Vehicle weight estimation model training method and device, equipment and storage medium

Abstract

The invention discloses a vehicle weight estimation model training method and device, equipment and a storage medium, and relates to the technical field of computers. The method comprises the steps of obtaining pre-training sample data; based on the pre-training sample data, performing iterative training processing on a to-be-trained pre-training model to obtain a trained pre-training model, the pre-training model comprising an encoder network; obtaining a to-be-trained vehicle weight estimation model based on the full-connection network and the trained encoder network; obtaining training sample data, wherein the training sample data comprises training sample vehicle driving data and vehicle weight label data corresponding to the training sample vehicle driving data; and performing iterative training processing on the to-be-trained vehicle weight estimation model based on the training sample data to obtain a trained vehicle weight estimation model.

Description

Technical Field

[0001] This application relates to the field of computer technology, specifically to the fields of intelligent transportation and vehicle control, and particularly to a training method, apparatus, device, and storage medium for a vehicle weight estimation model. Background Technology

[0002] Accurately estimating vehicle weight is typically a key step in achieving efficient transportation management, improving vehicle safety, and optimizing fuel economy.

[0003] Currently, vehicle weight estimation schemes in related technologies can include calculating vehicle weight based on sensor signals and calculating vehicle weight based on neural networks trained with big data. However, these schemes still suffer from unreliable vehicle weight estimation results. Summary of the Invention

[0004] This application provides a training method, apparatus, device, and storage medium for a vehicle weight estimation model, which solves the problem of poor performance of vehicle weight estimation models. The technical solution is as follows:

[0005] Firstly, a method for training a vehicle weight estimation model is provided, the method comprising:

[0006] Obtain pre-training sample data;

[0007] Based on the pre-training sample data, the pre-training model to be trained is subjected to iterative training to obtain a pre-trained model that has been trained. The pre-trained model includes an encoder network.

[0008] Based on the fully connected network and the trained encoder network, a vehicle weight estimation model to be trained is obtained.

[0009] Acquire training sample data, which includes training sample vehicle driving data and vehicle weight label data corresponding to the training sample vehicle driving data;

[0010] Based on the training sample data, the vehicle weight estimation model to be trained is iteratively trained to obtain a trained vehicle weight estimation model.

[0011] In one possible implementation, obtaining the pre-training sample data includes:

[0012] Acquire historical vehicle bus data;

[0013] The historical vehicle bus data is filtered using a vehicle longitudinal dynamics model to obtain candidate pre-training sample data.

[0014] The candidate pre-training sample data is preprocessed to obtain the pre-training sample data.

[0015] In one possible implementation, the preprocessing of the candidate pre-training sample data to obtain the pre-training sample data includes:

[0016] The candidate pre-training sample data is resampled.

[0017] Based on a preset duration, the results of the resampling process are sliced ​​to obtain the resampling results corresponding to multiple time slices.

[0018] The results of the resampling process for each time slice are normalized to obtain the pre-training sample data.

[0019] In one possible implementation, the pre-trained model further includes a decoder network, and the iterative training process of the pre-trained model to be trained based on the pre-trained sample data to obtain a trained pre-trained model includes:

[0020] The encoder network is used to encode the pre-trained sample data to obtain intermediate feature data.

[0021] The intermediate feature data is decoded using the decoder network to obtain the output of the pre-trained model to be trained.

[0022] Based on the pre-training sample data and output results, the pre-training model to be trained is subjected to iterative training to obtain a pre-trained model that has been trained.

[0023] In one possible implementation, based on the training sample data, the vehicle weight estimation model to be trained is iteratively trained to obtain a trained vehicle weight estimation model, including:

[0024] The trained encoder network is used to encode the training sample data to obtain encoded feature data.

[0025] The fully connected network is used to process the encoded feature data to obtain the vehicle weight output by the vehicle weight estimation model to be trained.

[0026] Based on the training sample data and the vehicle weight output by the vehicle weight estimation model to be trained, the vehicle weight estimation model to be trained is iteratively trained to obtain a trained vehicle weight estimation model.

[0027] Secondly, a method for estimating vehicle weight is provided, the method comprising:

[0028] Acquire the vehicle driving data to be processed;

[0029] The vehicle driving data is input into the vehicle weight estimation model to obtain the vehicle weight output by the vehicle weight estimation model.

[0030] The vehicle weight estimation model is trained using a model training method.

[0031] Thirdly, a training device for a vehicle weight estimation model is provided, the device comprising:

[0032] The first acquisition unit is used to acquire pre-training sample data;

[0033] The first pre-training unit is used to perform iterative training on the pre-training model to be trained based on the pre-training sample data to obtain a pre-trained model that has been trained. The pre-training model includes an encoder network.

[0034] The first acquisition unit is used to obtain the vehicle weight estimation model to be trained based on the fully connected network and the trained encoder network.

[0035] The second acquisition unit is used to acquire training sample data, which includes training sample vehicle driving data and vehicle weight label data corresponding to the training sample vehicle driving data.

[0036] The first training unit is used to iteratively train the vehicle weight estimation model to be trained based on the training sample data, so as to obtain the trained vehicle weight estimation model.

[0037] Fourthly, a vehicle weight estimation device is provided, the device comprising:

[0038] The acquisition unit is used to acquire vehicle driving data to be processed.

[0039] The estimation unit is used to input the vehicle driving data into the vehicle weight estimation model and obtain the vehicle weight output by the vehicle weight estimation model.

[0040] The vehicle weight estimation model is trained using the aforementioned device.

[0041] Fifthly, a computer-readable storage medium is provided, wherein at least one instruction is stored therein, the at least one instruction being loaded and executed by a processor to implement the aspects described above and any possible implementation thereof.

[0042] Sixthly, an electronic device is provided, comprising:

[0043] At least one processor; and

[0044] A memory communicatively connected to the at least one processor; wherein,

[0045] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the methods described above and any possible implementations.

[0046] Seventhly, a new energy vehicle is provided, including the electronic devices described above.

[0047] The beneficial effects of the technical solution provided in this application include at least the following:

[0048] As can be seen from the above technical solution, on the one hand, the embodiments of this application can obtain pre-training sample data, and based on the pre-training sample data, perform iterative training on the pre-training model to be trained to obtain a pre-trained model that has been trained. The pre-training model includes an encoder network. Based on the fully connected network and the trained encoder network, a vehicle weight estimation model to be trained is obtained. Training sample data is obtained, including training sample vehicle driving data and vehicle weight label data corresponding to the training sample vehicle driving data. Based on the training sample data, the vehicle weight estimation model to be trained is iteratively trained to obtain a vehicle weight estimation model that has been trained. Since the encoder network of the vehicle weight estimation model to be trained can be trained first based on the pre-training sample data, and then the vehicle weight estimation model to be trained can be iteratively trained based on the training sample data with vehicle weight label data, a more accurate and effective vehicle weight estimation model can be obtained, improving the learning performance and effectiveness of the vehicle weight estimation model, thereby ensuring the reliability of the estimated weight of new energy vehicles.

[0049] As can be seen from the above technical solution, on the other hand, the embodiments of this application can obtain the vehicle driving data to be processed, input the vehicle driving data into the vehicle weight estimation model, and obtain the vehicle weight estimation result. Since the pre-trained vehicle weight estimation model has better vehicle weight estimation performance, using the vehicle weight estimation model to estimate the vehicle weight of the obtained vehicle driving data can ensure the accuracy of the vehicle weight estimation result, and can also directly output the vehicle weight estimation result corresponding to the vehicle driving data based on the vehicle weight estimation model, thereby reducing the computing power overhead of vehicle weight estimation, optimizing the performance and efficiency of vehicle weight estimation, and thus improving the reliability of vehicle weight estimation for new energy vehicles.

[0050] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0052] Figure 1 This is a flowchart illustrating the training method of a vehicle weight estimation model provided in one embodiment of this application;

[0053] Figure 2 This is a flowchart illustrating a vehicle weight estimation method provided in another embodiment of this application;

[0054] Figure 3 This is a flowchart illustrating a vehicle weight estimation method provided in another embodiment of this application;

[0055] Figure 4 This is a schematic diagram of the resampling principle in a vehicle weight estimation method provided in another embodiment of this application;

[0056] Figure 5 This is a schematic diagram of the architecture of an application scenario for a vehicle weight estimation method provided in another embodiment of this application;

[0057] Figure 6 This is a structural block diagram of a training device for a vehicle weight estimation model provided in another embodiment of this application;

[0058] Figure 7 This is a structural block diagram of a vehicle weight estimation device provided in another embodiment of this application;

[0059] Figure 8 This is a block diagram of an electronic device used to implement the vehicle weight estimation method and model training method of the embodiments of this disclosure. Detailed Implementation

[0060] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0061] Obviously, the described embodiments are only some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0062] It should be noted that the terminal devices involved in the embodiments of this application may include, but are not limited to, smart devices such as mobile phones, personal digital assistants (PDAs), wireless handheld devices, and tablet computers; the display devices may include, but are not limited to, personal computers, televisions, and other devices with display functions.

[0063] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0064] Currently, on the one hand, methods for calculating vehicle weight based on sensor signals are built upon the collaborative work of multiple sensors and precise mathematical models. This method utilizes multiple sensors, such as accelerometers and GPS positioning systems, to collect signals and capture dynamic data in real time during vehicle operation. After acquiring the data, the method establishes state transition equations based on the vehicle's longitudinal dynamics model, linking the vehicle's forces, motion state, and other elements through a rigorous mathematical model. To accurately extract effective information from the sensor data, Kalman filtering and least squares methods are also combined. Kalman filtering effectively suppresses noise interference, ensuring the purity and reliability of the input data; while least squares optimization algorithms find the best-fit curve among numerous data points, ultimately achieving accurate calculation of the vehicle's mass. On the other hand, methods based on training neural networks using big data leverage the vast data network of the Internet of Vehicles (IoV) to collect multi-dimensional signals such as accelerator pedal opening, output torque, vehicle speed, and vehicle weight during the vehicle's operation, and input these signals into a neural network model for training. Compared to traditional methods, neural network models trained on big data have stronger adaptability and generalization capabilities, enabling them to adapt to the weight estimation needs of different vehicle models and driving conditions. In the automotive manufacturing and R&D process, this method can help engineers quickly assess the weight changes of a vehicle under different configurations, optimize vehicle design, and thus improve R&D efficiency and product performance.

[0065] However, the first method calculates vehicle weight using a longitudinal dynamics model of the entire vehicle. This model often neglects nonlinear elements in the vehicle system (such as hysteresis in the transmission system), thus introducing systematic errors. Furthermore, acceleration sensors and GPS position signals often contain noise, leading to measurement errors. The combined effect of measurement and systematic errors results in a significant deviation between the final calculation and the true value, resulting in slow convergence. The second method calculates vehicle weight using a trained neural network model, and the accuracy of the result is directly determined by the performance of the neural network. Training a better neural network for vehicle weight calculation requires a large amount of high-quality data with ground truth labels. While connected vehicle platforms contain a wealth of real-world vehicle driving data, they often lack crucial ground truth labels. This prevents the effective utilization of a large amount of connected vehicle data, hindering the training of a high-performance neural network model.

[0066] Therefore, there is an urgent need for a training method for a vehicle weight estimation model that can optimize the processing performance and accuracy of the model, thereby ensuring the reliability of the vehicle weight estimation results.

[0067] Please refer to Figure 1 This document illustrates a flowchart of a training method for a vehicle weight estimation model according to an embodiment of this application. The training method for this vehicle weight estimation model may specifically include:

[0068] Step 101: Obtain pre-training sample data.

[0069] Step 102: Based on the pre-training sample data, perform iterative training on the pre-training model to be trained to obtain a pre-trained model that has been trained. The pre-trained model includes an encoder network.

[0070] Step 103: Based on the fully connected network and the trained encoder network, obtain the vehicle weight estimation model to be trained.

[0071] Step 104: Obtain training sample data, which includes training sample vehicle driving data and vehicle weight label data corresponding to the training sample vehicle driving data.

[0072] Step 105: Based on the training sample data, iteratively train the vehicle weight estimation model to be trained to obtain the trained vehicle weight estimation model.

[0073] It should be noted that the vehicle weight estimation model to be trained can be a model based on an autoencoder neural network. The pre-trained model to be trained can also be a model based on an autoencoder neural network.

[0074] It should be noted that the pre-training sample data can be sample data excluding vehicle weight label data. The training sample data can be sample data including vehicle weight label data.

[0075] It should be noted that the driving data of the sample vehicles may include, but is not limited to, engine output torque, vehicle speed, gradient, equivalent acceleration resistance of the whole vehicle, turning radius, etc.

[0076] It should be noted that some or all of the execution entities in steps 101 to 105 may be applications located on the local terminal, or may be functional units such as plug-ins or software development kits (SDKs) set in applications located on the local terminal, or may be processing engines located on network-side servers, or may be distributed systems located on the network side, such as processing engines or distributed systems in network-side model training platforms, etc. This embodiment does not impose any special limitations on these.

[0077] It is understood that the application may be a native program installed on the local terminal, or it may be a web application of a browser on the local terminal. This embodiment does not limit this.

[0078] In this way, by first training the encoder network of the vehicle weight estimation model to be trained based on pre-trained sample data, and then iteratively training the vehicle weight estimation model to be trained based on training sample data with vehicle weight labels, a more accurate and effective vehicle weight estimation model can be obtained, improving the learning performance and effectiveness of the vehicle weight estimation model, thereby ensuring the reliability of the estimated vehicle weight of new energy vehicles.

[0079] Optionally, in one possible implementation of this embodiment, in step 101, firstly, historical vehicle bus data is acquired. Secondly, the historical vehicle bus data is filtered using a vehicle longitudinal dynamics model to obtain candidate pre-training sample data. Thirdly, the candidate pre-training sample data is preprocessed to obtain the pre-training sample data.

[0080] In this implementation, historical vehicle bus data can be obtained from a vehicle-to-everything (V2X) big data platform. Historical vehicle bus data can include all vehicle driving data.

[0081] In this implementation, the candidate pre-training sample data can be vehicle driving data related to vehicle weight, selected from historical vehicle bus data. The candidate pre-training sample data may include engine output torque, gradient, vehicle speed, equivalent acceleration drag of the entire vehicle, and real-time turning radius of the vehicle.

[0082] In one specific implementation of this method, firstly, the candidate pre-training sample data can be resampled. Secondly, based on a preset duration, the result of the resampling process is sliced ​​to obtain the resampling results corresponding to multiple time slices. Thirdly, the resampling results corresponding to each time slice are normalized to obtain the pre-training sample data.

[0083] One specific implementation involves using a linear interpolation algorithm to resample the candidate pre-training sample data to obtain the resampling result.

[0084] Here, preferably, the sampling frequency can be 10 Hz.

[0085] Another specific implementation involves using a maximum-minimum normalization algorithm to normalize the resampling results for each time slice, obtaining a normalized result for the resampling results for each time slice, and using the normalized result for the resampling results for each time slice as the pre-training sample data.

[0086] In this implementation, the preset duration can be determined according to actual business needs. Preferably, the preset duration can be 30 seconds.

[0087] In this way, by filtering and preprocessing historical vehicle bus data, effective vehicle driving data related to vehicle weight can be obtained, which can further improve the performance of the subsequent vehicle weight estimation model and reduce the amount of data processed by the model learning, thereby improving the efficiency of model training.

[0088] Optionally, in one possible implementation of this embodiment, the pre-trained model may further include a decoder network. In step 102, firstly, the encoder network can be used to encode the pre-trained sample data to obtain intermediate feature data. Secondly, the decoder network can be used to decode the intermediate feature data to obtain the output result of the pre-trained model to be trained. Thirdly, based on the pre-trained sample data and the output result, the pre-trained model to be trained can be iteratively trained to obtain a pre-trained model that has been trained.

[0089] In this implementation, the pre-trained model may include an encoder network and a decoder network.

[0090] In this implementation, the output of the pre-trained model to be trained may include the vehicle acceleration corresponding to the predicted pre-training sample data.

[0091] In a specific implementation of this method, firstly, pre-training sample data is input into the encoder network of the pre-trained model to be trained, outputting low-dimensional intermediate feature data. Secondly, the intermediate feature data is input into the decoder network of the pre-trained model to obtain the output of the pre-trained model, which includes the predicted vehicle acceleration corresponding to the pre-training sample data. Thirdly, based on the pre-training sample data, vehicle acceleration verification data, i.e., real vehicle acceleration, is obtained. Fourthly, based on the vehicle acceleration verification data and the predicted vehicle acceleration corresponding to the pre-training sample data, the pre-trained model to be trained is updated. Fifthly, based on the pre-training sample data, the updated pre-trained model is iteratively trained until a preset termination condition is met, resulting in a trained pre-trained model. Finally, the trained encoder network in the trained pre-trained model is obtained.

[0092] Understandably, the training of the pre-trained model is unsupervised. During the pre-training process, the pre-trained model learns the relationship between vehicle speed, output torque, and vehicle weight. The trained encoder network has already learned the relationship between vehicle speed, output torque, and vehicle weight.

[0093] In this way, the encoder of the vehicle weight estimation model can be pre-trained using encoder and decoder networks based on training sample data without labels, so that a better-performing vehicle weight estimation model can be trained subsequently based on a small amount of training sample data with labels.

[0094] It should be noted that the specific implementation process provided in this embodiment can be combined with various specific implementation processes provided in the aforementioned implementation methods to implement the training method of the vehicle weight estimation model in this embodiment. Detailed descriptions can be found in the relevant content of the aforementioned implementation methods, and will not be repeated here.

[0095] Optionally, in one possible implementation of this embodiment, in step 105, firstly, the trained encoder network can be used to encode the training sample data to obtain encoded feature data. Secondly, the fully connected network can be used to process the encoded feature data to obtain the vehicle weight output by the vehicle weight estimation model to be trained. Thirdly, based on the training sample data and the vehicle weight output by the vehicle weight estimation model to be trained, the vehicle weight estimation model to be trained can be iteratively trained to obtain a trained vehicle weight estimation model.

[0096] In this implementation, the vehicle weight estimation model to be trained may include a trained encoder network and a fully connected network (FCN).

[0097] In this implementation, the driving data of the training sample vehicles may include engine output torque, vehicle speed, gradient, equivalent acceleration resistance of the whole vehicle, turning radius, etc.

[0098] Here, the training sample data can be obtained by preprocessing the training sample data to be processed. Preprocessing may include filtering, resampling, slicing, and normalization.

[0099] In a specific implementation of this method, firstly, training sample data to be processed is acquired. This training sample data includes training sample vehicle bus data and vehicle weight label data corresponding to the training sample vehicle bus data. Secondly, the training sample data to be processed is filtered using a vehicle longitudinal dynamics model to obtain candidate training sample data. Thirdly, the candidate training sample data is resampled using a linear interpolation algorithm to obtain the resampling result. Next, the resampling result is sliced ​​based on a preset time duration to obtain resampling results corresponding to multiple time slices. Finally, the resampling result for each time slice is normalized using a maximum-minimum normalization algorithm to obtain a normalized result for each time slice, and this normalized result is used as the training sample data.

[0100] In another specific implementation of this method, training sample data is input into the trained encoder network to obtain low-dimensional encoded feature data corresponding to the training sample data. Next, the encoded feature data is input into the fully connected network to obtain the vehicle weight output by the vehicle weight estimation model to be trained, i.e., the predicted vehicle weight. Then, based on the training sample data and the vehicle weight output by the vehicle weight estimation model to be trained, the vehicle weight estimation model to be trained can be iteratively trained to obtain a fully trained vehicle weight estimation model.

[0101] Understandably, the training of the vehicle weight estimation model here can be supervised training.

[0102] In this way, a vehicle weight estimation model can be trained using training sample data with vehicle weight labels and a pre-trained encoder network. This allows the vehicle weight estimation model to learn the relationship between vehicle speed, output torque, and vehicle weight, further improving the performance and reliability of the vehicle weight estimation model.

[0103] It should be noted that the specific implementation process provided in this embodiment can be combined with various specific implementation processes provided in the aforementioned implementation methods to implement the training method of the vehicle weight estimation model in this embodiment. Detailed descriptions can be found in the relevant content of the aforementioned implementation methods, and will not be repeated here.

[0104] Figure 2 This is a flowchart illustrating a vehicle weight estimation method provided in another embodiment of this application, as shown below. Figure 2 As shown.

[0105] Step 201: Obtain the vehicle driving data to be processed.

[0106] Step 202: Input the vehicle driving data into the vehicle weight estimation model to obtain the vehicle weight output by the vehicle weight estimation model.

[0107] It should be noted that the vehicle weight estimation model is pre-trained using the same training method as the vehicle weight estimation model described in the aforementioned embodiments. The vehicle weight estimation model can be a model based on an autoencoder neural network.

[0108] It should be noted that vehicle driving data may include, but is not limited to, engine output torque, vehicle speed, gradient, equivalent acceleration resistance of the whole vehicle, turning radius, etc.

[0109] It should be noted that the execution entity of steps 201 to 202 may be part or all of an application located on a local terminal, or it may be a plugin or software development kit (SDK) or other functional unit set in an application located on a local terminal, or it may be a processing engine located on a network-side server, or it may be a distributed system located on the network side, such as a processing engine or distributed system in a vehicle weight estimation platform on the network side. This embodiment does not impose any particular limitations on this.

[0110] It is understood that the application may be a native program installed on the local terminal, or it may be a web application of a browser on the local terminal. This embodiment does not limit this.

[0111] In this way, the pre-trained vehicle weight estimation model has better vehicle weight estimation performance. By using the vehicle weight estimation model to estimate the vehicle weight of the acquired vehicle driving data, the accuracy of the vehicle weight estimation results can be guaranteed. At the same time, the vehicle weight estimation model can directly output the vehicle weight estimation results corresponding to the vehicle driving data, reducing the computational cost of vehicle weight estimation, optimizing the performance and efficiency of vehicle weight estimation, and thus improving the reliability of vehicle weight estimation for new energy vehicles.

[0112] Optionally, in one possible implementation of this embodiment, firstly, vehicle bus data to be processed is acquired. Secondly, the vehicle bus data to be processed is filtered using a vehicle longitudinal dynamics model to obtain candidate vehicle driving data. Thirdly, the candidate vehicle driving data is preprocessed to obtain the actual vehicle driving data.

[0113] In this implementation, the vehicle bus data can be obtained from a vehicle-to-everything (V2X) big data platform or from the vehicle itself. The vehicle bus data can include vehicle position, speed, orientation, engine output torque, gradient, vehicle speed, equivalent acceleration drag, real-time turning radius, and other driving data.

[0114] In this implementation, the candidate vehicle driving data can be vehicle driving data related to vehicle weight selected from the vehicle bus data. The candidate vehicle driving data may include total engine output torque, gradient, vehicle speed, equivalent acceleration drag of the whole vehicle, real-time turning radius of the vehicle, etc.

[0115] In one specific implementation of this method, firstly, the candidate vehicle driving data can be resampled. Secondly, based on a preset duration, the result of the resampling process is sliced ​​to obtain resampling results corresponding to multiple time slices. Thirdly, the resampling result corresponding to each time slice is normalized to obtain the vehicle driving data.

[0116] One specific implementation involves using a linear interpolation algorithm to resample the candidate pre-training sample data to obtain the resampling result.

[0117] Another specific implementation method is to use the maximum-minimum normalization algorithm to normalize the resampling results corresponding to each time slice, and then use the normalized results of the resampling results corresponding to each time slice as the vehicle driving data.

[0118] It is understood that the vehicle driving data includes vehicle driving data corresponding to multiple time slices of the same duration.

[0119] In this way, by filtering and preprocessing the vehicle bus data, effective vehicle driving data related to vehicle weight can be obtained, which can further improve the accuracy of the vehicle weight prediction results of the subsequent vehicle weight estimation model.

[0120] Optionally, in one possible implementation of this embodiment, the vehicle weight estimation model may include an encoder network and a fully connected network.

[0121] In one specific implementation of this method, vehicle driving data is input into the encoder network of the vehicle weight estimation model to obtain low-dimensional encoded feature data corresponding to the vehicle driving data. Then, the encoded feature data is input into the fully connected network of the vehicle weight estimation model to obtain the vehicle weight corresponding to the vehicle driving data output by the vehicle weight estimation model, i.e., the predicted vehicle weight.

[0122] It should be noted that the specific implementation process provided in this embodiment can be combined with various specific implementation processes provided in the aforementioned implementation methods to realize the vehicle weight estimation method of this embodiment. Detailed descriptions can be found in the relevant content of the aforementioned implementation methods, and will not be repeated here.

[0123] To better understand the methods of the embodiments of this application, the methods of the embodiments of this application will be described below in conjunction with the accompanying drawings and specific application scenarios.

[0124] Figure 3 This is a flowchart illustrating a vehicle weight estimation method provided in another embodiment of this application, as shown below. Figure 3 As shown. In this embodiment, the vehicle weight estimation method may include:

[0125] Step 301: Obtain historical vehicle bus data.

[0126] In this implementation, historical vehicle bus data can be vehicle driving data uploaded to the vehicle-to-everything (V2X) big data platform via the bus. This historical vehicle bus data can be stored in the V2X big data platform. The V2X big data platform can contain massive amounts of vehicle driving data.

[0127] Step 302: Using the vehicle longitudinal dynamics model, the historical vehicle bus data is filtered to obtain candidate pre-training sample data.

[0128] In this implementation, driving data related to vehicle weight calculation can be filtered from historical vehicle bus data.

[0129] Here, in order to obtain better driving data related to vehicle weight, the longitudinal dynamics model of the vehicle can be used for filtering.

[0130] Preferably, the longitudinal dynamics model of the vehicle can be expressed as formula (1):

[0131]

[0132] Among them, F trac For the driving force of the whole vehicle, F road F represents the rolling resistance of the entire vehicle. aero For the overall vehicle air resistance, F grad For the overall vehicle slope resistance, It is the total vehicle mass multiplied by the total vehicle acceleration.

[0133] Furthermore, the driving force of the entire vehicle can be expressed as formula (2):

[0134] F trac =T mot irη (2)

[0135] Among them, T mot Let T be the total engine output torque, i be the overall reduction ratio of the transmission system, r be the wheel end radius, and η be the transmission system efficiency. Here, the last three coefficients can be considered constants. Therefore, based on the vehicle's driving force, the total engine output torque T needs to be obtained from the bus data. mot .

[0136] Furthermore, the rolling resistance of the entire vehicle can be expressed as formula (3):

[0137] F road =mgcosθf (3)

[0138] Where m is the mass of the vehicle, g is the acceleration due to gravity, θ is the gradient, and f is the rolling resistance coefficient. The acceleration due to gravity and the rolling resistance coefficient can be considered constants; therefore, based on the vehicle's rolling resistance, the gradient θ needs to be obtained from the bus data.

[0139] Furthermore, the overall vehicle air resistance can be expressed as formula (4):

[0140]

[0141] Where A is the longitudinal projected area of ​​the vehicle, ρ is the air density, and C d Let v be the drag coefficient and v be the vehicle speed. Assuming the vehicle's projected area, air density, and drag coefficient are constants, the vehicle speed v needs to be obtained from the bus data based on the vehicle's air resistance.

[0142] Furthermore, the overall vehicle slope resistance can be expressed by formula (5):

[0143] F grad =mgsinθ (5)

[0144] Where θ is the slope degree, m is the mass of the vehicle, and g is the gravitational acceleration. Based on the vehicle's slope resistance, the slope degree θ needs to be obtained from the bus data.

[0145] Finally, optionally, the equivalent acceleration drag of the entire vehicle. It can be obtained by taking the derivative of the speed of the whole vehicle.

[0146] Understandably, based on the vehicle's longitudinal dynamics equations, the bus data required for the vehicle mass calculation neural network can be selected, namely, vehicle driving data related to vehicle weight. Furthermore, the vehicle's longitudinal dynamics equations only consider the straight-line driving condition; however, in reality, vehicles frequently encounter turning situations, which causes changes in vehicle drag. Therefore, the vehicle's real-time turning radius R needs to be added as reference data.

[0147] At this point, the pre-training sample data can include engine output torque, vehicle speed, gradient, equivalent acceleration resistance of the whole vehicle, turning radius, etc.

[0148] Step 303: Preprocess the candidate pre-training sample data to obtain pre-training sample data.

[0149] Understandably, model training requires continuous, uniform, and synchronous sample data. However, bus data obtained from the bus often has timestamps at different time intervals, and the data of different categories vary greatly in size. Therefore, it is necessary to preprocess the data obtained from the bus.

[0150] In this embodiment, data at different timestamps can be resampled to fit the size of different types of data at the same time. Figure 4 This is a schematic diagram of the resampling principle in a vehicle weight estimation method provided in another embodiment of this application, as shown below. Figure 4 As shown. The original data can be the filtered bus data, i.e., candidate pre-training sample data. During the sampling process, linear interpolation is used for resampling at a sampling frequency of 10Hz, and the resampled values ​​are sampled at equal intervals. By fitting the original data and the resampled values, the result of the resampling process can be obtained.

[0151] Here, resampling based on linear interpolation can ensure the consistency between the original data and the sampled data. It can avoid the problem that excessively high sampling frequency will add redundant data to the neural network, while excessively low sampling frequency will lead to data distortion, thus improving the reliability of the resampling process.

[0152] In this embodiment, after data resampling, further preprocessing is required for model training. To achieve batch training, the resampled pre-training sample data can be sliced ​​based on a preset time interval. The pre-training sample data used as model input can include pre-training sample data corresponding to multiple time segments.

[0153] Here, regarding the choice of slice length, too long a time will result in fewer usable training samples, while too short a time will result in insufficient information contained in a single sample. Therefore, 30 seconds was chosen as the slice length.

[0154] In this embodiment, optionally, the pre-training sample data for slice processing can be normalized using a maximum-minimum normalization algorithm.

[0155] Understandably, vehicle driving data generally doesn't conform to a normal distribution, and because this is actual data generated by the entire vehicle's operation, it often has strict maximum and minimum value limits. Vehicles themselves have pre-configured maximum and minimum performance limits; for example, regulations require heavy-duty tractor vehicles to have a maximum speed of 89 km / h. In such cases, a maximum-minimum normalization method can be used to normalize the data.

[0156] In this way, when performing maximum and minimum value normalization, the maximum and minimum values ​​can be obtained directly from the vehicle's own overall performance data instead of directly from the bus data, which can ensure that the distribution of training samples and validation data samples is more uniform.

[0157] Step 304: Input the pre-trained sample data into the encoder network of the pre-trained model to be trained, and output the intermediate feature data.

[0158] Step 305: Input the intermediate feature data into the decoder network of the pre-trained model to be trained, and output the predicted vehicle acceleration.

[0159] Step 306: Based on the pre-training sample data and the predicted vehicle acceleration, perform iterative training on the pre-training model to be trained to obtain the pre-trained model and the encoder network after training.

[0160] In this embodiment, Figure 5 This is a schematic diagram illustrating the architecture of an application scenario for a vehicle weight estimation method provided in another embodiment of this application, as shown below. Figure 5 As shown. In this application scenario, the training of the vehicle weight estimation model can include pre-training and formal training.

[0161] Here, unsupervised pre-training can be performed on a pre-trained model based on an autoencoder neural network. This pre-trained model uses an encoder-decoder structure as the network core. Based on bus data, pre-training sample data from multiple time slices is obtained. The encoder network takes as input any one time slice's pre-training sample data and outputs a low-dimensional encoding, i.e., intermediate feature data. The decoder network takes as input the encoder network's output encoding and outputs the predicted vehicle acceleration corresponding to the pre-training sample data in the input time slice. Therefore, all driving data during pre-training can be obtained from a vehicle-to-everything (V2X) big data platform, allowing for unsupervised training without the need for additional ground truth vehicle weight values. During pre-training, the encoder network first learns the relationship between vehicle power and speed, i.e., vehicle weight information, and embeds this weight information in the encoder's output encoding. The decoder then reconstructs the predicted vehicle acceleration within the time slice based on the weight information embedded in the encoding. The error between this predicted acceleration and the actual vehicle acceleration is calculated and backpropagated to complete the pre-training of the vehicle weight calculation neural network. This pre-trained model does not require additional ground truth vehicle weight data for training; it can directly utilize existing data to learn the relationship between vehicle speed, output torque, and vehicle weight. Here, the real vehicle acceleration can be obtained by taking the first-order forward differential of the real vehicle speed in the acquired sample data.

[0162] Step 307: Based on the trained encoder network and fully connected network, construct the vehicle weight estimation model to be trained.

[0163] Step 308: Obtain training sample data, which includes training sample vehicle driving data and vehicle weight label data corresponding to the training sample vehicle driving data.

[0164] Step 309: Input the training sample data into the trained encoder network to obtain the low-dimensional encoded feature data corresponding to the training sample data.

[0165] Step 310: Input the encoded feature data into the fully connected network to obtain the vehicle weight output by the vehicle weight estimation model to be trained.

[0166] Step 311: Based on the training sample data and the vehicle weight output by the vehicle weight estimation model to be trained, iteratively train the vehicle weight estimation model to be trained to obtain the trained vehicle weight estimation model.

[0167] In this embodiment, as Figure 5As shown, the pre-trained encoder network is transferred to the vehicle weight estimation model to be trained for transfer training. Based on the trained encoder network, a fully connected network (FCN) is added to form the vehicle weight estimation model to be trained for the next step of supervised training. Based on experimental data, training sample data from multiple time slices is obtained. The vehicle driving data in the training sample data has corresponding ground truth vehicle weight labels. The training sample data is input into the transfer-trained encoder network to obtain the corresponding encoding. This encoding is then input into the FCN to output the predicted vehicle weight. Based on the training sample data and the predicted vehicle weight, the vehicle weight estimation model to be trained is iteratively trained to obtain the trained vehicle weight estimation model. Here, since the encoder can already obtain the relationship between vehicle speed, output torque, and vehicle weight during pre-training, only a relatively small amount of training sample data with ground truth vehicle weight labels is needed for formal training to obtain a high-performance vehicle weight estimation model.

[0168] Furthermore, it's understandable that the training sample data for formal training is also filtered and preprocessed. Here, data slices of the same duration may contain varying amounts of high-quality data for calculating vehicle weight. Naturally, data slices containing high-quality information will yield more accurate vehicle weight results. To allow the network to evaluate the accuracy of its predictions, the vehicle weight estimation model outputs a probability distribution of predicted vehicle weights. The mean of this probability distribution is the vehicle weight predicted by the network, while the variance represents the accuracy of the prediction as determined by the vehicle weight estimation model.

[0169] Step 312: Obtain the vehicle driving data to be processed.

[0170] In this embodiment, the vehicle driving data to be processed may include real-time vehicle driving data used to estimate vehicle weight.

[0171] Step 313: Input the vehicle driving data to be processed into the vehicle weight estimation model, and output the vehicle weight corresponding to the vehicle driving data to be processed.

[0172] At this point, the trained vehicle weight estimation model can be used to predict the vehicle weight corresponding to the real-time vehicle driving data.

[0173] Thus, by adopting the technical solution in this embodiment, a vehicle weight estimation model that effectively learns the relationship between vehicle speed, output torque and vehicle weight can be obtained by using a certain amount of unlabeled pre-training sample data and a small amount of labeled training sample data for unsupervised pre-training and supervised formal training, respectively.

[0174] Furthermore, by adding probabilistic information as vehicle weight labels to the network during supervised formal training, enabling it to evaluate the accuracy of each model prediction, the accuracy of the calculation result can be estimated simultaneously with the calculation of the vehicle weight when training the vehicle weight estimation model.

[0175] Furthermore, by adopting the technical solution in this embodiment, the original bus data of the vehicle obtained from the Internet of Vehicles can be filtered according to the longitudinal dynamic equation of the whole vehicle to obtain training sample data related to vehicle weight, so as to train the vehicle weight estimation model and improve the reliability of model training.

[0176] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0177] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0178] Figure 6 This paper shows a structural block diagram of a training device for a vehicle weight estimation model according to an embodiment of this application, as follows: Figure 6 As shown. The training device 600 for the vehicle weight estimation model in this embodiment may include a first acquisition unit 601, a first pre-training unit 602, a first acquisition unit 603, a second acquisition unit 604, and a first training unit 605. Specifically, the first acquisition unit 601 is used to acquire pre-training sample data; the first pre-training unit 602 is used to iteratively train the pre-trained model to be trained based on the pre-training sample data to obtain a pre-trained model that has been trained, the pre-trained model including an encoder network; the first acquisition unit 603 is used to obtain a vehicle weight estimation model to be trained based on a fully connected network and the trained encoder network; the second acquisition unit 604 is used to acquire training sample data, the training sample data including training sample vehicle driving data and vehicle weight label data corresponding to the training sample vehicle driving data; and the first training unit 605 is used to iteratively train the vehicle weight estimation model to be trained based on the training sample data to obtain a vehicle weight estimation model that has been trained.

[0179] Optionally, in one possible implementation of this embodiment, the first acquisition unit 601 is used to acquire historical vehicle bus data; to filter the historical vehicle bus data using a vehicle longitudinal dynamics model to obtain candidate pre-training sample data; and to preprocess the candidate pre-training sample data to obtain the pre-training sample data.

[0180] Optionally, in one possible implementation of this embodiment, the first acquisition unit 601 is used to resample the candidate pre-training sample data; slice the result of the resampling process based on a preset duration to obtain the resampling process results corresponding to multiple time slices; and normalize the resampling process results corresponding to each time slice to obtain the pre-training sample data.

[0181] Optionally, in one possible implementation of this embodiment, the pre-trained model further includes a decoder network and a first pre-training unit 602, which is used to encode the pre-training sample data using the encoder network to obtain intermediate feature data; to decode the intermediate feature data using the decoder network to obtain the output result of the pre-trained model to be trained; and to perform iterative training on the pre-trained model to be trained based on the pre-training sample data and the output result to obtain a pre-trained model that has been trained.

[0182] Optionally, in one possible implementation of this embodiment, the first training unit 605 is used to encode the training sample data using the trained encoder network to obtain encoded feature data; to process the encoded feature data using the fully connected network to obtain the vehicle weight output by the vehicle weight estimation model to be trained; and to iteratively train the vehicle weight estimation model to be trained based on the training sample data and the vehicle weight output by the vehicle weight estimation model to be trained to obtain the trained vehicle weight estimation model.

[0183] In this embodiment, pre-training sample data can be acquired by a first acquisition unit. Then, the first pre-training unit can iteratively train the pre-training model to be trained based on the pre-training sample data to obtain a trained pre-training model. The pre-training model includes an encoder network. The first acquisition unit obtains a vehicle weight estimation model to be trained based on a fully connected network and the trained encoder network. The second acquisition unit acquires training sample data, which includes training sample vehicle driving data and vehicle weight label data corresponding to the training sample vehicle driving data. The first training unit iteratively trains the vehicle weight estimation model to be trained based on the training sample data to obtain a trained vehicle weight estimation model. Since the encoder network of the vehicle weight estimation model to be trained can be trained first based on the pre-training sample data, and then the vehicle weight estimation model to be trained can be iteratively trained based on the training sample data with vehicle weight label data, a more accurate and effective vehicle weight estimation model can be obtained, improving the learning performance and effectiveness of the vehicle weight estimation model, thereby ensuring the reliability of the estimated weight of new energy vehicles.

[0184] Figure 7 A structural block diagram of a vehicle weight estimation device according to an embodiment of this application is shown, as follows: Figure 7 As shown. The vehicle weight estimation device 700 in this embodiment may include an acquisition unit 701 and an estimation unit 702. The acquisition unit 701 is used to acquire vehicle driving data to be processed; the estimation unit 702 is used to input the vehicle driving data into a vehicle weight estimation model to obtain the vehicle weight output by the vehicle weight estimation model; the vehicle weight estimation model is trained using the aforementioned vehicle weight estimation model training device.

[0185] In this embodiment, the vehicle driving data to be processed is acquired by the third acquisition unit, and then the first estimation unit inputs the vehicle driving data into the vehicle weight estimation model to obtain the vehicle weight output by the vehicle weight estimation model. Since the pre-trained vehicle weight estimation model has better vehicle weight estimation performance, using the vehicle weight estimation model to estimate the vehicle weight of the acquired vehicle driving data can ensure the accuracy of the vehicle weight estimation result, and can also directly output the vehicle weight estimation result corresponding to the vehicle driving data based on the vehicle weight estimation model, thereby reducing the computing power overhead of vehicle weight estimation, optimizing the performance and efficiency of vehicle weight estimation, and thus improving the reliability of vehicle weight estimation for new energy vehicles.

[0186] The technical solution of this application involves the collection, storage, use, processing, transmission, provision, and disclosure of user personal information, such as user image and attribute data, which comply with relevant laws and regulations and do not violate public order and good morals.

[0187] According to embodiments of this application, this application also provides an electronic device, a readable storage medium, and a computer program product.

[0188] According to embodiments of this application, a new energy vehicle including the provided electronic equipment is further provided. This new energy vehicle may include Level 2 or higher autonomous driving vehicles. For example, it may be an autonomous commercial vehicle, an autonomous logistics vehicle, or an autonomous logistics vehicle.

[0189] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of this application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the application described and / or claimed herein.

[0190] like Figure 8 As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. The RAM 803 may also store various programs and data required for the operation of the electronic device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0191] Multiple components in electronic device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of displays, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows electronic device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0192] The computing unit 801 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as methods for vehicle weight estimation and methods for model training. For example, in some embodiments, the methods for vehicle weight estimation and methods for model training can be implemented as computer software programs tangibly contained in a machine-readable medium, such as storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the methods for vehicle weight estimation and methods for model training described above can be performed. Alternatively, in other embodiments, the computing unit 801 may be configured in any other suitable manner (e.g., by means of firmware) to perform methods for vehicle weight estimation and model training.

[0193] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0194] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0195] In the context of this application, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0196] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0197] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0198] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0199] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this application can be achieved, and this is not limited herein.

[0200] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A training method of a vehicle weight estimation model, characterized by, The method comprises: obtaining pre-training sample data; based on the pre-training sample data, the pre-training model to be trained is iteratively trained to obtain a trained pre-training model, the pre-training model comprising an encoder network; based on the full connection network and the trained encoder network, a vehicle weight estimation model to be trained is obtained; obtain training sample data, the training sample data comprising training sample vehicle driving data and vehicle weight label data corresponding to the training sample vehicle driving data; based on the training sample data, the vehicle weight estimation model to be trained is iteratively trained to obtain a trained vehicle weight estimation model.

2. The method of claim 1, wherein, The method comprises: obtaining historical vehicle bus data; using a vehicle longitudinal dynamics model, the historical vehicle bus data is filtered to obtain candidate pre-training sample data; the candidate pre-training sample data is preprocessed to obtain the pre-training sample data.

3. The method of claim 2, wherein, The method comprises: resampling the candidate pre-training sample data; based on a preset time length, the resampling result is sliced to obtain a plurality of time slice corresponding resampling results; the resampling result corresponding to each time slice is normalized to obtain the pre-training sample data.

4. The method of claim 1, wherein, The pre-training model further comprises a decoder network, and the method comprises: using the encoder network, the pre-training sample data is encoded to obtain intermediate feature data; using the decoder network, the intermediate feature data is decoded to obtain the output result of the pre-training model to be trained; based on the pre-training sample data and the output result, the pre-training model to be trained is iteratively trained to obtain a trained pre-training model.

5. The method of claim 1, wherein, based on the training sample data, the vehicle weight estimation model to be trained is iteratively trained to obtain a trained vehicle weight estimation model, comprising: using the trained encoder network, the training sample data is encoded to obtain encoded feature data; using the full connection network, the encoded feature data is processed to obtain the vehicle weight output by the vehicle weight estimation model to be trained; based on the training sample data and the vehicle weight output by the vehicle weight estimation model to be trained, the vehicle weight estimation model to be trained is iteratively trained to obtain a trained vehicle weight estimation model.

6. A method of estimating the weight of a vehicle, characterized in that The method comprises: obtaining vehicle driving data to be processed; the vehicle driving data is input into the vehicle weight estimation model to obtain the vehicle weight output by the vehicle weight estimation model; wherein the vehicle weight estimation model is trained by the method of any one of claims 1 to 5.

7. A training device for a vehicle weight estimation model, characterized in that, The device comprises: a first obtaining unit configured to obtain pre-training sample data; The first pre-training unit is configured to perform iterative training on the pre-training model to be trained based on the pre-training sample data, to obtain a trained pre-training model, wherein the pre-training model comprises an encoder network. The first obtaining unit is configured to obtain a vehicle weight estimation model to be trained based on the full connection network and the trained encoder network. The second obtaining unit is configured to obtain training sample data, wherein the training sample data comprises training sample vehicle driving data and vehicle weight label data corresponding to the training sample vehicle driving data. The first training unit is configured to perform iterative training on the vehicle weight estimation model to be trained based on the training sample data, to obtain a trained vehicle weight estimation model.

8. An apparatus for estimating a vehicle weight, characterized by The device comprises: An obtaining unit configured to obtain vehicle driving data to be processed; An estimation unit configured to input the vehicle driving data into a vehicle weight estimation model, to obtain vehicle weight output by the vehicle weight estimation model; The vehicle weight estimation model is trained by the device of claim 7.

9. An electronic device comprising: at least one processor; and a memory connected to the at least one processor in communication; wherein The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1-6.

10. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to make the computer execute the method according to any one of claims 1-6.

11. A new energy vehicle comprising the electronic device of claim 9.

Images