Method and device for determining vehicle performance boundary, terminal equipment and storage medium

Abstract

The application is suitable for the automatic driving technical field, and provides a vehicle performance boundary determination method and device, a terminal device and a storage medium. In the embodiment of the application, a first test sample parameter is obtained from a parameter space, the first test sample parameter is used to describe a specific scene in which a test vehicle is located; the first test sample parameter is input into a preset test platform to test the test vehicle, a first test result corresponding to the first test sample parameter is obtained, the first test result is used to describe a driving state of the test vehicle simulated by the test platform in the specific scene; a preset agent model is trained according to the first test sample parameter and the first test result; if the agent model meets a preset termination condition, a performance boundary of the test vehicle is determined according to distribution information of the first test sample parameter, so that the automatic driving capability of the vehicle is clearly displayed by determining the performance boundary of the test vehicle.

Description

Technical Field

[0001] This application belongs to the field of autonomous driving technology, and in particular relates to a method, apparatus, terminal device and storage medium for determining vehicle performance boundaries. Background Technology

[0002] With the development of society, autonomous driving technology is receiving increasing attention. Effectively verifying the safety of autonomous driving technology is an important foundation for its implementation. Current safety verification methods for autonomous driving technology are usually based on testing and verification of specific functions. These tests and verifications cannot clearly demonstrate the autonomous driving capabilities of vehicles in certain driving scenarios. Summary of the Invention

[0003] This application provides a method, apparatus, terminal device, and storage medium for determining vehicle performance boundaries, which can solve the problem of not being able to clearly demonstrate the autonomous driving capabilities of a vehicle in some driving scenarios.

[0004] In a first aspect, embodiments of this application provide a method for determining vehicle performance boundaries, including:

[0005] The first test sample parameters are obtained from the preset parameter space. The first test sample parameters are used to describe the specific scenario in which the test vehicle is located.

[0006] The first test sample parameters are input into a preset test platform to test the test vehicle and obtain the first test result corresponding to the first test sample parameters. The first test result is used to describe the driving state of the test vehicle simulated by the test platform in the specific scenario.

[0007] The preset proxy model is trained based on the parameters of the first test sample and the results of the first test.

[0008] If the above proxy model meets the preset termination condition, the performance boundary of the above test vehicle is determined based on the distribution information of the parameters of the first test sample.

[0009] Secondly, embodiments of this application provide a device for determining vehicle performance boundaries, comprising:

[0010] The parameter acquisition module is used to acquire the first test sample parameters from the preset parameter space. The first test sample parameters are used to describe the specific scenario in which the test vehicle is located.

[0011] The testing module is used to input the first test sample parameters into a preset testing platform to test the test vehicle and obtain the first test result corresponding to the first test sample parameters. The first test result is used to describe the driving state of the test vehicle simulated by the testing platform in the specific scenario.

[0012] The training module is used to train the preset proxy model based on the parameters of the first test sample and the first test result.

[0013] The boundary determination module is used to determine the performance boundary of the test vehicle based on the distribution information of the parameters of the first test sample if the above proxy model meets the preset termination condition.

[0014] Thirdly, embodiments of this application provide a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the above-mentioned methods for determining vehicle performance boundaries.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described methods for determining vehicle performance boundaries.

[0016] Fifthly, embodiments of this application provide a computer program product that, when run on a terminal device, causes the terminal device to execute any of the vehicle performance boundary determination methods described in the first aspect.

[0017] In this embodiment, a first test sample parameter is obtained from a preset parameter space to describe the specific scenario in which the test vehicle is located. This specific scenario is the vehicle's driving scenario. The first test sample parameter is then input into a preset test platform to test the test vehicle, and a first test result corresponding to the first test sample parameter is obtained. This first test result is used to describe the driving state of the test vehicle in the specific scenario simulated by the test platform. Based on the first test sample parameter and the first test result, a preset proxy model is trained. If the proxy model meets the preset termination condition, it means that the output result of the current proxy model has met the accuracy requirement for determining the performance boundary of the vehicle's autonomous driving function in the parameter space. Then, the performance boundary of the test vehicle is determined based on the distribution information of the first test sample parameter, so as to clearly demonstrate the vehicle's autonomous driving capability in a specific driving scenario through the performance boundary. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the first method for determining the vehicle performance boundary provided in the embodiments of this application;

[0020] Figure 2 This is a schematic diagram of vehicle performance boundaries provided in an embodiment of this application;

[0021] Figure 3 This is a schematic diagram of the second process of the method for determining the vehicle performance boundary provided in the embodiments of this application;

[0022] Figure 4 This is a schematic diagram of the structure of the vehicle performance boundary determination device provided in the embodiments of this application;

[0023] Figure 5 This is a schematic diagram of the structure of the terminal device provided in the embodiments of this application. Detailed Implementation

[0024] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0025] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0026] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0027] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0028] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0029] Figure 1 The diagram shown is a flowchart illustrating a method for determining vehicle performance boundaries according to an embodiment of this application. The execution subject of this method can be a terminal device, such as... Figure 1 As shown, the method for determining the vehicle performance boundary may include the following steps:

[0030] Step S101: Obtain the first test sample parameters from the preset parameter space. The first test sample parameters are used to describe the specific scenario in which the test vehicle is located.

[0031] In this embodiment, a specific scenario in which a vehicle is located can be described by parameters of different dimensions. This specific scenario is equivalent to a special case of various driving scenarios during the vehicle's operation. For example, in the scenario of two illegally parked vehicles, the scenario is described by the specific value of the relative distance parameter between the two vehicles. However, since the values ​​of parameters of different dimensions under the same logical scenario are different, the corresponding specific scenarios are also different, which leads to different degrees of impact on vehicle safety. Therefore, in order to clearly define the autonomous driving capability of a vehicle in a specific scenario, the above-mentioned parameter space is composed of key parameters of at least one dimension corresponding to the logical scenario. Thus, by obtaining the first test sample parameters from the parameter space, the purpose of describing a specific scenario is achieved. Here, the above-mentioned first test sample parameters are the parameters currently obtained to describe the specific scenario of the test vehicle. If the current scenario is the nth test scenario, then the first test sample parameters are the sample parameters to be obtained in the nth test. They may contain the specific values ​​of at least one dimension of the parameter to be studied, which are determined according to the number of key parameters constituting the parameter space. Each round of testing requires acquiring a certain number of first test sample parameters to correspond to a certain number of specific scenarios. Based on these specific scenarios, the test vehicle is then tested. The number of the first test sample parameters can be any integer obtained according to user requirements, such as 50. The test vehicle can be an autonomous vehicle, including Level 1-Level 5 autonomous vehicles. In some cases, the test vehicle can also be a non-autonomous vehicle, such as a vehicle that can provide driving suggestions to assist the driver. This embodiment uses an autonomous vehicle as an example for illustration.

[0032] Specifically, step S101 may include: the terminal device obtaining first test sample parameters from the parameter space according to a preset random algorithm. This random algorithm can determine the first test sample parameters based on a seed number. Determining the seed number facilitates the traceability of the results. The seed number is not limited to a fixed combination; any combination of these seed numbers constitutes a design change in this example. Furthermore, the terminal device may also obtain the first test sample parameters from the parameter space using a Bayesian optimization algorithm based on previously tested test sample parameters.

[0033] In one embodiment, before step S101 above, the terminal device needs to determine the logical scenario faced by the test vehicle, that is, to clarify which driving scenario the user wants to explore and the corresponding performance boundary of the vehicle when performing autonomous driving. This logical scenario includes, but is not limited to, dangerous scenarios obtained through road testing, dangerous scenarios obtained through simulation, common scenarios in the autonomous driving process, or any scenario involving risks. It is understood that the user can improve the relevant performance based on the performance boundary of the vehicle. Therefore, the above logical scenario should be a scenario that helps improve the vehicle's performance. For example, a scenario of two illegally parked vehicles, that is, a scenario in which the test vehicle uses the autonomous driving function to drive when faced with two illegally parked vehicles. During the test, the test vehicle may collide or scrape due to its own autonomous driving technology not being sensitive enough in avoiding the illegally parked two vehicles, or it may successfully pass through. Therefore, the performance boundary under this scenario can improve the obstacle avoidance performance of the autonomous driving technology.

[0034] Understandably, the aforementioned performance boundary is equivalent to a boundary defining whether a vehicle can achieve safe driving in a certain logical scenario. By determining this performance boundary, we can obtain the range in which the vehicle can successfully navigate all possible situations within that logical scenario, as well as the range in which it cannot or is uncertain whether it can successfully navigate. All possible situations of the logical scenario constitute the parameter space corresponding to that scenario, and the aforementioned range corresponds to the various regions in the parameter space divided by the performance boundary. Therefore, by determining this performance boundary, we can determine the vehicle's autonomous driving capability; that is, the smaller the range in which it cannot or is uncertain whether it can successfully navigate, the better the autonomous driving capability. Moreover, since the vehicle's autonomous driving capability can be determined based on this performance boundary, we can quantitatively provide the verification pass conditions for the safety verification of autonomous driving technology based on the relevant requirements for autonomous driving safety. For example, we can set verification scenarios based on key parameters near the performance boundary to determine whether the vehicle can successfully navigate, thereby improving the verification effect when verifying the safety of autonomous driving technology by quantitatively providing verification pass conditions. In addition, we can also modify the operating conditions set by the autonomous driving system functions and define functions through this performance boundary, thereby further improving the autonomous driving function.

[0035] In one embodiment, before step S101, the terminal device needs to determine the parameter space corresponding to the above logical scenario. Specifically, the terminal device can determine the key parameters of each dimension corresponding to the logical scenario from a preset parameter set, and further determine the parameter range values ​​corresponding to each key parameter. Thus, the parameter space is determined based on the key parameters of at least one dimension corresponding to the logical scenario and the parameter range values ​​corresponding to the key parameters. The parameter range values ​​can be set according to the range values ​​that can exist under the logical scenario, the parameter range values ​​that may exist in dangerous situations according to big data analysis, or the user's needs. For example, in the scenario of two vehicles illegally parked, it is set that both vehicles can illegally park outside the lane. Since the road width is fixed, the parameter range value of the relative width Δx in the relative distance between the two vehicles does not exceed the road width. The relative distance is the distance between the centroids of the two vehicles.

[0036] For example, when facing a scenario of two vehicles illegally parked, the key parameters are the y-coordinate of the first vehicle, the yaw angle of the first vehicle, the relative distance (Δx, Δy) between the second vehicle and the first vehicle, and the yaw angle of the second vehicle. The parameter space corresponding to the scenario of two vehicles illegally parked is determined based on the y-coordinate of the first vehicle, the yaw angle of the first vehicle, the relative distance between the second vehicle and the first vehicle, and the yaw angle of the second vehicle.

[0037] In one embodiment, before determining the parameter space, the method may further include: preprocessing the parameter range values ​​corresponding to the key parameters, that is, transforming the parameter range values ​​corresponding to the key parameters into a range that conforms to multi-dimensional parameters and is easy to process. For example, if the yaw angle of the first vehicle is obtained relative to location A and the yaw angle of the second vehicle is obtained relative to location B, the yaw angle of the second vehicle can be transformed into the yaw angle relative to location A. In addition, when there are missing values ​​in the parameter range values ​​corresponding to the key parameters, missing value filling can be performed; when there are duplicate values ​​in the parameter range values ​​corresponding to the key parameters, duplicate value deletion can be performed, etc.

[0038] Step S102: Input the first test sample parameters into the preset test platform to test the test vehicle and obtain the first test result corresponding to the first test sample parameters. The first test result is used to describe the driving state of the test vehicle simulated by the test platform in a specific scenario.

[0039] In this embodiment, the terminal device inputs the first test sample parameters obtained above into the test platform, enabling the test platform to simulate specific scenarios under different conditions within the same logical scenario. The test vehicle then drives within these simulated scenarios, obtaining the driving state of the test vehicle corresponding to each scenario. This driving state is the first test result, which includes, but is not limited to, collision, friction, and successful passage. It is understood that since the first test result may not necessarily satisfy the termination condition for the proxy model, it needs to be stored after obtaining the first test result. This storage path can be set by the user. Furthermore, for ease of data processing, the first test result can be formatted, for example, converted to a commonly used data tag format.

[0040] Specifically, the aforementioned testing platforms include, but are not limited to, simulation software, self-developed simulation platforms, closed road testing platforms, and public road testing grounds. The aforementioned simulation software includes, but is not limited to, Carla, VTD, Carsim, and Sumo. Open-source autonomous driving systems or self-developed autonomous driving systems can be set on the aforementioned test vehicles, such as Advanced Driving Assistance Systems (ADAS), which can correspond to L1-L2 level or L3 and above autonomous driving systems.

[0041] Step S103: Train the preset proxy model based on the parameters of the first test sample and the first test result.

[0042] In this embodiment, the aforementioned proxy model includes, but is not limited to, machine learning proxy models such as Gaussian process proxy classification model, logistic regression proxy model, and random forest proxy model. By training the proxy model, the probability and standard deviation in the distribution information of the parameters of the first test sample can be obtained, that is, the predicted value and uncertainty of the proxy model for a certain probability of the parameters of the first test sample can be obtained. The distribution information includes, but is not limited to, density, predicted classification probability and standard deviation.

[0043] It is understandable that if the first test parameter is obtained from the parameter space based on the test sample parameters that have been tested before, then the above proxy model needs to be trained based on the test sample parameters that have been tested before, the test results corresponding to the test sample parameters that have been tested before, the first test sample parameters, and the first test results.

[0044] Step S104: If the proxy model meets the preset termination condition, then determine the performance boundary of the test vehicle based on the distribution information of the parameters of the first test sample.

[0045] In this embodiment, if the proxy model meets the preset termination condition, it means that the number of sample parameters currently trained by the proxy model has met the user's accuracy requirements. Then, the performance boundary of the test vehicle in the parameter space can be determined based on the distribution information of the first test sample parameters used to train the proxy model. It is understood that since the performance boundary depends on the logical scenario, the performance boundary of each key parameter in the parameter space has a small floating value. For example, the y-coordinate of the first vehicle floats at (-2m, +2m), and the yaw angle of the first vehicle floats at (-30°, +30°).

[0046] For example, such as Figure 2 As shown, if the current parameter space contains only two key parameters, namely key parameter A and key parameter B, and the parameter ranges of key parameter A and key parameter B are both 0.0 to 1.0, then the performance boundary of the test vehicle can be determined based on the probability in the distribution information. Figure 2 Different colors in the diagram correspond to different probabilities. Figure 2 The dashed ring in the diagram represents the performance boundary. The area outside the dashed ring is the normal functional area, meaning that the test vehicle can successfully pass the tests in all the specific scenarios corresponding to this area. For example... Figure 2 When the test sample parameter is (0.2, 0.2), the corresponding test result is a successful pass; the area within the dashed ring is the functional failure area, which means that the test vehicle cannot pass successfully in any specific scenario corresponding to this area, for example... Figure 2 When the test sample parameters are (0.6, 0.4), the corresponding test result may be a collision; the area between the dashed rings is a functionally ambiguous area, meaning that in the specific scenarios corresponding to this area, the test vehicle may or may not successfully pass through, for example... Figure 2 When the test sample parameters are (0.8, 0.2), the corresponding test result may be friction.

[0047] Specifically, the termination conditions mentioned above include, but are not limited to, the convergence of the surrogate model and the number of training iterations of the surrogate model reaching a preset number. These termination conditions can be designed according to different requirements and objectives. Specifically, convergence of the surrogate model can be determined by judging the magnitude of the change in its training parameters relative to a preset value. For example, if the change in the surrogate model's training parameters is not greater than a preset value, the surrogate model is considered convergent, thus terminating the test. The aforementioned training parameters include, but are not limited to, the kernel parameters of the Gaussian process radial basis function (RBF). Furthermore, to ensure the accuracy of the results, it can be further specified that if the change in the surrogate model's training parameters is not greater than a preset value for a consecutive preset number of iterations, the surrogate model is considered convergent.

[0048] It is understandable that if the first test parameter is obtained from the parameter space based on the test sample parameters that have been tested previously, then after the surrogate model meets the preset termination condition, the performance boundary of the test vehicle needs to be determined based on the distribution information of the test sample parameters that have been tested previously and the distribution information of the first test sample parameters.

[0049] It is understood that in this embodiment, specific scenarios for different situations are set in the parameter space corresponding to the logical scenario, and the test vehicle is tested based on the set specific scenarios. In this way, when facing certain specific scenarios where the truth value label cannot be clearly defined before driving, i.e. driving scenarios with unknown dangers, the performance boundary under unknown danger driving scenarios can be given in a targeted manner, thereby improving the vehicle's autonomous driving system's ability to control risks.

[0050] In one embodiment, such as Figure 3 As shown, when the parameters of the first test sample are obtained from the parameter space based on the parameters of previously tested test samples, the method for determining the vehicle performance boundary may include the following steps:

[0051] Step S301: Obtain the first test sample parameter from the parameter space based on the distribution information of at least one second test sample parameter.

[0052] In this embodiment, the second test sample parameter is a test sample parameter that has already been tested. For example, if the first test sample parameter is the sample parameter to be obtained in the nth iteration, then the second test sample parameter is the sample parameter that has already been tested and obtained from the 1st to the (n-1th)th iteration. It is understood that, compared to methods that obtain the first test sample parameter by random sampling or interval sampling from the parameter space, this embodiment can obtain the distribution state of the already tested test sample parameters in the parameter space through the distribution information of the already tested test sample parameters, thereby selectively obtaining the current test sample parameters. This can reduce the number of tests, improve testing efficiency, and further improve the accuracy of performance boundary determination. Experiments show that when the model meets the above termination conditions, compared to interval sampling and random sampling, this embodiment reduces the number of sample parameters to be tested by 37.5%, and improves testing efficiency by more than 60%.

[0053] In one embodiment, step S301 may include: the terminal device generating a sampling function based on the distribution information of the second test sample parameters, generating a target distribution map corresponding to the parameter space based on the sampling function, determining an initial performance boundary in the target distribution map, and selecting the first test sample parameters from the target distribution map based on the initial performance boundary. Furthermore, after obtaining the target distribution map, the test sample parameters to be sampled in the map can be corrected before sampling, thereby improving data accuracy.

[0054] In one embodiment, since the performance boundary includes a functionally normal region, a functionally failed region, and a functionally ambiguous region, the selection of the first test sample parameters from the target distribution map based on the initial performance boundary may include: determining the functionally normal region, functionally failed region, and functionally ambiguous region in the target distribution map based on the initial performance boundary, such as... Figure 2 As shown, the first test sample parameters are selected from the normal function region, the malfunction region, and the ambiguous function region according to the preset selection ratio.

[0055] In one embodiment, to improve the accuracy of performance boundaries, the key region selected for the test sample parameters is the functionally ambiguous region. Therefore, the value corresponding to the functionally ambiguous region in the above selection ratio can be set to be greater than the values ​​corresponding to the other two regions, thereby increasing the accuracy of performance boundary determination. For example, the selection ratio of the normal functional region, the functionally failed region, and the functionally ambiguous region is 1:1:2, and the preset number of the first test sample parameters is 40. Then, the number of the first test sample parameters corresponding to the normal functional region is 10, the number of the first test sample parameters corresponding to the functionally failed region is 10, and the number of the first test sample parameters corresponding to the functionally ambiguous region is 40.

[0056] In one embodiment, the above-mentioned selection of the first test sample parameters from the normal functional region, the functional failure region, and the functional ambiguity region according to the preset selection ratio may include: dividing the normal functional region, the functional failure region, and the functional ambiguity region into their respective target regions, selecting the target region with the fewest second test sample parameters from each target region, and selecting the first test sample parameters from the target region with the fewest second test sample parameters.

[0057] In one embodiment, when the distribution information of the second test sample parameters includes probability, standard deviation, and location, generating the acquisition function based on the distribution information of the second test sample parameters may include: the terminal device generating a probability function, a standard deviation function, and a density function based on the probability, standard deviation, and location corresponding to the second test sample parameters, and then generating an acquisition function based on the probability function, standard deviation function, and density function, as shown in the following formula:

[0058] AF(x)=p(x)+λ1σ(x)+λ2ρ(x)

[0059] Wherein, AF is the acquisition function; x is the parameter vector used to represent the entire parameter space; p is the probability function; σ is the standard deviation function; ρ is the density function; and λ1 and λ2 are weighting coefficients.

[0060] In one embodiment, the above-mentioned acquisition function can also be determined based on a derived combination of the probability, standard deviation and position corresponding to the second test sample parameters, such as p(x)σ(x), p(x)ρ(x), etc.

[0061] In one embodiment, the terminal device can generate a predicted probability distribution map, a standard deviation distribution map, and a density map in the entire parameter space based on the above probability function, standard deviation function, and density function, respectively, so that the user can view them.

[0062] Step S302: Train the surrogate model based on the first test sample parameters, the first test result, the second test sample parameters, and the second test result corresponding to the second test sample parameters.

[0063] In this embodiment, the second test result corresponding to the second test sample parameters can be obtained by inputting the second test sample parameters into a preset test platform to test the test vehicle.

[0064] Step S303: Determine whether the proxy model meets the termination condition.

[0065] If yes, proceed to step S304; otherwise, proceed to step S301 and subsequent steps.

[0066] Step S304: Determine the performance boundary of the test vehicle based on the distribution information of the parameters of the first test sample and the distribution information of the parameters of the second test sample.

[0067] In this embodiment, if the termination condition is met, the terminal device directly determines the performance boundary based on the distribution information of the test sample parameters that have been tested so far. The test sample parameters that have been tested so far are the first test sample parameters and the second test sample parameters.

[0068] Understandably, if the termination condition is not met, step S301 and subsequent steps need to be executed again. Test sample parameters are obtained again based on the distribution information of the test sample parameters that have been tested so far, and training is performed. That is, the third test sample parameter is obtained from the parameter space based on the distribution information of the first test sample parameter and the distribution information of the second test sample parameter. For example, if the first test sample parameter is the sample parameter obtained in the nth test, and the second test sample parameter is the sample parameter obtained from the 1st to the (n-1)th test, then the third test sample parameter is the sample parameter obtained in the (n+1)th test, which corresponds to the specific scenario of the test vehicle in the (n+1)th test. Training is then performed based on the third test sample parameter until the surrogate model meets the termination condition.

[0069] In this embodiment, a first test sample parameter is obtained from a preset parameter space to describe the specific scenario in which the test vehicle is located. This specific scenario is the vehicle's driving scenario. The first test sample parameter is then input into a preset test platform to test the test vehicle, and a first test result corresponding to the first test sample parameter is obtained. This first test result is used to describe the driving state of the test vehicle in the specific scenario simulated by the test platform. Based on the first test sample parameter and the first test result, a preset proxy model is trained. If the proxy model meets the preset termination condition, it means that the output result of the current proxy model has met the accuracy requirement for determining the performance boundary of the vehicle's autonomous driving function in the parameter space. Then, the performance boundary of the test vehicle is determined based on the distribution information of the first test sample parameter, so as to clearly demonstrate the vehicle's autonomous driving capability in a specific driving scenario through the performance boundary.

[0070] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0071] Corresponding to the method for determining vehicle performance boundaries described above, Figure 4 The diagram shown is a structural schematic of a vehicle performance boundary determination device according to an embodiment of this application. Figure 4 As shown, the device for determining the vehicle performance boundary may include:

[0072] The parameter acquisition module 401 is used to acquire first test sample parameters from a preset parameter space. The first test sample parameters are used to describe the specific scenario in which the test vehicle is located.

[0073] The test module 402 is used to input the first test sample parameters into a preset test platform to test the test vehicle and obtain the first test result corresponding to the first test sample parameters. The first test result is used to describe the driving state of the test vehicle simulated by the test platform in a specific scenario.

[0074] Training module 403 is used to train a preset proxy model based on the parameters of the first test sample and the first test result.

[0075] The boundary determination module 404 is used to determine the performance boundary of the test vehicle based on the distribution information of the parameters of the first test sample if the proxy model meets the preset termination condition.

[0076] In one embodiment, the parameter acquisition module 401 may include:

[0077] The first parameter acquisition submodule is used to acquire the first test sample parameters from the parameter space based on the distribution information of at least one second test sample parameter.

[0078] Accordingly, the training module 403 mentioned above may include:

[0079] The first training submodule is used to train the surrogate model based on the first test sample parameters, the first test result, the second test sample parameters, and the second test result corresponding to the second test sample parameters.

[0080] Accordingly, the boundary determination module 404 may include:

[0081] The first boundary determination submodule is used to determine the performance boundary of the test vehicle based on the distribution information of the first test sample parameters and the distribution information of the second test sample parameters if the surrogate model meets the termination condition.

[0082] In one embodiment, the above-mentioned vehicle performance boundary determination device may further include:

[0083] The second parameter acquisition submodule is used to obtain the third test sample parameters from the parameter space based on the distribution information of the first test sample parameters and the distribution information of the second test sample parameters if the proxy model does not meet the termination condition.

[0084] In one embodiment, the aforementioned first parameter acquisition submodule may include:

[0085] The function generation unit is used to generate a data acquisition function based on the distribution information of the parameters of the second test sample.

[0086] The boundary determination unit is used to generate a target distribution map corresponding to the parameter space based on the acquisition function, and to determine the initial performance boundary in the target distribution map.

[0087] The parameter selection unit is used to select the parameters of the first test sample from the target distribution map based on the initial performance boundary.

[0088] In one embodiment, the parameter selection unit may include:

[0089] The region determination sub-unit is used to determine the functionally normal region, functionally failed region, and functionally ambiguous region in the target distribution map based on the initial performance boundary.

[0090] The parameter selection subunit is used to select the first test sample parameters from the normal function region, the functional failure region, and the functional ambiguity region according to the preset selection ratio.

[0091] In one embodiment, the distribution information of the second test sample parameters includes probability, standard deviation, and location, and the function generation unit may include:

[0092] The first function generates a sub-unit, which is used to generate the probability function, standard deviation function, and density function based on the probability, standard deviation, and location, respectively.

[0093] The second function generation sub-unit is used to generate the acquisition function based on the probability function, standard deviation function, and density function.

[0094] In one embodiment, the parameter acquisition module 401 may further include:

[0095] The third parameter acquisition submodule is used to obtain the parameters of the first test sample from the parameter space according to a preset random algorithm.

[0096] In this embodiment, first test sample parameters describing the specific scenario of the test vehicle are obtained from a preset parameter space. This specific scenario is the vehicle's driving scenario. The first test sample parameters are then input into a preset test platform to test the test vehicle, resulting in a first test result. This first test result describes the driving state of the test vehicle in the specific scenario as simulated by the test platform. Based on the first test sample parameters and the first test result, a preset proxy model is trained. If the proxy model meets a preset termination condition, it indicates that the output of the current proxy model has met the accuracy requirements for determining the performance boundary of the vehicle's autonomous driving function in the parameter space. The performance boundary of the test vehicle is then determined based on the distribution information of the first test sample parameters, so as to clearly demonstrate the vehicle's autonomous driving capability in a specific driving scenario.

[0097] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described device and module can be referred to the corresponding process in the foregoing system embodiments and method embodiments, and will not be repeated here.

[0098] Figure 5 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. For ease of explanation, only the parts related to the embodiments of this application are shown.

[0099] like Figure 5 As shown, the terminal device 5 in this embodiment includes: at least one processor 500 ( Figure 5 (Only one is shown in the image), a memory 501 connected to the processor 500, and a computer program 502 stored in the memory 501 and executable on at least one processor 500, such as a vehicle performance boundary determination program. When the processor 500 executes the computer program 502, it implements the steps in the various vehicle performance boundary determination method embodiments described above, for example... Figure 1 Steps S101 to S104 are shown. Alternatively, when the processor 500 executes the computer program 502, it implements the functions of each module in the above-described device embodiments, for example... Figure 4 The functions of modules 401 to 404 are shown.

[0100] For example, the computer program 502 described above can be divided into one or more modules. One or more of these modules are stored in the memory 501 and executed by the processor 500 to complete this application. The one or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 502 in the terminal device 5. For example, the computer program 502 can be divided into a parameter acquisition module 401, a testing module 402, a training module 403, and a boundary determination module 404, with the specific functions of each module as follows:

[0101] The parameter acquisition module 401 is used to acquire first test sample parameters from a preset parameter space. The first test sample parameters are used to describe the specific scenario in which the test vehicle is located.

[0102] Test module 402 is used to input the first test sample parameters into a preset test platform to test the test vehicle and obtain the first test result corresponding to the first test sample parameters. The first test result is used to describe the driving state of the test vehicle simulated by the test platform in a specific scenario.

[0103] Training module 403 is used to train a preset proxy model based on the parameters of the first test sample and the first test result;

[0104] The boundary determination module 404 is used to determine the performance boundary of the test vehicle based on the distribution information of the parameters of the first test sample if the proxy model meets the preset termination condition.

[0105] The aforementioned terminal device 5 may include, but is not limited to, a processor 500 and a memory 501. Those skilled in the art will understand that... Figure 5 This is merely an example of terminal device 5 and does not constitute a limitation on terminal device 5. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, it may also include input / output devices, network access devices, buses, etc.

[0106] The processor 500 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0107] In some embodiments, the aforementioned memory 501 may be an internal storage unit of the terminal device 5, such as a hard disk or memory of the terminal device 5. In other embodiments, the aforementioned memory 501 may be an external storage device of the terminal device 5, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the terminal device 5. Furthermore, the aforementioned memory 501 may include both internal storage units and external storage devices of the terminal device 5. The aforementioned memory 501 is used to store the operating system, applications, boot loader, data, and other programs, such as the program code of the aforementioned computer programs. The aforementioned memory 501 may also be used to temporarily store data that has been output or will be output.

[0108] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the above device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments 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. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. In the above embodiments, the descriptions of each embodiment have different focuses; parts not described or recorded in detail in a certain embodiment can be referred to the relevant descriptions of other embodiments.

[0109] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0110] In the embodiments provided in this application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are merely illustrative. For instance, the division of modules or units described above is only 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 coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0111] The units described above 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 network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0112] If the integrated units described above are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0113] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for determining the performance boundary of a vehicle, characterized in that, include: Obtain the first test sample parameters from the preset parameter space. The first test sample parameters are used to describe the specific scenario in which the test vehicle is located. The first test sample parameters are input into a preset test platform to test the test vehicle, and a first test result corresponding to the first test sample parameters is obtained. The first test result is used to describe the driving state of the test vehicle in the specific scenario simulated by the test platform. The preset proxy model is trained based on the parameters of the first test sample and the first test result; If the proxy model meets the preset termination condition, the performance boundary of the test vehicle in the parameter space is determined according to the distribution information of the parameters of the first test sample. The performance boundary is equivalent to a boundary of whether the vehicle can achieve safe driving in a certain logical scenario. By determining the performance boundary, the range in which the vehicle can successfully pass in all possible situations of the logical scenario, as well as the range in which it cannot or is uncertain whether it can successfully pass, can be obtained. The performance boundary includes a normal function region, a function failure region, and a function ambiguity region. The step of obtaining the first test sample parameters from the preset parameter space includes: The first test sample parameter is obtained from the parameter space based on the distribution information of at least one second test sample parameter, wherein the second test sample parameter is a test sample parameter that has been tested. Accordingly, training the preset proxy model based on the parameters of the first test sample and the first test result includes: The proxy model is trained based on the first test sample parameters, the first test result, the second test sample parameters, and the second test result corresponding to the second test sample parameters. Accordingly, if the proxy model satisfies the preset termination condition, then determining the performance boundary of the test vehicle based on the distribution information of the first test sample parameters includes: If the proxy model satisfies the termination condition, the performance boundary of the test vehicle is determined based on the distribution information of the first test sample parameters and the distribution information of the second test sample parameters.

2. The method for determining the vehicle performance boundary as described in claim 1, characterized in that, Also includes: If the proxy model does not meet the termination condition, then the third test sample parameter is obtained from the parameter space based on the distribution information of the first test sample parameter and the distribution information of the second test sample parameter.

3. The method for determining the vehicle performance boundary as described in claim 1, characterized in that, The step of obtaining the first test sample parameters from the parameter space based on the distribution information of at least one second test sample parameter includes: A data acquisition function is generated based on the distribution information of the parameters of the second test sample; The target distribution map corresponding to the parameter space is generated according to the acquisition function, and the initial performance boundary is determined in the target distribution map; The parameters of the first test sample are selected from the target distribution map based on the initial performance boundary.

4. The method for determining the vehicle performance boundary as described in claim 3, characterized in that, The step of selecting the first test sample parameters from the target distribution map based on the initial performance boundary includes: Based on the initial performance boundary, determine the functionally normal region, functionally failed region, and functionally ambiguous region in the target distribution map; The first test sample parameters are selected from the functional normal region, the functional failure region, and the functional ambiguous region according to the preset selection ratio.

5. The method for determining the vehicle performance boundary as described in claim 3, characterized in that, The distribution information of the second test sample parameters includes probability, standard deviation, and location. Generating the acquisition function based on the distribution information of the second test sample parameters includes: Generate a probability function, a standard deviation function, and a density function based on the probability, the standard deviation, and the location, respectively; The acquisition function is generated based on the probability function, the standard deviation function, and the density function.

6. The method for determining the vehicle performance boundary as described in claim 1, characterized in that, The step of obtaining the first test sample parameters from the preset parameter space includes: The first test sample parameters are obtained from the parameter space according to a preset random algorithm.

7. A device for determining the performance boundary of a vehicle, characterized in that, include: The parameter acquisition module is used to acquire the first test sample parameters from the preset parameter space. The first test sample parameters are used to describe the specific scenario in which the test vehicle is located. The testing module is used to input the first test sample parameters into a preset testing platform to test the test vehicle and obtain a first test result corresponding to the first test sample parameters. The first test result is used to describe the driving state of the test vehicle simulated by the testing platform in the specific scenario. The training module is used to train a preset proxy model based on the parameters of the first test sample and the first test result. The boundary determination module is used to determine the performance boundary of the test vehicle in the parameter space based on the distribution information of the parameters of the first test sample if the proxy model meets the preset termination condition. The performance boundary is equivalent to a boundary of whether the vehicle can achieve safe driving in a certain logical scenario. By determining the performance boundary, the range in which the vehicle can successfully pass in all possible situations of the logical scenario can be obtained, as well as the range in which it cannot or is uncertain whether it can successfully pass. The performance boundary includes a normal function region, a function failure region, and a function ambiguity region. The parameter acquisition module includes: The first parameter acquisition submodule is used to acquire the first test sample parameter from the parameter space based on the distribution information of at least one second test sample parameter, wherein the second test sample parameter is a test sample parameter that has been tested. Accordingly, the training module includes: The first training submodule is used to train the surrogate model based on the first test sample parameters, the first test result, the second test sample parameters, and the second test result corresponding to the second test sample parameters. The boundary determination module includes: The first boundary determination submodule is used to determine the performance boundary of the test vehicle based on the distribution information of the first test sample parameters and the distribution information of the second test sample parameters if the surrogate model meets the termination condition.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for determining the vehicle performance boundary as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of a method for determining vehicle performance boundaries as described in any one of claims 1 to 6.

Images