A training method and an evaluation method of a vehicle evaluation model

Abstract

This application provides a training method and an evaluation method for a vehicle evaluation model. The training method obtains a test training set and its true labels for a target vehicle under target indicators. The test training set includes test sub-item data for several indicator sub-items and sub-item labels for each test sub-item. The test training set is input into a first neural network for training, resulting in a trained first neural network and several sub-item prediction data output by the trained first neural network. The true labels and all sub-item prediction data are input into a second neural network for training, resulting in a trained second neural network. Based on the trained first and second neural networks, a trained vehicle evaluation model is constructed. This training method can provide a vehicle evaluation model that can effectively improve the accuracy and consistency of vehicle evaluation. This invention relates to the field of vehicle technology.

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and in particular to a training method and evaluation method for a vehicle evaluation model. Background Technology

[0002] In the development of vehicle dynamics performance, subjective evaluation of the vehicle is highly valued by relevant personnel because it has a direct impact on the optimization of vehicle dynamics performance.

[0003] Currently, related technologies typically rely on professional car evaluators conducting real-vehicle experiments at customized test tracks and evaluating vehicle performance based on their driving experience. This approach heavily depends on the personal experience and feelings of professional car evaluators, and the evaluation results are often discrete and inconsistent, making it difficult to quantify effectively. This limits the correlation between the evaluation results and objective test data on vehicle dynamics, resulting in unsatisfactory accuracy and consistency of the evaluation results.

[0004] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the Invention

[0005] The purpose of this invention is to at least partially solve one of the technical problems existing in the related art.

[0006] The main objective of this application is to propose a training method and an evaluation method for a vehicle evaluation model. The training method can provide a vehicle evaluation model that can effectively improve the accuracy and consistency of vehicle evaluation.

[0007] To achieve the above objectives, one aspect of this application proposes a method for training a vehicle evaluation model, comprising: Obtain the test training set of the target vehicle under the target indicator item and the real label of the test training set. The test training set includes test sub-item data of several indicator sub-items and sub-item labels of each test sub-item data. The test training set is input into the first neural network for training to obtain a trained first neural network and several sub-item prediction data output by the trained first neural network. The real labels and all the predicted data of the sub-items are input into the second neural network for training to obtain the trained second neural network. A trained vehicle evaluation model is constructed based on the trained first neural network and the trained second neural network.

[0008] In addition, the training method for a vehicle evaluation model according to the above embodiments of this application may also have the following additional technical features: In some embodiments, obtaining the test training set includes: Obtain experimental test data for the target vehicle; Based on each sub-item of the target indicator, the experimental test data is cropped and aligned to obtain the test sub-item data and sub-item label for each sub-item. The test training set is constructed based on all the test sub-item data and the sub-item labels.

[0009] In some embodiments, the first neural network includes a feature extraction module and a prediction module. The step of inputting the test training set into the first neural network for training to obtain a trained first neural network, and the training first neural network outputting a plurality of sub-item prediction data, includes: The test training set is input into the feature extraction module for feature extraction to obtain several test features, each of which corresponds to a test sub-item data. Each of the test features is input into the prediction module to perform sub-item prediction, thereby obtaining the sub-item prediction data; Based on the sub-item label, a first loss value is calculated for the corresponding sub-item prediction data to obtain the first loss value; Based on the first loss value, the parameters of the first neural network are updated to obtain the trained first neural network.

[0010] In some embodiments, the first neural network includes a feature extraction module and a prediction module. The feature extraction module includes several extraction sub-modules. The test sub-item data includes several signal data, each signal data corresponding to one extraction sub-module. The test sub-item data and the sub-item labels are input into the first neural network for training to obtain a trained first neural network and sub-item prediction data output by the trained first neural network, including: Each signal data is input into the corresponding extraction submodule for feature extraction, resulting in several signal features; All the signal features are input into the prediction module to perform sub-item prediction, and the sub-item prediction data is obtained. Based on the sub-item label, a second loss value is calculated for the corresponding sub-item prediction data to obtain the second loss value; Based on the second loss value, the parameters of the first neural network are updated to obtain the trained first neural network.

[0011] In some embodiments, target data is input into the target module for feature extraction to obtain target features, including: Local feature extraction is performed on the target data to obtain intermediate feature vectors; The intermediate feature vector is reduced in dimension and pooled to obtain the target feature; If the target data is a test training set, then the target module is a feature extraction module, and the target feature is a test feature; or, if the target data is signal data, then the target module is an extraction sub-module, and the target feature is a signal feature.

[0012] In some embodiments, the step of inputting the true labels and all the predicted sub-items into a second neural network for training to obtain a trained second neural network includes: A nonlinear transformation is performed on all the predicted sub-items to obtain the sub-item transformation features; The transformation features of the sub-items are evaluated and predicted to obtain the evaluation and prediction results corresponding to the target index item; Based on the actual labels, a third loss value is calculated on the evaluation prediction results to obtain the third loss value; Based on the third loss value, the parameters of the second neural network are updated to obtain the trained third neural network.

[0013] To achieve the above objectives, another aspect of this application proposes an evaluation method for a vehicle evaluation model, comprising: Obtain the vehicle test set for the vehicle to be evaluated; The vehicle test set is input into the trained vehicle evaluation model to evaluate the vehicle and obtain the vehicle evaluation result of the vehicle to be evaluated.

[0014] To achieve the above objectives, another aspect of this application proposes a method for training a vehicle evaluation model, comprising: The first processing unit is used to obtain the test training set of the vehicle under the target indicator item and the real label of the test training set. The test training set includes test sub-item data of several indicator sub-items and sub-item labels of each test sub-item data. The second processing unit is used to input the test training set into the first neural network for training, to obtain the trained first neural network and several sub-item prediction data output by the trained first neural network. The third processing unit is used to input the real labels and all the predicted data of the sub-items into the second neural network for training, so as to obtain the trained second neural network. The fourth processing unit is used to construct a trained vehicle evaluation model based on the trained first neural network and the trained second neural network.

[0015] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned auxiliary control method.

[0016] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned auxiliary control method.

[0017] On the other hand, embodiments of this application also provide a computer program product, which includes a computer program stored in a computer-readable storage medium. The processor of an electronic device reads the computer program from the computer-readable storage medium and executes the computer program, causing the electronic device to perform the method described above.

[0018] The embodiments of this application include at least the following beneficial effects: This application provides a training method and evaluation method for a vehicle evaluation model. The training method obtains a test training set of a target vehicle for a target indicator item and the true labels of the test training set. The test training set includes test sub-item data for several indicator sub-items and sub-item labels for each test sub-item data. The test training set is input into a first neural network for training to obtain a trained first neural network and several sub-item prediction data output by the trained first neural network. The true labels and all the sub-item prediction data are input into a second neural network for training to obtain a trained second neural network. Based on the trained first neural network and the trained second neural network, a trained vehicle evaluation model is constructed. This training method trains the first neural network using the test sub-item data and sub-item labels in the test training set, and trains the second neural network using the sub-item prediction data output by the first neural network and the true labels of the test training set. This allows the trained vehicle evaluation model to capture and learn the correlation between the evaluation results and the vehicle test data, which is beneficial to improving the accuracy and consistency of the evaluation results. Attached Figure Description

[0019] Figure 1 This is a flowchart of a training method for a vehicle evaluation model provided in an embodiment of this application; Figure 2 This is a flowchart illustrating how to obtain a test training set, as provided in an embodiment of this application. Figure 3 This is a detailed flowchart of the first method for training a first neural network provided in the embodiments of this application; Figure 4This is a detailed flowchart of the second method for training the first neural network provided in the embodiments of this application; Figure 5 This is a detailed flowchart of feature extraction provided in an embodiment of this application; Figure 6 This is a detailed flowchart of step S130 provided in an embodiment of this application; Figure 7 This is a schematic diagram of a network framework for a second neural network provided in an embodiment of this application; Figure 8 This is a flowchart of an evaluation method for a vehicle evaluation model provided in an embodiment of this application; Figure 9 This is a schematic diagram of the structure of a training system for a vehicle evaluation model provided in an embodiment of this application; Figure 10 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] This application will be further described in detail below. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of this application; they are merely examples of apparatuses / devices and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0021] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0022] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0024] Currently, related technologies typically rely on professional car reviewers conducting real-vehicle tests at customized test tracks and evaluating vehicle performance based on their driving experience. While this subjective evaluation method is widely used in the automotive industry, it suffers from the following problems: 1. Cost and resource dependence: Traditional subjective evaluation methods require a large amount of human resources and expensive test sites and equipment, resulting in high costs.

[0025] 2. Discrepancy and lack of quantification in evaluation results: Due to reliance on personal experience and feelings, evaluation results are often discrete and difficult to quantify effectively, which limits the correlation between evaluation results and objective test data of vehicle dynamics.

[0026] 3. Difficulty in design optimization: Designers find it difficult to translate the vague feelings of professional car reviewers into specific engineering parameters. The product optimization process relies on repeated trial production and adjustments, which not only increases the R&D cycle but also increases costs.

[0027] In summary, the subjective evaluation methods mentioned above have drawbacks such as unsatisfactory accuracy and consistency of evaluation results, high cost, and long R&D efficiency.

[0028] It should be noted that the aforementioned related technologies are only used to assist in understanding the technical solutions of this application and do not mean that they belong to the publicly disclosed prior art.

[0029] In view of this, this application provides a training method and an evaluation method for a vehicle evaluation model. The training method can learn the relationship between each indicator sub-item and its corresponding sub-item label from a local perspective through a first neural network, and learn the relationship between the real label and the sub-item prediction data of each indicator sub-item from a global perspective through a second neural network. This enables the trained vehicle evaluation model to capture and learn the relationship between the evaluation results and the vehicle test data, which is beneficial to improving the accuracy and consistency of the evaluation results.

[0030] The method provided in this application can be applied to a terminal, a server, or software running on a terminal or server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or in-vehicle terminal, but is not limited thereto; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application implementing the method, but is not limited to the above forms.

[0031] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0032] Figure 1 This is an optional flowchart of a training method for a vehicle evaluation model provided in an embodiment of this application. Figure 1 The method may include, but is not limited to, steps S110 to S140.

[0033] Step S110: Obtain the test training set of the target vehicle under the target indicator item and the real label of the test training set. The test training set includes test sub-item data of several indicator sub-items and sub-item labels of each test sub-item data. In the embodiments of this application, the target index item and index sub-item can be subjective evaluation indexes in the subjective evaluation test of the vehicle during the development process of vehicle dynamics performance. Specifically, the evaluation system of the vehicle subjective evaluation test usually includes several levels of subjective evaluation indexes. For example, the evaluation system of a vehicle manufacturer includes several primary evaluation indexes, each primary evaluation index contains several secondary evaluation indexes, and each secondary evaluation index contains several tertiary evaluation indexes. Primary evaluation indexes are usually handling stability, smoothness, comfort, etc., while the sub-evaluation indexes under each primary rating index (such as secondary evaluation indexes, or tertiary rating indexes under secondary evaluation indexes) are usually set differently depending on the vehicle manufacturer, and this application does not impose any restrictions here.

[0034] It is understood that, for ease of understanding, this application embodiment takes the target indicator item as the first-level evaluation indicator, and the indicator sub-item as the second-level evaluation indicator under the first-level evaluation indicator as an example. The content of the target indicator item being the second-level evaluation indicator and so on can be simply deduced by analogy.

[0035] Specifically, if the target indicator is operational stability, the sub-indicators can be steering feel, understeer, tilt characteristics, diagonal sway, etc. The test data corresponding to the steering feel indicator can be steering wheel angle, steering wheel torque, etc., and the sub-label can be a subjective evaluation value, which is usually provided by relevant personnel in the vehicle subjective evaluation test. The test data corresponding to the understeer indicator can be the understeer degree of the vehicle; the test data corresponding to the tilt characteristics indicator can be the body roll angle, vehicle speed pitch angle, etc.; the test data corresponding to the diagonal sway indicator can be the yaw rate of the vehicle, etc. The examples of target indicators and indicator sub-items in this application are only illustrative and are not intended to limit this application. Specific target indicators and indicator sub-items have been implemented in various ways in the vehicle field. For example, when the target indicator is ride comfort, its corresponding indicator sub-items can be the vehicle's straight-line driving performance (such as steering disturbance stability, crosswind stability, steering wheel centering feel, steering wheel force feel uniformity, etc.) and braking performance (such as straight-line braking stability, cornering braking, etc.) related to smooth driving.

[0036] It should be noted that the sub-item labels for the other indicator sub-items are similar to those for the aforementioned steering feel indicator sub-item, and can be easily deduced by analogy. The vehicle's test sub-item data can be calculated from objective vehicle data (such as acceleration, speed, and steering angle) collected during vehicle operation. This objective vehicle data can be acquired by sensors deployed at various locations on the vehicle; various implementation methods already exist, which will not be elaborated upon here.

[0037] Reference Figure 2 In some embodiments, obtaining the test training set includes: Step S210: Obtain experimental test data of the target vehicle; Step S220: Based on each sub-item of the target indicator, the experimental test data is cropped and aligned to obtain the test sub-item data and sub-item label for each sub-item of the indicator; Step S230: Construct the test training set based on all the test sub-item data and the sub-item labels.

[0038] In this embodiment, basic parameters of the target vehicle can be collected by sensors deployed at various locations on the vehicle to obtain experimental test data. For any sub-item of the target indicator, data clipping can be performed by clipping the objective vehicle data under the test operation corresponding to the sub-item from the experimental test data, thereby obtaining the test sub-item data of the sub-item and corresponding it with the subjective evaluation value (i.e., sub-item label) under the test operation. The test sub-item data and sub-item labels of the other sub-items can be derived in the same way. The test sub-item data and sub-item labels of each sub-item under the target indicator are used to construct the test training set of the target indicator. The true label of the test training set is also the true label of the target indicator.

[0039] It is understood that in the first implementation, the true label can be calculated by weighted summation of all sub-labels, and there are already various specific implementation methods; while in the second implementation, the true label can be provided by relevant personnel in the vehicle subjective evaluation test, and this application does not restrict the method of obtaining the true label.

[0040] Step S120: Input the test training set into the first neural network for training to obtain the trained first neural network and several sub-item prediction data output by the trained first neural network; In this embodiment, the test training set can be input into the first neural network. The first neural network learns the test sub-item data of each sub-item under the target index and performs prediction scoring, thereby obtaining the trained first neural network and several sub-item prediction data output by the trained first neural network. Each sub-item prediction data corresponds to one index sub-item.

[0041] It is understandable that there are multiple ways to implement the specific training process of the first neural network. For example, the parameters of the first neural network can be updated by the backpropagation algorithm to obtain the trained first neural network. This application will not elaborate on the specific training process of the first neural network here.

[0042] Reference Figure 3 In some embodiments, the first neural network includes a feature extraction module and a prediction module. The step of inputting the test training set into the first neural network for training to obtain a trained first neural network, and the training first neural network outputting several sub-item prediction data, includes: Step S310: Input the test training set into the feature extraction module for feature extraction to obtain several test features, each of which corresponds to a test sub-item data; Step S320: Input each of the test features into the prediction module to perform sub-item prediction and obtain the sub-item prediction data; Step S330: Calculate the first loss value for the corresponding predicted data of the sub-item based on the sub-item label to obtain the first loss value; Step S340: Update the parameters of the first neural network according to the first loss value to obtain the trained first neural network.

[0043] In the embodiments of this application, in the first implementation, the first neural network includes several feature extraction modules connected in parallel and a prediction module. The feature extraction module may be a convolutional neural network (CNN), and the network structure of each feature extraction module is the same. The prediction module may be a long short-term memory network (LSTM).

[0044] Understandably, step S310 can involve inputting each test sub-item data in the test training set into a corresponding feature extraction module for feature extraction, thereby obtaining several test features. Specifically, if the test training set includes three test sub-item data, and the corresponding first neural network includes parallel feature extraction modules, then the first test sub-item data can be input into the first feature extraction module for feature extraction to obtain the first test feature; the second test sub-item data can be input into the second feature extraction module for feature extraction to obtain the second test feature; and the third test sub-item data can be input into the third feature extraction module for feature extraction to obtain the third test feature.

[0045] It should be noted that sub-item prediction can be achieved by capturing the long-term temporal dependencies of each test feature through the prediction module and outputting sub-item prediction data for each indicator sub-item. The first loss value calculation is used to measure the degree of difference between the sub-item label and the sub-item prediction data. Specifically, it can be calculated by using a loss function to calculate the loss value corresponding to the sub-item label and the sub-item prediction data. There are many commonly used loss functions, such as 0-1 loss function, squared loss function, absolute loss function, log loss function, and cross-entropy loss function, which will not be elaborated on in this application. In the embodiments of this application, any loss function can be selected to determine the training loss value, such as the squared loss function. Based on the calculated first loss value, the parameters of the first neural network are updated using the backpropagation algorithm. After several iterations, the trained first neural network can be obtained. The specific number of iterations can be preset, or training can be considered complete when the test set reaches the accuracy requirement.

[0046] Reference Figure 4 In some embodiments, the first neural network includes a feature extraction module and a prediction module. The feature extraction module includes several extraction sub-modules. The test sub-item data includes several signal data, each signal data corresponding to one extraction sub-module. The test sub-item data and the sub-item labels are input into the first neural network for training to obtain a trained first neural network and sub-item prediction data output by the trained first neural network, including: Step S410: Input each of the signal data into the corresponding extraction submodule for feature extraction to obtain several signal features; Step S420: Input all the signal features into the prediction module to perform sub-item prediction and obtain the sub-item prediction data; Step S430: Calculate the second loss value for the corresponding sub-item prediction data based on the sub-item label to obtain the second loss value; Step S440: Update the parameters of the first neural network according to the second loss value to obtain the trained first neural network.

[0047] In this embodiment, in the second implementation, the first neural network includes a feature extraction module and a prediction module. The feature extraction module includes several extraction sub-modules connected in parallel, and each extraction sub-module can be a convolutional neural network (CNN). The network structure of each extraction sub-module is consistent, while the prediction module can be a long short-term memory network.

[0048] Understandably, for any test sub-item data in the test training set, each signal data point in the test sub-item data can be input into a corresponding extraction sub-module for feature extraction, thereby obtaining the signal features corresponding to each signal data point. Sub-item prediction can be achieved by inputting all signal features under the test sub-item data into the prediction module. The prediction module then captures the long-term temporal dependencies of all signal features and outputs the sub-item prediction data for the corresponding indicator sub-item. In practical applications, this sub-item prediction data is usually the predicted score of the corresponding indicator sub-item.

[0049] It is worth mentioning that the content of steps S430 to S440 is similar to that of steps S330 to S340 mentioned above, and can be easily deduced by analogy. Furthermore, in practical applications, many sub-items are related to various objective vehicle data. For example, the test data for the steering feel sub-item includes time-series data composed of the steering wheel angle during vehicle operation, denoted as steering wheel angle time-series data, and time-series data composed of steering wheel torque, denoted as steering wheel torque time-series data. Therefore, for the test data of the steering feel sub-item, step S410 can be to determine the steering wheel angle time-series data as the first signal data and perform feature extraction, and to determine the steering wheel torque time-series data as the second signal data and perform feature extraction. The test data for other sub-items are similarly deduced by analogy.

[0050] Reference Figure 5 In some embodiments, target data is input into the target module for feature extraction to obtain target features, including: Step S510: Extract local features from the target data to obtain intermediate feature vectors; Step S520: Perform dimensionality reduction pooling on the intermediate feature vector to obtain the target feature; If the target data is a test training set, then the target module is a feature extraction module, and the target feature is a test feature; or, if the target data is signal data, then the target module is an extraction sub-module, and the target feature is a signal feature.

[0051] In this embodiment of the application, taking the target module as the feature extraction module as an example, step S510 can be based on a one-dimensional convolutional layer in the convolutional neural network to perform local feature extraction on the test sub-item data in the test training set to obtain the corresponding feature values, which are denoted as intermediate feature vectors. Then, the intermediate feature vectors are sequentially input into the batch normalization (BN) layer, ReLU activation layer, pooling layer, dropout layer and fully connected layer in the convolutional neural network. The feature values ​​output by the ReLU activation layer are reduced in dimensionality by the pooling layer, and the feature values ​​are integrated by the fully connected layer to obtain the target features.

[0052] It is understandable that the target module is the content of the extraction sub-module, which is similar to the target module being the content of the feature extraction module mentioned above, and can be easily deduced by analogy.

[0053] Step S130: Input the real labels and all the predicted data of the sub-items into the second neural network for training to obtain the trained second neural network; In this embodiment of the application, the sub-item prediction data of all sub-items under a certain target indicator can be input into the second neural network. The second neural network learns the test prediction data of each sub-item under the target indicator and performs prediction scoring. The trained second neural network is obtained based on the real labels.

[0054] Reference Figure 6 and Figure 7 In some embodiments, step S130, inputting the true labels and all the predicted sub-items into a second neural network for training to obtain a trained second neural network, includes: Step S610: Perform a nonlinear transformation on all the predicted sub-item data to obtain the sub-item transformation features; Step S620: Evaluate and predict the transformation features of the sub-items to obtain the evaluation and prediction results corresponding to the target index item; Step S630: Calculate the third loss value based on the real label and the evaluation prediction result to obtain the third loss value; Step S640: Update the parameters of the second neural network according to the third loss value to obtain the trained third neural network.

[0055] In this embodiment of the application, the second neural network may be a backpropagation neural network, which includes an input layer, a hidden layer and an output layer. The input layer is used to receive all sub-item prediction data, the hidden layer is used to learn the complex nonlinear relationship between all sub-item prediction data, and the output layer is used to perform the final score prediction.

[0056] For example, in this embodiment of the application, the number of predicted sub-item data is 3. Figure 7 The input layer can receive three sub-item prediction data and provide the output of all received sub-item prediction data to the hidden layer. Step S610 can be a nonlinear transformation of the input layer output through the hidden layer. Specifically, it can be a weighted summation of the input layer output through a fully connected layer in the hidden layer, and a standardization of the fully connected layer output through a batch normalization layer. Then, a nonlinear transformation of the feature values ​​of the batch normalization layer output is performed through a ReLU activation layer in the hidden layer. The output of the ReLU activation layer is then passed through a Dropout layer to obtain the sub-item transformation features.

[0057] Understandably, the evaluation prediction can be achieved by processing the sub-item transformation features through the output layer of the feedforward neural network. This output layer includes a fully connected layer and a Softmax activation layer. The output layer is used to map and convert the sub-item transformation features into the final predicted value, i.e., the evaluation prediction result. The content of steps S630 to S640 is similar to that of steps S330 to S340 mentioned above and can be easily deduced by analogy.

[0058] Step S140: Construct a trained vehicle evaluation model based on the trained first neural network and the trained second neural network.

[0059] In the embodiments of this application, a trained vehicle evaluation model can be obtained by combining a trained first neural network and a trained second neural network.

[0060] Reference Figure 8 This application provides an evaluation method for a vehicle evaluation model, including: Step S810: Obtain the vehicle test set of the vehicle to be evaluated; Step S820: Input the vehicle test set into the trained vehicle evaluation model to perform vehicle evaluation and obtain the vehicle evaluation result of the vehicle to be evaluated.

[0061] In practical applications, a vehicle test set of the vehicle to be evaluated can be obtained. This vehicle test set can be a collection of objective vehicle data during the driving process of the vehicle to be evaluated. Then, the vehicle test set is input into the pre-trained vehicle evaluation model, and the vehicle is evaluated by the pre-trained vehicle evaluation model to obtain the vehicle evaluation result of the vehicle to be evaluated.

[0062] Please see Figure 9 This application also provides a training system for a vehicle evaluation model, which can implement the above-described training method for the vehicle evaluation model. The training system includes: The first processing unit 901 is used to obtain the test training set of the vehicle under the target indicator item and the real label of the test training set. The test training set includes test sub-item data of several indicator sub-items and sub-item labels of each test sub-item data. The second processing unit 902 is used to input the test training set into the first neural network for training, to obtain the trained first neural network and several sub-item prediction data output by the trained first neural network. The third processing unit 903 is used to input the real label and all the sub-item prediction data into the second neural network for training, so as to obtain the trained second neural network. The fourth processing unit 904 is used to construct a trained vehicle evaluation model based on the trained first neural network and the trained second neural network.

[0063] It is understood that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0064] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0065] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0066] Please see Figure 10 , Figure 10 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 1001 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 1002 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1002 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1002 and is called and executed by the processor 1001. Input / output interface 1003 is used to implement information input and output; The communication interface 1004 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 1005 transmits information between various components of the device (e.g., processor 1001, memory 1002, input / output interface 1003, and communication interface 1004); The processor 1001, memory 1002, input / output interface 1003 and communication interface 1004 are connected to each other within the device via bus 1005.

[0067] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0068] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0069] This application also discloses a computer program product or computer program, which includes computer instructions stored in the aforementioned computer-readable storage medium; the processor of the aforementioned electronic device can read the computer instructions from the aforementioned computer-readable storage medium, and the processor executes the computer instructions, causing the electronic device to perform the aforementioned method embodiment.

[0070] It is understood that the content of the above method embodiments is applicable to this computer program product or computer program embodiment. The specific functions implemented by this computer program product or computer program embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0071] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0072] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0073] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0074] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0075] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0076] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application 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 this application 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 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.

[0077] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0078] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the 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 apparatuses or units may be electrical, mechanical, or other forms.

[0079] 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.

[0080] Furthermore, the functional units in the various embodiments of this application 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.

[0081] 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 this application, 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 multiple 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 of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0082] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A training method for a vehicle evaluation model, characterized in that, include: Obtain the test training set of the target vehicle under the target indicator item and the real label of the test training set. The test training set includes test sub-item data of several indicator sub-items and sub-item labels of each test sub-item data. The test training set is input into the first neural network for training to obtain a trained first neural network and several sub-item prediction data output by the trained first neural network. The real labels and all the predicted data of the sub-items are input into the second neural network for training to obtain the trained second neural network. A trained vehicle evaluation model is constructed based on the trained first neural network and the trained second neural network.

2. The method according to claim 1, characterized in that, Obtain the test training set, including: Obtain experimental test data for the target vehicle; Based on each sub-item of the target indicator, the experimental test data is cropped and aligned to obtain the test sub-item data and sub-item label for each sub-item. The test training set is constructed based on all the test sub-item data and the sub-item labels.

3. The method according to claim 1, characterized in that, The first neural network includes a feature extraction module and a prediction module. The test training set is input into the first neural network for training to obtain a trained first neural network, and the trained first neural network outputs several sub-item prediction data, including: The test training set is input into the feature extraction module for feature extraction to obtain several test features, each of which corresponds to a test sub-item data. Each of the test features is input into the prediction module to perform sub-item prediction, thereby obtaining the sub-item prediction data; Based on the sub-item label, a first loss value is calculated for the corresponding sub-item prediction data to obtain the first loss value; Based on the first loss value, the parameters of the first neural network are updated to obtain the trained first neural network.

4. The method according to claim 1, characterized in that, The first neural network includes a feature extraction module and a prediction module. The feature extraction module includes several extraction sub-modules. The test sub-item data includes several signal data, each signal data corresponding to one extraction sub-module. The test sub-item data and the sub-item labels are input into the first neural network for training to obtain a trained first neural network and the sub-item prediction data output by the trained first neural network, including: Each signal data is input into the corresponding extraction submodule for feature extraction, resulting in several signal features; All the signal features are input into the prediction module to perform sub-item prediction, and the sub-item prediction data is obtained. Based on the sub-item label, a second loss value is calculated for the corresponding sub-item prediction data to obtain the second loss value; Based on the second loss value, the parameters of the first neural network are updated to obtain the trained first neural network.

5. The method according to claim 3 or 4, characterized in that, The target data is input into the target module for feature extraction to obtain the target features, including: Local feature extraction is performed on the target data to obtain intermediate feature vectors; The intermediate feature vector is reduced in dimension and pooled to obtain the target feature; If the target data is a test training set, then the target module is a feature extraction module, and the target feature is a test feature; or, if the target data is signal data, then the target module is an extraction sub-module, and the target feature is a signal feature.

6. The method according to claim 1, characterized in that, The step of inputting the real labels and all the predicted sub-items into the second neural network for training to obtain the trained second neural network includes: A nonlinear transformation is performed on all the predicted sub-items to obtain the sub-item transformation features; The transformation features of the sub-items are evaluated and predicted to obtain the evaluation and prediction results corresponding to the target index item; Based on the actual labels, a third loss value is calculated on the evaluation prediction results to obtain the third loss value; Based on the third loss value, the parameters of the second neural network are updated to obtain the trained third neural network.

7. An evaluation method for a vehicle evaluation model, characterized in that, include: Obtain the vehicle test set for the vehicle to be evaluated; The vehicle test set is input into the trained vehicle evaluation model as described in any one of claims 1-6 to perform vehicle evaluation, thereby obtaining the vehicle evaluation result of the vehicle to be evaluated.

8. A training system for a vehicle evaluation model, characterized in that, include: The first processing unit is used to obtain the test training set of the vehicle under the target indicator item and the real label of the test training set. The test training set includes test sub-item data of several indicator sub-items and sub-item labels of each test sub-item data. The second processing unit is used to input the test training set into the first neural network for training, to obtain the trained first neural network and several sub-item prediction data output by the trained first neural network. The third processing unit is used to input the real labels and all the predicted data of the sub-items into the second neural network for training, so as to obtain the trained second neural network. The fourth processing unit is used to construct a trained vehicle evaluation model based on the trained first neural network and the trained second neural network.

9. An electronic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images