the weighted augmented sampling algorithm and the importance sampling weights to obtain at least one second test scenario.

[0012] According to another aspect of the present invention, a testing apparatus for a vehicle assisted driving system is also provided, comprising: an acquisition module for acquiring a preset traffic operation dataset, wherein the preset traffic operation dataset contains a sequence of driving condition parameters for multiple vehicles, the sequence of driving condition parameters being driving condition parameters in multiple consecutive time frames of multiple vehicles driving on a road; a determination module for determining at least two target parameter types from multiple initial parameter types corresponding to the driving condition parameters, according to the importance of their impact on testing the assisted driving system installed on the multiple vehicles; an information extraction module for extracting test scenarios from the preset traffic operation dataset based on the at least two target parameter types, to obtain multiple target test scenarios; and a simulation testing module for performing simulation testing on the assisted driving system based on the multiple target test scenarios, to obtain simulation test results of the assisted driving system.

[0013] The determination module is used to determine the importance scores corresponding to multiple initial parameter types according to the importance of their impact on the testing of the assisted driving system using an ensemble learning model; and to determine at least two target parameter types based on the importance scores corresponding to the multiple initial parameter types.

[0014] The determination module is used to sort multiple initial parameter types in descending order based on their importance scores to obtain an importance ranking queue; and to select the first preset number of parameter types from the importance ranking queue to obtain at least two target parameter types.

[0015] The information extraction module is used to construct a joint probability density function corresponding to at least two target parameter types. The joint probability density function is used to characterize the collaborative distribution characteristics of at least two target parameter types in real traffic operation scenarios. Based on the joint probability density function, test scenarios are extracted from a preset traffic operation dataset to obtain multiple target test scenarios.

[0016] The information extraction module is used to extract test scenarios from a preset traffic operation dataset based on a state acceptance mechanism sampling algorithm and a joint probability density function to obtain at least one first test scenario; to extract test scenarios from the preset traffic operation dataset based on a weighted augmentation sampling algorithm and a joint probability density function to obtain at least one second test scenario; and to integrate at least one first test scenario and at least one second test scenario to obtain multiple target test scenarios.

[0017] The information extraction module is used to construct a scenario rationality evaluation function corresponding to the state acceptance mechanism sampling algorithm based on the joint probability density function; and to extract test scenarios from the preset traffic operation dataset based on the state acceptance mechanism sampling algorithm and the scenario rationality evaluation function to obtain at least one first test scenario.

[0018] The information extraction module is used to construct the importance sampling weights corresponding to the weighted augmented sampling algorithm based on the joint probability density function; and to extract test scenarios from the preset traffic operation dataset based on the weighted augmented sampling algorithm and the importance sampling weights to obtain at least one second test scenario.

[0019] According to another aspect of the present invention, an electronic device is also provided, comprising: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods of various embodiments of the present invention during runtime.

[0020] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is executed, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of the present invention.

[0021] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the methods of various embodiments of the present invention.

[0022] According to another aspect of the present invention, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods of various embodiments of the present invention.

[0023] According to another aspect of the present invention, a computer program is also provided, which, when executed by a processor, implements the methods of the various embodiments of the present invention.

[0024] In this embodiment of the invention, firstly, a preset traffic operation dataset is acquired; then, from multiple initial parameter types corresponding to driving condition parameters, at least two target parameter types are determined according to their importance in testing the assisted driving systems installed on multiple vehicles; next, based on the at least two target parameter types, information is extracted from the preset traffic operation dataset to obtain multiple target test scenarios; finally, based on the multiple target test scenarios, simulation tests are performed on the assisted driving system to obtain the simulation test results of the assisted driving system. This application constructs multiple target test scenarios based on a preset traffic operation dataset, ensuring that the data used truly reflects the dynamic characteristics of vehicles in actual traffic operations and avoiding biases introduced by manual simulation. By determining at least two target parameter types based on the importance of testing the driver assistance systems installed on multiple vehicles, the efficiency and relevance of test scenario generation can be improved. Only variables that have a greater impact on the testing of driver assistance systems are retained, eliminating noise and computational burden introduced by redundant parameters. This enhances the correlation between the generated scenarios and the real response mechanism of the driver assistance systems. Based on the obtained multiple target test scenarios, simulation tests are conducted on the driver assistance systems. The simulation test results reflect the decision robustness and action effectiveness of the driver assistance systems under real-world scenario distributions, providing quantifiable and traceable empirical evidence for the improvement of driver assistance systems. This solves the technical problem of low testing efficiency for driver assistance systems in related technologies. Attached Figure Description

[0025] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0026] Figure 1 This is a flowchart of a test method for a vehicle driver assistance system according to an embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of a test process for an optional driver assistance system according to an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of the testing process for another optional driver assistance system according to an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of the testing process for another optional driver assistance system according to an embodiment of the present invention;

[0030] Figure 5 This is a schematic diagram of the testing process for another optional driver assistance system according to an embodiment of the present invention;

[0031] Figure 6This is a schematic diagram of a test apparatus for a vehicle driver assistance system according to an embodiment of the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] According to one aspect of the present invention, a test method for a vehicle-assisted driving system is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0035] Figure 1 This is a flowchart of a test method for a vehicle driver assistance system according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0036] Step S102: Obtain the preset traffic operation dataset.

[0037] The preset traffic operation dataset contains a sequence of driving condition parameters for multiple vehicles. The sequence of driving condition parameters consists of driving condition parameters for multiple vehicles in multiple consecutive time frames while they are driving on the road.

[0038] The aforementioned preset traffic operation dataset refers to a set of structured traffic operation data collected from real road environments and preprocessed. The preset traffic operation dataset contains multi-dimensional driving status information of multiple vehicles across multiple time frames, such as vehicle number, lane number, driving direction, frame number, following vehicle number, longitudinal speed, lateral speed, longitudinal acceleration, and lateral acceleration. The preset traffic operation dataset is used as the basic data source for generating subsequent test scenarios.

[0039] The aforementioned multiple vehicles can refer to various types of vehicles participating in real traffic flow, recorded in a pre-defined traffic operation dataset. These can include passenger cars and trucks, and the driving behaviors of these vehicles encompass dynamic processes such as normal cruising, acceleration, deceleration, and lane changing. Among these vehicles, some are identified as the leading vehicle (the vehicle performing the cutting-in action) and the master vehicle (the vehicle being cut in, the simulated object of the test vehicle). The relative motion relationship between the leading and master vehicles can serve as an important basis for generating hazardous test scenarios.

[0040] The aforementioned driving condition parameters can refer to quantified physical quantities that describe the vehicle's motion state within a specific time frame. These parameters may include, but are not limited to, vehicle number, lane number, driving direction, frame number, longitudinal velocity, lateral velocity, longitudinal acceleration, lateral acceleration, longitudinal distance to the vehicle in front, lateral distance, headway, and collision time. These driving condition parameters constitute the multidimensional state vector for each time frame.

[0041] In one optional embodiment, a preset traffic operation dataset can be obtained, which may include multi-dimensional parameters such as vehicle number, lane number, driving direction, frame number, longitudinal and lateral position, speed, and acceleration. The data in the preset traffic operation dataset can also be preprocessed to remove abnormal records such as speeding, reversing, and driving in the wrong direction, while retaining the trajectories of vehicles traveling normally. Subsequently, the data is grouped by vehicle number, and the driving status parameter sequence of each vehicle within consecutive time frames is extracted to form a multi-dimensional vector arranged in chronological order. Each vector contains state information such as position, speed, and acceleration.

[0042] The above settings ensure that the data used accurately reflects the dynamic characteristics of vehicles in actual traffic operations, avoiding biases introduced by artificial simulation. Parameter sequences based on real-world scenarios possess complete spatiotemporal consistency and statistical distribution characteristics, providing a high-quality, high-fidelity input foundation for subsequent parameter importance assessment, joint probability modeling, and scenario extraction. This guarantees that the generated test scenarios are realistically reproducible and engineering-effective.

[0043] Step S104: From the multiple initial parameter types corresponding to the driving condition parameters, determine at least two target parameter types according to the importance of their impact on testing the driver assistance systems installed on multiple vehicles.

[0044] The aforementioned initial parameter types can refer to parameter categories in the preset traffic operation dataset that can be used to describe the vehicle's motion state and its interaction with the environment. These may include, but are not limited to: longitudinal speed, lateral speed, longitudinal acceleration, lateral acceleration, cut-in duration, cut-in direction, longitudinal speed of the main vehicle, lateral speed of the main vehicle, longitudinal acceleration of the main vehicle, lateral acceleration of the main vehicle, relative longitudinal distance, relative lateral distance, relative longitudinal speed, relative lateral speed, driving direction, and number of lanes.

[0045] The aforementioned driver assistance systems may refer to systems such as the automatic emergency braking system being tested. The function of the automatic emergency braking system is to automatically trigger braking when a collision risk is detected ahead to avoid or mitigate a collision.

[0046] The aforementioned at least two target parameter types can refer to parameter types that have a significant impact on the performance of the driver assistance system, selected after an importance assessment of the initial parameter types.

[0047] In one optional embodiment, an ensemble learning model is constructed based on multiple initial parameter types from a pre-set traffic operation dataset, such as vehicle longitudinal and lateral speeds, acceleration, relative distance, number of lanes, and vehicle type. The model is trained to quantify the importance of each parameter in testing the assisted driving system. After determining the optimal number of trees through cross-validation, the trained ensemble learning model outputs an importance score for each initial parameter type. Subsequently, the multiple initial parameter types are ranked according to their importance scores, and at least two parameters with higher scores are selected to obtain at least two target parameter types, thus completing parameter dimensionality compression.

[0048] By implementing the above settings, the efficiency and relevance of test scenario generation can be improved. Only variables that have a greater impact on testing the assisted driving system can be retained, eliminating noise and computational burden introduced by redundant parameters. This ensures that the subsequent joint probability density function modeling and sampling process focuses on important dimensions, reduces computational complexity, enhances the correlation between the generated scenario and the real response mechanism of the assisted driving system, and improves the engineering validity and discriminative power of the test results.

[0049] Step S106: Based on at least two target parameter types, information is extracted from the preset traffic operation dataset to obtain multiple target test scenarios.

[0050] The aforementioned target test scenarios can refer to representative test instances extracted or generated from a pre-defined traffic operation dataset based on target parameter types. Each target test scenario contains combinations of target parameter values ​​and is mapped to executable vehicle trajectories and behaviors in a simulation environment.

[0051] In one optional embodiment, after determining at least two target parameter types, a joint probability density function is constructed using kernel density estimation based on observations of these target parameter types in a pre-defined traffic operation dataset. This function characterizes the collaborative distribution patterns of at least two target parameter types in a real-world scenario. Subsequently, a Monte Carlo two-layer sampling strategy can be combined to generate a basic test scenario based on the joint probability density function, covering high-density areas. Then, an enhanced proposal distribution is constructed, and enhanced test scenarios are generated through importance sampling, focusing on low-probability but high-risk tail regions. Finally, the basic and enhanced test scenarios are integrated to form multiple target test scenarios, covering a complete distribution from low risk to dangerous levels.

[0052] The above settings achieve an efficient balance between realism and challenge in test scenarios, ensuring that the scenario distribution closely resembles real traffic behavior, overcoming the bottleneck of raw data sparsity, and systematically enhancing the coverage of high-risk edge scenarios. This avoids distribution distortion and coverage blind spots caused by manual construction or random generation, improving the sufficiency of testing and the reliability of evaluation of driver assistance systems under extreme conditions.

[0053] Step S108: Based on multiple target test scenarios, perform simulation tests on the driver assistance system to obtain the simulation test results of the driver assistance system.

[0054] The simulation test results mentioned above refer to the quantitative evaluation data obtained after running the assisted driving system in a target test scenario on a collaborative simulation platform. The simulation test results are used to verify the safety and response effectiveness of the assisted driving system under different risk levels.

[0055] In one optional embodiment, multiple generated target test scenarios can be imported into a simulation platform to construct a lane scenario. The initial positions, speeds, and entry trajectories of the test vehicle and target vehicles are set, all reconstructed based on at least two target parameter types extracted from the target test scenarios. Each target test scenario runs the simulation independently, recording output results such as whether the driver assistance system is activated, braking pressure level, vehicle distance, and whether a collision occurs, thus forming the simulation test results.

[0056] The above setup enables systematic and reproducible closed-loop verification of the driver assistance system (ADAS) under extreme conditions in real-world distributed driving scenarios. The data-driven, high-fidelity scenarios cover a full range of conditions, from routine to high-risk, eliminating reliance on human assumptions or a few extreme cases and enhancing the statistical significance and engineering representativeness of the evaluation. Simulation test results reflect the decision robustness and action effectiveness of the ADAS under real-world distributed scenarios, providing quantifiable and traceable empirical evidence for the improvement of ADAS.

[0057] In this embodiment of the invention, firstly, a preset traffic operation dataset is acquired; then, from multiple initial parameter types corresponding to driving condition parameters, at least two target parameter types are determined according to their importance in testing the assisted driving systems installed on multiple vehicles; next, based on the at least two target parameter types, information is extracted from the preset traffic operation dataset to obtain multiple target test scenarios; finally, based on the multiple target test scenarios, simulation tests are performed on the assisted driving system to obtain the simulation test results of the assisted driving system. This application constructs multiple target test scenarios based on a preset traffic operation dataset, ensuring that the data used truly reflects the dynamic characteristics of vehicles in actual traffic operations and avoiding biases introduced by manual simulation. By determining at least two target parameter types based on the importance of testing the driver assistance systems installed on multiple vehicles, the efficiency and relevance of test scenario generation can be improved. Only variables that have a greater impact on the testing of driver assistance systems are retained, eliminating noise and computational burden introduced by redundant parameters. This enhances the correlation between the generated scenarios and the real response mechanism of the driver assistance systems. Based on the obtained multiple target test scenarios, simulation tests are conducted on the driver assistance systems. The simulation test results reflect the decision robustness and action effectiveness of the driver assistance systems under real-world scenario distributions, providing quantifiable and traceable empirical evidence for the improvement of driver assistance systems. This solves the technical problem of low testing efficiency for driver assistance systems in related technologies.

[0058] In this embodiment of the invention, from multiple initial parameter types corresponding to driving condition parameters, at least two target parameter types are determined according to the importance of their impact on testing the assisted driving systems installed on multiple vehicles. This includes: using an ensemble learning model to determine the importance scores corresponding to multiple initial parameter types according to the importance of their impact on testing the assisted driving systems; and determining at least two target parameter types based on the importance scores corresponding to multiple initial parameter types.

[0059] The aforementioned ensemble learning model can refer to a machine learning model built based on the random forest algorithm, used to evaluate the importance of multiple initial parameter types in a pre-defined traffic operation dataset. The ensemble learning model can consist of multiple decision trees, each independently constructed with randomly selected samples and features, and outputs an importance score through voting or averaging. By calculating the average information gain of each initial parameter type when splitting nodes in multiple decision trees, the ensemble learning model outputs the relative importance of each parameter to the prediction, thereby achieving nonlinear, non-parametric importance ranking in a high-dimensional parameter space.

[0060] The aforementioned importance score can refer to the numerical weight index output by the ensemble learning model for each initial parameter type, used to quantify the relative influence of that initial parameter type on the simulation test of the driver assistance system.

[0061] In one optional embodiment, an ensemble learning model of random forest is constructed using multiple initial parameter types from a preset traffic operation dataset as input features. After determining the optimal number of decision trees through cross-validation, the model is trained to learn the nonlinear mapping relationship between each parameter and the risk score. After the ensemble learning model is trained, the contribution of each parameter to the prediction error is calculated using the average impurity reduction, which is used as the importance score. Subsequently, the multiple initial parameter types are sorted according to their importance scores, at least two target parameter types are selected, and the remaining low-scoring parameters are removed, thus completing the parameter space dimensionality reduction.

[0062] By employing the above settings, variables that play a dominant role in the decision-making of the assisted driving system can be accurately identified, avoiding redundant or weakly correlated parameters from interfering with model building and simulation efficiency. Data-driven importance assessment ensures that the screening results possess physical interpretability and engineering relevance, enabling subsequent scenario generation to focus on dimensions that realistically affect braking behavior. This enhances the relevance and effectiveness of test scenarios, providing a foundation for a high-fidelity, low-redundancy testing system.

[0063] In this embodiment of the invention, determining at least two target parameter types based on the importance scores corresponding to multiple initial parameter types includes: sorting the multiple initial parameter types in descending order based on the importance scores corresponding to multiple initial parameter types to obtain an importance ranking queue; and selecting the first preset number of parameter types in the importance ranking queue to obtain at least two target parameter types.

[0064] The aforementioned importance ranking queue can be considered an ordered list formed by arranging the initial parameter types in descending order of importance scores output by the ensemble learning model. The importance ranking queue is structured as pairs of parameter names and corresponding scores, with each item representing an initial parameter type and its degree of impact on the simulation testing of the driver assistance system.

[0065] The aforementioned preset number can refer to a threshold number of target parameter types that is pre-set based on technical objectives and engineering practices and used to extract from the importance ranking queue. The preset number can be determined based on the model output distribution characteristics and complexity balance requirements.

[0066] In one optional embodiment, after obtaining importance scores for multiple initial parameter types output by the ensemble learning model, the initial parameter types are sorted in descending order of importance score to form an importance ranking queue. For example, relative longitudinal speed is ranked first, relative longitudinal distance is ranked second, the longitudinal speed of the main vehicle is ranked third, the lateral speed of the cutting vehicle is ranked fourth, and the remaining parameter scores are ranked lower. Subsequently, based on engineering experience and experimental verification, a preset number of parameter types are selected from the head of the importance ranking queue as target parameter types, and the remaining low-scoring parameters are systematically eliminated, thus completing the parameter selection.

[0067] By implementing the above settings, quantified priorities replace subjective judgment, ensuring that the selected target parameter types have clear statistical significance and decision relevance, avoiding the human error of omitting important variables or introducing irrelevant redundancy. By truncating a pre-set number of parameter types, the repeatability and standardization of the dimensionality reduction process are achieved, improving the computational efficiency and stability of subsequent joint probabilistic modeling and scenario generation. This ensures focus on the variables that truly drive the behavior of the assisted driving system, providing a reliable input foundation for high-precision, low-noise simulation evaluation.

[0068] In this embodiment of the invention, information is extracted from a preset traffic operation dataset based on at least two target parameter types to obtain multiple target test scenarios, including: constructing a joint probability density function corresponding to at least two target parameter types, wherein the joint probability density function is used to characterize the cooperative distribution characteristics of at least two target parameter types in real traffic operation scenarios; and extracting test scenarios from the preset traffic operation dataset based on the joint probability density function to obtain multiple target test scenarios.

[0069] The aforementioned joint probability density function can refer to a mathematical function that describes the multidimensional probability distribution characteristics of at least two selected target parameter types occurring simultaneously in a real traffic operation environment. The joint probability density function characterizes the complex, nonlinear statistical correlations and coordinated changes of target parameter types in a high-speed oncoming vehicle scenario. For example, when the relative longitudinal speed is large, the relative longitudinal distance is often small; when the speed of the preceding vehicle is high, the lateral speed of the oncoming vehicle may exhibit a concentrated distribution within a specific range. The joint probability density function is learned from real-world observed target parameter samples in a pre-defined traffic operation dataset using kernel density estimation methods.

[0070] In one optional embodiment, after determining at least two target parameter types, a joint probability density function is constructed based on measured samples of these target parameter types in a preset traffic operation dataset using a kernel density estimation method. A Gaussian kernel function is used to smooth the sample distribution, accurately characterizing the non-independent cooperative relationship between the at least two target parameter types in real traffic. Subsequently, based on this joint probability density function, a Monte Carlo two-layer sampling strategy is used for scene extraction. Multiple basic test scenarios are generated in high-probability regions. Then, using the joint probability density function as a reference, multiple enhanced test scenarios are generated through importance sampling in low-probability but high-risk tail regions. These enhanced test scenarios are then merged into multiple target test scenarios.

[0071] The above settings can realistically reproduce the complex coupling relationship between at least two target parameter types, avoiding distribution distortion caused by independent assumptions and ensuring that the generated scenarios closely resemble real traffic behavior. The joint probability density function serves as a physical constraint benchmark, ensuring that the extracted scenarios conform to actual statistical laws. Simultaneously, a dual-layer sampling strategy improves the coverage of high-risk edge scenarios, enhancing the testing completeness and evaluation effectiveness of the assisted driving system under real, complex conditions.

[0072] In this embodiment of the invention, test scenarios are extracted from a preset traffic operation dataset based on a joint probability density function to obtain multiple target test scenarios, including: extracting test scenarios from the preset traffic operation dataset based on a state acceptance mechanism sampling algorithm and a joint probability density function to obtain at least one first test scenario; extracting test scenarios from the preset traffic operation dataset based on a weighted augmentation sampling algorithm and a joint probability density function to obtain at least one second test scenario; and integrating at least one first test scenario and at least one second test scenario to obtain multiple target test scenarios.

[0073] The aforementioned state acceptance mechanism sampling algorithm refers to a sampling algorithm based on the Markov chain Monte Carlo method, which is the baseline sampling method used in this application to generate test scenarios that conform to the real traffic statistical distribution from the joint probability density function. The state acceptance mechanism sampling algorithm constructs a Markov chain and, in each iteration, generates a candidate state from a proposal distribution, such as a Gaussian distribution, i.e., a set of target parameter types. It calculates the acceptance probability of the candidate state and the current state relative to the target joint probability density function; and decides whether to accept the candidate state as the next chain state based on the uniform distribution sampling results. The characteristic of the state acceptance mechanism sampling algorithm is the state acceptance mechanism, that is, it does not directly sample the target distribution, but uses a probabilistic acceptance and rejection mechanism to make the stationary distribution of the Markov chain converge to the target joint probability density function. The state acceptance mechanism sampling algorithm can effectively explore regions with high probability density in the parameter space and generate a large number of test scenarios with distributions consistent with real data.

[0074] At least one of the aforementioned first test scenarios can refer to a set of test scenarios that are highly consistent with the statistical distribution of real traffic data, generated by the sampling algorithm of the aforementioned state acceptance mechanism from a preset traffic operation dataset. The first test scenarios are mainly concentrated in the high-density region of the joint probability density function, and constitute the main body of the test scenario library, used to cover the normal response capability of the assisted driving system under normal dangerous conditions.

[0075] The aforementioned weighted augmented sampling algorithm refers to an augmented sampling method based on importance sampling, which is the second-stage sampling technique used in this application for efficiently generating high-risk, low-probability edge scenes. The weighted augmented sampling algorithm introduces an easily sampled proposal distribution, typically an extension of high-density regions or a high-variance distribution, and then assigns importance weights to each sampling point to correct sampling bias, giving higher weight to originally low-probability but high-risk scenes in the statistical estimation. Essentially, the weighted augmented sampling algorithm uses samples from high-probability regions to weighted estimate the expectation of low-probability regions, thereby improving the efficiency of generating dangerous scenes.

[0076] The aforementioned at least one second test scenario can refer to a set of test scenarios with high-risk characteristics generated by a weighted augmented sampling algorithm. The second test scenarios are mainly distributed in the low-density tail region of the joint probability density function. In this application, the weighted augmented sampling algorithm distribution, after weight correction, can effectively reflect the statistical characteristics of real high-risk events. The second test scenarios are systematically amplified and included in the test set through a weighting mechanism.

[0077] In one optional embodiment, based on the joint probability density function, a state acceptance mechanism sampling algorithm is used for sampling. Through Markov chain iteration, at least one first test scenario is stably generated in the high probability density region to ensure that the sample distribution closely approximates the real data distribution. A weighted enhancement sampling algorithm is used to construct a proposed distribution with the joint probability density function as a reference, assigning higher sampling weights to low-probability but high-risk regions, such as small relative distances and high relative velocities, to generate at least one second test scenario, thereby specifically enhancing edge-dangerous scenarios. At least one first test scenario and at least one second test scenario are merged to form multiple target test scenarios.

[0078] Through the above settings, a synergistic enhancement of realism and challenge is achieved. The state acceptance mechanism sampling algorithm ensures the physical rationality of the overall scene distribution, while the weighted augmented sampling algorithm overcomes the bottleneck of original data sparsity and improves the generation efficiency of high-risk scenes. The state acceptance mechanism sampling algorithm and the weighted augmented sampling algorithm complement each other, avoiding the ineffective redundancy of pure random sampling and overcoming the insufficient coverage of tail risks by single sampling. This achieves a better balance between statistical consistency and extreme condition coverage in the test scenarios, improving the reliability of the assisted driving system evaluation in real and complex traffic environments.

[0079] In this embodiment of the invention, based on the state acceptance mechanism sampling algorithm and the joint probability density function, information test scenarios are extracted from a preset traffic operation dataset to obtain at least one first test scenario, including: constructing a scenario rationality evaluation function corresponding to the state acceptance mechanism sampling algorithm based on the joint probability density function; and extracting test scenarios from the preset traffic operation dataset based on the state acceptance mechanism sampling algorithm and the scenario rationality evaluation function to obtain at least one first test scenario.

[0080] The scenario rationality evaluation function mentioned above can be a mathematical function used in the state acceptance mechanism sampling algorithm to quantify whether a candidate test scenario is reasonable, i.e., whether it conforms to the statistical laws of real traffic operation. The scenario rationality evaluation function is learned from a preset traffic operation dataset by the kernel density estimation method, and it represents the cooperative distribution characteristics of the target parameter type in the real scenario.

[0081] In one optional embodiment, a scenario rationality evaluation function is constructed using the joint probability density function as the target distribution. This function employs the density value of the joint probability density function as the basis for calculating the acceptance probability, characterizing the likelihood of a certain parameter combination occurring in real traffic. During sampling, candidate states are randomly generated from the neighborhood of the current state, the acceptance rate is calculated, and a decision is made based on uniformly distributed random numbers to determine whether to accept a state as the next state. This process can be iterated repeatedly to generate a Markov chain that converges to the real distribution. At least one first test scenario is extracted from the Markov chain to ensure that the parameter combinations of each scenario conform to the collaborative statistical laws of variables in real traffic.

[0082] Through the above settings, the state acceptance mechanism sampling algorithm ensures that the sampling process follows the data distribution constraints of the real world, avoiding distortions caused by manual assumptions or independent distributions. Using a probability-driven acceptance mechanism, the state acceptance mechanism sampling algorithm automatically identifies and retains high-probability, highly representative scenarios, effectively filtering out unreasonable or rare combinations. This ensures that the first test scenario has good distribution consistency and physical feasibility, providing stable and reliable baseline samples for subsequent high-risk scenario enhancements.

[0083] In this embodiment of the invention, test scenarios are extracted from a preset traffic operation dataset based on a weighted augmented sampling algorithm and a joint probability density function to obtain at least one second test scenario. This includes: constructing importance sampling weights corresponding to the weighted augmented sampling algorithm based on the joint probability density function; and extracting test scenarios from the preset traffic operation dataset based on the weighted augmented sampling algorithm and the importance sampling weights to obtain at least one second test scenario.

[0084] The aforementioned importance sampling weight can refer to the correction factor assigned to each test scenario sampled from the proposed distribution in the weighted augmented sampling algorithm, used to eliminate systematic bias caused by the inconsistency between the sampling distribution and the target distribution.

[0085] In one optional embodiment, an importance sampling weight is constructed based on a joint probability density function. To enhance the detection capability of high-risk scenarios, a test scenario is extracted from a preset traffic operation dataset based on a weighted augmented sampling algorithm and importance sampling weights, and at least one second test scenario is selected, focusing on covering test scenarios that are scarce but high-risk in the original data.

[0086] Through the above settings and weight correction mechanism, sampling is proactively guided towards areas with real dangers but scarce samples, improving the generation efficiency and coverage of high-risk scenarios. Compared to uniform or blind sampling, this method obtains greater risk information gain with less sampling overhead, allowing the assisted driving system to undergo sufficient stress testing under extreme but real-world conditions. This effectively exposes the decision-making blind spots of the assisted driving system under edge conditions, enhancing the engineering value and robustness verification capabilities of the test.

[0087] The technical solution proposed in this application is described below with reference to an optional embodiment, using an automatic emergency braking system as an example of an assisted driving system. This application proposes a method for generating and testing scenarios of vehicles cutting in front to evaluate the vehicle's automatic emergency braking system. Based on a pre-set traffic operation dataset, such as the HighD dataset, this application comprehensively applies dimensionality reduction algorithms and scene generation methods to create dangerous test scenarios, providing data and methodological support for improving the safety of autonomous driving in high-speed scenarios. The comprehensive solution proposed in this application ensures effective coverage of dangerous operating conditions on real roads and captures the spatiotemporal distribution characteristics of scene elements.

[0088] This application extracts test scenario elements for autonomous vehicle systems, specifically extracting scenarios of vehicles cutting in from ahead on highways. After filtering out problematic data from a pre-set traffic operation dataset, such as speeding, reversing, and driving in the wrong direction, the preceding vehicle cutting-in scenario is extracted. From the pre-set traffic operation dataset, data with vehicles having the same lane number but different lane number values ​​are extracted; this represents a lane-changing behavior, and the vehicle performing the lane change is defined as the preceding vehicle. The start time, cut-in time, and end time of the preceding vehicle's cut-in process are determined as follows: Before the lane number of the merging vehicle changes, for six consecutive frames where the difference in the y-axis between the merging vehicles is positive or negative, the frame with the smaller difference is taken as the lane-change start time. The change in the lane number of the merging vehicle is defined as the cut-in moment. After the lane number of the merging vehicle changes, if the y-axis difference between the merging vehicles does not satisfy the condition of being positive or negative for six consecutive frames, the frame with the smaller difference is defined as the lane-change end time. The complete cut-in process includes at least 6 seconds of vehicle data. Since each frame in the dataset contains 0.04 seconds, data with fewer than 150 frames from start to end time will be excluded. In the data recording lane change points, the later number is defined as the lead vehicle. Information about the lead vehicle and the vehicle in front is integrated into the same document.

[0089] Next, scene element analysis based on the random forest method is performed. Time to Collision (TTC) is used as an important indicator to evaluate driving safety in rear-end collisions, side collisions, and lane-changing scenarios. The smaller the collision time value, the higher the degree of danger. However, when the relative distance between vehicles is small, even if the collision time appears large due to the small speed difference between the lead vehicle and the cutting vehicle, the actual collision risk is still severe. Therefore, collision time and time headway (THW) are used together to evaluate safety. This application uses the Risk Perception (RP) equation to estimate the driver's risk perception when approaching the vehicle in front, with constant weight values ​​for the steady-state term A and the transient term B, to quantify the danger of the autonomous driving cutting process. The calculation expression of the Risk Perception (RP) equation is as follows:

[0090] ;

[0091] Here, A and B are weighting coefficients, which can be 1 and 5 respectively.

[0092] Because the dataset of the preceding vehicle entry scene has high dimensionality and may have inconsistent dimensionality, a random forest algorithm is used to evaluate the importance of risk perception (RP) among scene elements in order to reduce dimensionality and unify dimensionality. The steps of the random forest algorithm are as follows: After preprocessing the data, cross-validation is used to traverse different decision trees to determine the optimal number. A random forest model is built and trained, and the importance of each element is calculated using the trained model. The results of the factor importance analysis are discussed, that is, at least two target parameter types are obtained.

[0093] Next, a preceding vehicle intrusion scene is generated based on Monte Carlo Double-Layer Sampling (MCDLS). This application employs Monte Carlo Double-Layer Sampling, including state acceptance mechanism sampling and weighted augmented sampling. When the target distribution is difficult to sample directly, a Markov chain is constructed using the state acceptance mechanism sampling algorithm. New states are generated based on the proposed distribution, and a certain probability of acceptance is set, thereby converging the stationary distribution of the Markov chain to the target distribution, achieving effective sampling. For scenarios involving rare events, when the target distribution is not easy to sample directly, a proposed distribution that is easy to sample is selected using the weighted augmented sampling algorithm, and importance sampling weights are used to correct bias, thereby estimating the statistical characteristics of the target distribution. Parameterized scenes are extracted from the preceding vehicle intrusion data. The joint probability density function (PDF) of the scene parameters is estimated using the kernel density estimation (KDE) method. KDE is used to derive continuous probability densities by fully utilizing information from the neighborhood of sample points. The formula for estimating kernel density can be expressed as follows:

[0094] ;

[0095] in, It is the joint probability density function obtained through estimation, representing the sample size. Indicates a sample, It is a smoothing parameter. It is a Gaussian kernel.

[0096] After obtaining the joint probability density function of at least two target parameter types using kernel density estimation, at least one first test scenario is generated by sampling using the state acceptor sampling algorithm. The steps of the state acceptor sampling algorithm include: initialization, selecting the average value as the initial value. The Markov chain iterates through the loop, and its state is... , Sample a state from the proposed distribution. Then from a uniform distribution Perform sampling. If Then it accepts iteration. Otherwise, refuse to iterate. Test case generation and estimation involves using a state acceptance mechanism sampling algorithm to obtain parameter samples to generate at least one first test scenario and estimating the probability of events occurring. The relative error of the estimation result can be expressed as follows:

[0097] ;

[0098] in, It is the relative error of the estimation result. It is a sample function.

[0099] A weighted augmented sampling algorithm is used to generate samples based on the state acceptance mechanism sampling algorithm. This algorithm generates a larger number of random test cases, improving scene coverage and increasing the probability of generating edge-hazardous scenarios. The Monte Carlo method is then used to sample from the original distribution. In the process of random sampling, the estimation formula for the mathematical expectation of the sample can be expressed as follows:

[0100] ;

[0101] When it is difficult to obtain samples from a given distribution, an easier-to-sample distribution is introduced. It can be represented as follows:

[0102] ;

[0103] For the above equation to hold true, and It must be absolutely continuous, satisfying the following formula:

[0104] ;

[0105] The unbiased estimate of IS sampling based on the weighted augmentation sampling algorithm is shown in the following formula:

[0106] ;

[0107] ;

[0108] In the given and In this case, it is necessary to determine The distribution of the sample variance. Generally, the smaller the sample variance, the faster the sample expected value converges. The variance formula can be expressed as follows, and the following formula should be made smaller.

[0109] ;

[0110] therefore, Should be in and Matching in common high-value regions The size of the probability density function. Test cases involving collisions or extreme conditions are typically located in regions with low probability density function values. The target probability density function is obtained by sampling using a weighted augmentation sampling algorithm. Assigning higher sampling probabilities to regions with lower values ​​to generate test cases can effectively reduce the variance of the unbiased estimation results and increase the probability of critical events occurring. The relative error of the weighted augmented sampling algorithm estimation results can be expressed as follows:

[0111] ;

[0112] Next, an automatic emergency braking scenario based on time-of-collision (TTC) is constructed. The scenario is defined as a dangerous scenario. The scenario is defined as a high-risk scenario. The scenario is defined as a medium-risk scenario. The scenarios were defined as low-risk scenarios. A large number of test cases were generated under these risk levels, and K-means clustering analysis was applied for simulation testing. Samples were assigned to K cluster centers based on the shortest distance. The center positions were updated by recalculating the within-group means of the labeled samples, and this process was repeated until convergence. The silhouette coefficient was used to determine the optimal number of clusters; the closer the silhouette coefficient was to 1, the better the cluster assignment, and the higher the within-group tightness and between-group separation.

[0113] Finally, a simulation test scenario is run, constructed within the software. The test scenario can be a two-way four-lane straight road, with a test vehicle and a target vehicle configured. Based on the lane-merging scenario element data from the clustering results, the speeds of both vehicles are set, and their driving paths are planned. Alternatively, the test scenario can be configured such that the test vehicle travels straight while the target vehicle cuts in from one side. Two imaging sensors are used to detect the vehicle's environmental information.

[0114] This application proposes a comprehensive model combining random forest, Monte Carlo bilayer sampling, K-means clustering, and simulation for generating and testing high-speed vehicle-to-the-front (VTL) cut-off scenarios. The random forest method is used to evaluate the importance of scenario elements and risk perception parameters. Monte Carlo bilayer sampling is applied to generate automatic emergency braking test cases for VTL cut-off scenarios under high-speed conditions and compared with state acceptance sampling and weighted augmentation sampling algorithms. Monte Carlo bilayer sampling generates more hazardous, high-risk, and medium-risk scenarios. The random forest, Monte Carlo bilayer sampling, and K-means clustering algorithms are also applicable to scenarios such as emergency braking of the vehicle ahead, pedestrian crossings, and vehicles changing lanes laterally.

[0115] Figure 2 This is a schematic diagram illustrating the testing process of an optional driver assistance system according to an embodiment of the present invention, such as... Figure 2 As shown, a pre-defined traffic operation dataset is obtained. Using an ensemble learning model, importance scores are determined for multiple initial parameter types according to their impact on the testing of the assisted driving system. Based on these importance scores, at least two target parameter types are determined. Information is extracted from the pre-defined traffic operation dataset based on these at least two target parameter types to obtain multiple target test scenarios. Simulation tests are then conducted on the assisted driving system based on these target test scenarios to obtain the simulation test results.

[0116] Figure 3 This is a schematic diagram of the testing process for another optional driver assistance system according to an embodiment of the present invention, such as... Figure 3 As shown, a preset traffic operation dataset is obtained. From multiple initial parameter types corresponding to driving condition parameters, at least two target parameter types are determined according to their importance in testing the assisted driving systems installed on multiple vehicles. A joint probability density function corresponding to the at least two target parameter types is constructed. Based on the joint probability density function, test scenarios are extracted from the preset traffic operation dataset to obtain multiple target test scenarios. Based on the multiple target test scenarios, simulation tests are performed on the assisted driving system to obtain the simulation test results of the assisted driving system.

[0117] Figure 4 This is a schematic diagram of the testing process for another optional driver assistance system according to an embodiment of the present invention, such as... Figure 4 As shown, a pre-defined traffic operation dataset is obtained. Using an ensemble learning model, importance scores are determined for multiple initial parameter types according to their impact on the testing of the assisted driving system. Based on these importance scores, at least two target parameter types are determined. A joint probability density function is constructed for each of the at least two target parameter types. Based on this joint probability density function, test scenarios are extracted from the pre-defined traffic operation dataset to obtain multiple target test scenarios. Simulation tests are then conducted on the assisted driving system based on these target test scenarios to obtain the simulation test results.

[0118] Figure 5 This is a schematic diagram of the testing process for another optional driver assistance system according to an embodiment of the present invention, such as... Figure 5As shown, a preset traffic operation dataset is obtained. From multiple initial parameter types corresponding to driving condition parameters, at least two target parameter types are determined according to their importance in testing the assisted driving systems installed on multiple vehicles. A joint probability density function corresponding to the at least two target parameter types is constructed. Based on the state acceptance mechanism sampling algorithm and the joint probability density function, test scenarios are extracted from the preset traffic operation dataset to obtain at least one first test scenario. Based on the weighted augmentation sampling algorithm and the joint probability density function, test scenarios are extracted from the preset traffic operation dataset to obtain at least one second test scenario. The at least one first test scenario and the at least one second test scenario are integrated to obtain multiple target test scenarios. Based on the multiple target test scenarios, the assisted driving system is simulated to obtain the simulation test results of the assisted driving system.

[0119] According to another aspect of the present invention, a test apparatus for a vehicle assisted driving system is also provided. This apparatus can execute the test method for the vehicle assisted driving system described in the above embodiments. The specific implementation method and preferred application scenarios are the same as those described in the above embodiments, and will not be repeated here.

[0120] Figure 6 This is a schematic diagram of a test device for a vehicle driver assistance system according to an embodiment of this application, as shown below. Figure 6 As shown, the device includes the following: an acquisition module 602, a determination module 604, an information extraction module 606, and a simulation test module 608.

[0121] The system comprises the following modules: an acquisition module 602, which acquires a preset traffic operation dataset containing a sequence of driving condition parameters for multiple vehicles, the driving condition parameter sequence being the driving condition parameters of multiple vehicles in multiple consecutive time frames while driving on the road; a determination module 604, which determines at least two target parameter types from multiple initial parameter types corresponding to the driving condition parameters, according to the importance of their impact on testing the assisted driving systems installed on multiple vehicles; an information extraction module 606, which extracts test scenarios from the preset traffic operation dataset based on at least two target parameter types, obtaining multiple target test scenarios; and a simulation test module 608, which performs simulation tests on the assisted driving system based on the multiple target test scenarios, obtaining the simulation test results of the assisted driving system.

[0122] The determination module is used to determine the importance scores corresponding to multiple initial parameter types according to the importance of their impact on the testing of the assisted driving system using an ensemble learning model; and to determine at least two target parameter types based on the importance scores corresponding to the multiple initial parameter types.

[0123] The determination module is used to sort multiple initial parameter types in descending order based on their importance scores to obtain an importance ranking queue; and to select the first preset number of parameter types from the importance ranking queue to obtain at least two target parameter types.

[0124] The information extraction module is used to construct a joint probability density function corresponding to at least two target parameter types. The joint probability density function is used to characterize the collaborative distribution characteristics of at least two target parameter types in real traffic operation scenarios. Based on the joint probability density function, test scenarios are extracted from a preset traffic operation dataset to obtain multiple target test scenarios.

[0125] The information extraction module is used to extract test scenarios from a preset traffic operation dataset based on a state acceptance mechanism sampling algorithm and a joint probability density function to obtain at least one first test scenario; to extract test scenarios from the preset traffic operation dataset based on a weighted augmentation sampling algorithm and a joint probability density function to obtain at least one second test scenario; and to integrate at least one first test scenario and at least one second test scenario to obtain multiple target test scenarios.

[0126] The information extraction module is used to construct a scenario rationality evaluation function corresponding to the state acceptance mechanism sampling algorithm based on the joint probability density function; and to extract test scenarios from the preset traffic operation dataset based on the state acceptance mechanism sampling algorithm and the scenario rationality evaluation function to obtain at least one first test scenario.

[0127] The information extraction module is used to construct the importance sampling weights corresponding to the weighted augmented sampling algorithm based on the joint probability density function; and to extract test scenarios from the preset traffic operation dataset based on the weighted augmented sampling algorithm and the importance sampling weights to obtain at least one second test scenario.

[0128] Embodiments of this application also provide an electronic device, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of the present invention during runtime.

[0129] The aforementioned memory can refer to devices inside a computer used to store data and programs, including RAM, hard disks, etc. RAM can be used to temporarily store running programs and data, while hard disks can be used to store programs and data long-term. Memory enables the computer to read and write data and execute programs. The aforementioned processor is responsible for executing instructions in computer programs and performing data processing. It can also be responsible for controlling and executing various operations, including arithmetic operations, logical operations, and data transmission.

[0130] Embodiments of this application also provide a computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of the present invention.

[0131] The aforementioned computer storage media can refer to the media used in computer memory to store certain discontinuous physical quantities. Computer storage media mainly include semiconductors, magnetic cores, magnetic drums, magnetic tapes, laser discs, etc. Computer-readable storage media include stored programs, which can be a set of instructions that a computer can recognize and execute, running on an electronic computer to meet certain information needs.

[0132] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the methods of various embodiments of the present invention.

[0133] The aforementioned computer program products can refer to software programs that have been written, tested, and released, and can run on computers or other devices. Computer program products can include application programs, operating systems, utility software, etc., used to achieve specific functions or solve specific problems.

[0134] Embodiments of this application also provide a computer program product, including a non-volatile computer-readable storage medium for storing a computer program that, when executed by a processor, implements the methods in various embodiments of the present invention.

[0135] The aforementioned non-volatile computer-readable storage medium can refer to a medium for storing data. Non-volatile computer-readable storage media can retain data without loss when power is off and can be used to store long-term data, such as operating systems, applications, and user files. Non-volatile storage media can include hard disk drives, solid-state drives, optical disks, and flash memory storage devices, etc.

[0136] Embodiments of this application also provide a computer program that, when executed by a processor, implements the methods described in the various embodiments of the present invention.

[0137] The aforementioned computer program can refer to a set of instructions used to tell the computer to perform specific tasks or operations. Computer programs can be written by programmers using specific programming languages ​​and can include algorithms, data structures, logic, and control flow. Computer programs can be used for a variety of purposes, including application software, operating systems, etc.

[0138] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0139] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection can be through some interfaces; the indirect coupling or communication connection between units or modules can be electrical or other forms.

[0140] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0141] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0142] If the integrated unit is implemented as 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 the present invention, in essence, or the part that contributes to the prior art, or all or part of the 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 cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0143] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A test method for a vehicle driver assistance system, characterized in that, include: Obtain a preset traffic operation dataset, wherein the preset traffic operation dataset contains a sequence of driving condition parameters for multiple vehicles, and the sequence of driving condition parameters is the driving condition parameters of the multiple vehicles in multiple consecutive time frames when they are driving on the road; From the multiple initial parameter types corresponding to the driving condition parameters, at least two target parameter types are determined according to the importance of their impact on testing the driver assistance systems installed on the multiple vehicles. Based on the at least two target parameter types, information is extracted from the preset traffic operation dataset to obtain multiple target test scenarios; Based on the multiple target test scenarios, the driver assistance system is simulated and tested to obtain the simulation test results of the driver assistance system.

2. The test method for a vehicle driver assistance system according to claim 1, characterized in that, From the multiple initial parameter types corresponding to the driving condition parameters, at least two target parameter types are determined according to their importance in testing the driver assistance systems installed on the multiple vehicles, including: Using an ensemble learning model, importance scores are determined for multiple initial parameter types according to their impact on the testing of the assisted driving system. Based on the importance scores corresponding to the multiple initial parameter types, the at least two target parameter types are determined.

3. The test method for a vehicle driver assistance system according to claim 2, characterized in that, Based on the importance scores corresponding to the multiple initial parameter types, the at least two target parameter types are determined, including: Based on the importance scores corresponding to the multiple initial parameter types, the multiple initial parameter types are sorted in descending order to obtain an importance sorting queue; Select the first preset number of parameter types from the importance sorting queue to obtain the at least two target parameter types.

4. The test method for the vehicle driver assistance system according to any one of claims 1 to 3, characterized in that, Based on the at least two target parameter types, information is extracted from the preset traffic operation dataset to obtain multiple target test scenarios, including: Construct a joint probability density function corresponding to the at least two target parameter types, wherein the joint probability density function is used to characterize the cooperative distribution characteristics of the at least two target parameter types in real traffic operation scenarios; Based on the joint probability density function, test scenarios are extracted from the preset traffic operation dataset to obtain the multiple target test scenarios.

5. The test method for a vehicle driver assistance system according to claim 4, characterized in that, Based on the joint probability density function, test scenarios are extracted from the preset traffic operation dataset to obtain the multiple target test scenarios, including: Based on the state acceptance mechanism sampling algorithm and the joint probability density function, test scenarios are extracted from the preset traffic operation dataset to obtain at least one first test scenario. Based on the weighted enhanced sampling algorithm and the joint probability density function, test scenarios are extracted from the preset traffic operation dataset to obtain at least one second test scenario. The at least one first test scenario and the at least one second test scenario are integrated to obtain the plurality of target test scenarios.

6. The test method for a vehicle driver assistance system according to claim 5, characterized in that, Based on the state acceptance mechanism sampling algorithm and the joint probability density function, information test scenarios are extracted from the preset traffic operation dataset to obtain at least one first test scenario, including: Based on the joint probability density function, construct the scenario rationality evaluation function corresponding to the state acceptance mechanism sampling algorithm; Based on the state acceptance mechanism sampling algorithm and the scenario rationality evaluation function, test scenarios are extracted from the preset traffic operation dataset to obtain at least one first test scenario.

7. The test method for a vehicle driver assistance system according to claim 5, characterized in that, Based on the weighted augmented sampling algorithm and the joint probability density function, test scenarios are extracted from the preset traffic operation dataset to obtain at least one second test scenario, including: Based on the joint probability density function, the importance sampling weights corresponding to the weighted enhanced sampling algorithm are constructed; Based on the weighted enhanced sampling algorithm and the importance sampling weight, test scenarios are extracted from the preset traffic operation dataset to obtain at least one second test scenario.

8. A testing device for a vehicle driver assistance system, characterized in that, include: The acquisition module is used to acquire a preset traffic operation dataset, wherein the preset traffic operation dataset contains a sequence of driving condition parameters for multiple vehicles, and the sequence of driving condition parameters is the driving condition parameters of the multiple vehicles in multiple consecutive time frames when they are driving on the road. The determination module is used to determine at least two target parameter types from multiple initial parameter types corresponding to the driving condition parameters, according to the importance of their impact on testing the driver assistance systems installed on the multiple vehicles. The information extraction module is used to extract test scenarios from the preset traffic operation dataset based on the at least two target parameter types, thereby obtaining multiple target test scenarios; The simulation testing module is used to perform simulation tests on the driver assistance system based on the multiple target test scenarios, and obtain the simulation test results of the driver assistance system.

9. An electronic device, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, executes the test method for the vehicle driver assistance system according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the storage medium is located to perform a test method for the vehicle driver assistance system according to any one of claims 1 to 7.

11. A computer program product, characterized in that, Includes computer instructions that, when executed by a processor, implement the test method for the vehicle driver assistance system according to any one of claims 1 to 7.

Images