Vehicle driving condition model training method, device, equipment, medium and product

Abstract

The invention discloses a vehicle driving condition model training method, device and equipment, a medium and a product. The method comprises the following steps: acquiring driving sample data and a corresponding standard driving condition category; determining a current model hyper-parameter of the initial network model in the current iteration period; performing model training on the initial network model based on the current model hyper-parameter of the initial network model according to the driving sample data and the corresponding standard driving condition category to obtain a trained current driving condition model in the current iteration period; and performing model testing on the current driving condition model, and if a model testing result meets a preset model testing condition, determining the current driving condition model as a vehicle driving condition model. According to the technical scheme of the embodiment of the invention, the training precision of the vehicle driving condition model is improved, and then the accuracy of road condition recognition in the vehicle driving process is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of automotive communication, and particularly to a control method, device, equipment, storage medium and product for vehicle satellite positioning. Background Art

[0002] The commercial vehicle mechanical automatic transmission (AMT, Automated Manual Transmission) realizes functions such as gear selection and clutch control through a transmission control unit (TCU) on the basis of a manual transmission (MT, Manual Transmission), and the control strategy highly depends on complex and changeable driving conditions. Therefore, accurately identifying the driving conditions is crucial for strategies such as transmission gear selection and clutch control.

[0003] Currently, the identification of existing vehicle driving conditions is mainly based on the method of model training, which highly depends on model empirical parameters and a large amount of real sample data. However, there are obvious deficiencies in both sample construction and model parameter selection for the existing training of the driving condition recognition model, resulting in low model accuracy of the trained driving condition recognition model, thereby reducing the accuracy of road condition recognition during vehicle driving. Summary of the Invention

[0004] The present invention provides a method, device, equipment, medium and product for training a vehicle driving condition model to improve the training accuracy of the vehicle driving condition model, and further improve the accuracy of road condition recognition during vehicle driving.

[0005] According to one aspect of the present invention, a method for training a vehicle driving condition model is provided, the method comprising:

[0006] Obtaining driving sample data and its corresponding standard driving condition categories;

[0007] Determining current model hyperparameters of an initial network model in a current iteration cycle;

[0008] Based on the driving sample data and its corresponding standard driving condition categories, and based on the current model hyperparameters of the initial network model, training the initial network model to obtain a currently trained current driving condition model in the current iteration cycle;

[0009] Testing the current driving condition model, and if the model test result meets a preset model test condition, determining the current driving condition model as a vehicle driving condition model; the vehicle driving condition model is used to predict the driving road conditions during vehicle driving.

[0010] According to another aspect of the present invention, there is provided a vehicle driving condition model training device, the device comprising:

[0011] A driving condition category acquisition module for acquiring driving sample data and its corresponding standard driving condition category;

[0012] A hyperparameter determination module for determining the current model hyperparameters of the initial network model in the current iteration cycle;

[0013] A model training module for training the initial network model based on the driving sample data and its corresponding standard driving condition category and the current model hyperparameters of the initial network model to obtain the currently trained current driving condition model in the current iteration cycle;

[0014] A model testing module for testing the current driving condition model, and if the model test result meets the preset model test conditions, determining the current driving condition model as the vehicle driving condition model; the vehicle driving condition model is used to predict the driving road conditions during vehicle driving.

[0015] According to another aspect of the present invention, there is provided an electronic device, the electronic device comprising:

[0016] At least one processor; and

[0017] A memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the vehicle driving condition model training method according to any embodiment of the present invention.

[0019] According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing a processor to implement the vehicle driving condition model training method according to any embodiment of the present invention when executed.

[0020] According to another aspect of the present invention, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the above-mentioned vehicle driving condition model training method.

[0021] In the technical solution of the embodiment of the present invention, the current model hyperparameters of the initial network model in the current iteration period are determined, and based on the current model hyperparameters of the initial network model and the driving sample data and their corresponding standard driving condition categories, the initial network model is trained until the preset model test conditions are met, and a vehicle driving condition model is obtained. By continuously iteratively optimizing the model hyperparameters during the model training process of the vehicle driving condition model, the above technical solution trains the vehicle driving condition model based on the continuously optimized model hyperparameters, making the model training more accurate, significantly improving the model performance, efficiency and robustness. The method of training the model based on the continuously optimized model hyperparameters and vehicle driving condition data improves the training accuracy of the vehicle driving condition model, and further improves the accuracy of identifying road conditions during vehicle driving.

[0022] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0024] Figure 1 is a flowchart of a method for training a vehicle driving condition model according to Embodiment 1 of the present invention;

[0025] Figure 2 is a flowchart of a method for training a vehicle driving condition model according to Embodiment 2 of the present invention;

[0026] Figure 3 is a schematic structural diagram of a device for training a vehicle driving condition model according to Embodiment 3 of the present invention;

[0027] Figure 4 is a schematic structural diagram of an electronic device for implementing the method for training a vehicle driving condition model of the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] To enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings 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 under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0030] Embodiment 1

[0031] Figure 1 It is a flowchart of a method for training a vehicle driving condition model provided in Embodiment 1 of the present invention. This embodiment is applicable to the situation of identifying road conditions during the driving of commercial vehicles. This method can be executed by a vehicle driving condition model training device, which can be implemented in the form of hardware and / or software, and the vehicle driving condition model training device can be configured in an electronic device. As Figure 1 shown, the method includes:

[0032] S110. Obtain driving sample data and its corresponding standard driving condition categories.

[0033] S120. Determine the current model hyperparameters of the initial network model in the current iteration cycle.

[0034] S130. Based on the driving sample data and its corresponding standard driving condition categories, and based on the current model hyperparameters of the initial network model, train the initial network model to obtain the current driving condition model that has been trained in the current iteration cycle.

[0035] S140. Perform model testing on the current driving condition model. If the model testing result meets the preset model testing conditions, then determine the current driving condition model as the vehicle driving condition model; the vehicle driving condition model is used to predict the driving road conditions during vehicle driving.

[0036] Among them, the driving sample data is the driving data generated during the vehicle driving process required for model training. The standard driving condition category corresponding to the driving sample data is the real condition category, such as urban roads, provincial roads, and highway roads, etc.

[0037] The driving sample data and its corresponding standard driving condition category can be pre-constructed or determined. In an alternative embodiment, before obtaining the driving sample data and its corresponding standard driving condition category, it further includes:

[0038] Step a: Obtain the vehicle driving data, perform feature field processing on the vehicle driving data, and determine the target feature field.

[0039] Step b: Based on the target feature field and the vehicle driving data, construct the driving sample data.

[0040] Step c: Determine the standard driving condition category corresponding to the driving sample data.

[0041] Among them, the vehicle driving data may include speed-related data, torque-related data, rotation speed-related data, pedal-related data, etc. at different driving time points or time periods. The vehicle driving data can be data collected by the vehicle networking T-Box (Telematics BOX) of different commercial vehicles. For example, the screening range can be the driving data of 30 heavy commercial vehicles within 3 months.

[0042] Exemplarily, the fields in the vehicle driving data can be extracted to determine all non-repeated fields included in the vehicle driving data, and the extracted fields are used as the target feature fields.

[0043] Further, since the target feature field is used for subsequent model training, in an alternative embodiment, to further ensure the rationality of field screening, performing feature field processing on the vehicle driving data to determine the target feature field includes:

[0044] Step a1: Perform feature field processing on the vehicle driving data to determine the initial feature field.

[0045] Step a2: Determine the field data quality of the initial feature field, and based on the field data quality, determine the intermediate feature field.

[0046] Step a3: Based on the field correlation between the intermediate feature fields, determine the target feature field.

[0047] Specifically, determine the fields included in the vehicle driving data. Exemplarily, field extraction can be performed through field extraction functions in existing third-party libraries for data operation and analysis. For example, read the vehicle driving data through the read_excel() function provided by the open-source third-party library Pandas, and collect all the fields included in the vehicle driving data through the columns() function to obtain the initial feature fields.

[0048] Among them, the field data quality can be the number of times a field appears in the vehicle driving data. For example, the vehicle driving data includes the driving data of vehicle A, the driving data of vehicle B, and the driving data of vehicle C. If field a is included in the driving data of vehicle A, vehicle B, and vehicle C, field b is included in the driving data of vehicle A and vehicle B, and only field c is included in the driving data of vehicle A, it can be considered that the quality of field a is higher than that of field b, and the quality of field b is higher than that of field c. Exemplarily, the field data quality can be viewed through missingno.matrix().

[0049] Determine the intermediate feature fields according to the field data quality and / or actual vehicle experience. For example, screen and obtain a preset number of initial feature fields with higher field quality as the intermediate feature fields. Or, according to actual vehicle experience, screen and obtain the first feature fields from the initial feature fields, and, according to the field data quality, screen and obtain the second feature fields; take the intersection of the first feature fields and the second feature fields as the intermediate feature fields.

[0050] The number of fields of the intermediate feature fields is multiple. Determine the field correlation between the intermediate feature fields, and screen out the intermediate feature fields with higher field correlation. For example, if the feature correlation between intermediate feature field A and intermediate feature field B is greater than the preset correlation threshold, then either intermediate feature field A or intermediate feature field B can be retained; the specific retention criterion can also be combined with the correlation between the corresponding intermediate feature field and other intermediate feature fields. Feature fields with low correlation can avoid multicollinearity.

[0051] Exemplarily, a heatmap of each intermediate feature field can be drawn through the heatmap() function, and the correlation coefficient between each intermediate feature field can be determined according to the heatmap. This correlation coefficient can be used to characterize the degree of field correlation. Determine the intermediate feature fields with a correlation coefficient greater than the preset coefficient threshold (such as 0.9 or 0.95) for selective elimination. For example, the intermediate feature fields include instrument vehicle speed, engine speed, actual engine torque, driver demand torque, and throttle pedal opening. The target feature fields obtained through field correlation screening are instrument vehicle speed and driver demand torque.

[0052] The above technical solution screens fields in vehicle driving data based on factors such as field data quality and field correlation, improves the screening quality of target feature fields, and reduces the complexity of the subsequent model training process; avoids multicollinearity, enhances the generalization ability, and avoids over-reliance on redundant information during model training, thereby improving model performance.

[0053] According to the target feature fields, driving sample data is constructed based on vehicle driving data. Specifically, the missing value filling of the fields in the vehicle driving data lacking the target feature fields can be performed based on the upward filling method. To further enhance the model performance and effect and enhance the feature expression ability, feature engineering construction can be performed on the target feature fields, thereby improving the integrity of feature data. For example, if the target feature fields are instrument vehicle speed and driver demand torque, the constructed feature fields after feature engineering construction are driving duration, maximum speed, idle time ratio, average speed, average demand torque, maximum demand torque, and acceleration time ratio.

[0054] According to the target feature fields and the constructed feature fields, driving sample data is generated. Specifically, the driving segment of any vehicle from the last stop to the next stop in the vehicle driving data is used as a kinematic segment, and the target feature fields, the constructed feature fields, and their corresponding field values of each kinematic segment are determined to form target sample data. 80% of the target sample data is used as the driving sample data for model training; 20% of the target sample data is used as the test sample data for model testing.

[0055] Determine the standard driving condition category corresponding to the driving sample data. Specifically, relevant technicians can manually label the driving sample data to determine its corresponding standard driving condition category; or a label annotation tool can also be used to perform automated or semi-automated category annotation on the driving sample data. To further improve the determination accuracy and efficiency of the standard driving condition category, in an alternative embodiment, determining the standard driving condition category corresponding to the driving sample data includes:

[0056] Step c1: Determine the condition category value according to the driving sample data.

[0057] Step c2: Based on the condition category value, perform cluster analysis on the driving sample data to obtain at least one set of cluster sets; the number of sets of the cluster sets is the same as the condition category value.

[0058] Step c3: For any cluster set, use the set category of the cluster set as the standard driving condition category of each driving sample data under the cluster set.

[0059] Specifically, according to the driving sample data, the elbow method can be applied to determine the optimal value of the working condition category (K value). For example, if the calculated K value is 3, it means that the number of categories obtained by clustering is 3. Based on the K-means clustering algorithm, the driving sample data is clustered and analyzed to obtain at least one set of clustering sets. Among them, the number of clustering sets is the same as the value of the working condition category. According to the value of the working condition category (when K is 3), three driving working conditions are defined, namely urban working condition, provincial road and highway working condition. According to the clustering results, labels for each driving sample data are generated, that is, the standard driving working condition category. Optionally, if it is found through correlation analysis and variance analysis that there is still a high degree of correlation among the features in the constructed driving sample data, feature screening can be further performed.

[0060] The above technical solution realizes the automatic determination of the working condition category of the driving sample data by means of clustering analysis and data labeling, improving the determination accuracy and efficiency. According to the driving sample data, the method for determining the value of the working condition category realizes the reasonable determination of the number of clusters, clarifies the specific types of driving working condition categories, and improves the reliability and efficiency of sample annotation.

[0061] Among them, the initial network model can be pre-selected by relevant technical personnel according to actual needs. Different initial network models correspond to different model hyperparameters.

[0062] Optionally, the initial network model is a support vector machine SVM (Support Vector Machine); the model hyperparameters of SVM include the penalty coefficient and the kernel density width; correspondingly, determining the current model hyperparameters of the initial network model in the current iteration cycle includes: determining the parameter optimization range of the model hyperparameters; based on the parameter optimization range, generating several initial individuals corresponding to the model hyperparameters; based on the genetic algorithm, determining the optimal individual among the initial individuals and taking the optimal individual as the current model hyperparameters in the current iteration cycle.

[0063] This optional embodiment performs model hyperparameter optimization based on the genetic algorithm. Before each iteration and optimization of the genetic algorithm, a set of initial solutions, that is, a population, needs to be randomly generated, and each solution is an individual. Usually, the initial solutions are randomly generated. To further accurately determine the optimization range of the genetic algorithm, this embodiment also provides a method for determining the parameter optimization range.

[0064] In an alternative embodiment, determining the parameter search range of the model hyperparameters includes: determining the search range and search step of the model hyperparameters through the grid search method; generating at least one set of reference hyperparameters according to the search range and search step; selecting the optimal parameters for rough search from each set of reference hyperparameters based on the cross-validation algorithm; and determining the parameter search range according to the optimal parameters for rough search.

[0065] Specifically, define the search ranges and search steps corresponding to the penalty coefficient C and the kernel density width g respectively. For example, the search range of the penalty coefficient C is [10 -3 , 10 3 , and the search step is incremented by a factor of 10, such as 10 -3 , 10 -2 , …, 10 3 . The search range of the kernel density width g is [10 -3 , 10 3 , and the search step is incremented by a factor of 10. According to the search range and search step, generate at least one set of reference hyperparameters (C1, g1), ……, (C n , g n ). Use the cross-validation method to evaluate the classification accuracy of each set of (C, g), and select the set of (C, g) with the highest accuracy as the optimal parameters for rough search. Generate the parameter search range centered on the optimal parameters for rough search (C, g). For example, set a smaller range centered on the optimal parameters for rough search (C, g), such as [C / 10, C*10] and [g / 10, g*10].

[0066] Based on this parameter search range, generate several initial individuals corresponding to the model hyperparameters of the genetic algorithm. And based on the genetic algorithm, perform fitness evaluation, selection, crossover, and mutation on each initial individual to obtain the optimal individual in the current population as the current model hyperparameters in the middle stage of the current iteration of the genetic algorithm: the penalty coefficient C and the kernel density width g.

[0067] Use the current model hyperparameters as the model hyperparameters of the SVM to participate in model training. Input the driving sample data and its corresponding standard driving condition categories into the SVM to obtain the predicted driving condition categories output by the SVM; according to the driving sample data, its corresponding standard driving condition categories, and the predicted driving condition categories, perform model training on the SVM until the model training iteration conditions are met, such as the loss value reaches the set threshold or the number of iterations reaches the preset number threshold, etc., to obtain the current driving condition model that has completed training in the current iteration cycle.

[0068] Use the test sample data to perform a model test on the current driving condition model, and determine whether the current driving condition model meets the model test conditions. For example, the model test conditions can be that the model accuracy reaches a set threshold, etc. This embodiment does not limit the specific model test conditions, as long as the requirements for model test and verification can be met.

[0069] If the model test result meets the preset model test conditions, then determine the current driving condition model as the vehicle driving condition model; if the model test result does not meet the preset model test conditions, then continue to iterate the model hyperparameters of the SVM until the trained driving condition model meets the preset model test conditions.

[0070] The technical solution of the embodiment of the present invention determines the current model hyperparameters of the initial network model in the current iteration cycle, and based on the current model hyperparameters of the initial network model, performs model training on the initial network model according to the driving sample data and its corresponding standard driving condition categories until the preset model test conditions are met, and a vehicle driving condition model is obtained. The above technical solution continuously iteratively optimizes the model hyperparameters during the model training process of the vehicle driving condition model, so as to train the vehicle driving condition model based on the continuously optimized model hyperparameters, making the model training more accurate, significantly improving the model performance, efficiency and robustness. The method of performing model training based on the continuously optimized model hyperparameters and vehicle driving condition data improves the training accuracy of the vehicle driving condition model, and further improves the accuracy of identifying road conditions during vehicle driving.

[0071] The above-trained vehicle driving condition model is used to predict the driving road conditions during vehicle driving. For the usage scenario of the vehicle driving condition model, for example, obtain the to-be-predicted driving condition data of the target vehicle, and input the to-be-predicted driving condition data into the vehicle driving condition model for driving road condition prediction to obtain a condition prediction result, which can be, for example, an urban road, a provincial road or a highway, etc.

[0072] Embodiment 2

[0073] Figure 2 It is a flowchart of a method for training a vehicle driving condition model provided by Embodiment 2 of the present invention. Based on the above embodiment, this embodiment provides a preferred example.

[0074] As Figure 2 shown, the method includes the following specific steps:

[0075] S21. Obtain vehicle driving data, perform feature field processing on the vehicle driving data, and determine target feature fields.

[0076] Specifically, perform feature field processing on vehicle driving data to determine initial feature fields; determine the field data quality of the initial feature fields, and based on the field data quality, determine intermediate feature fields; determine target feature fields according to the field correlation between the intermediate feature fields.

[0077] Exemplarily, the vehicle driving data comes from the data collected by the vehicle networking Tbox, and the screening range is the driving data of 30 heavy commercial vehicles for 3 months. First, use the read_excel function provided by the pandas library to read the data, and collect the fields included in the data through the columns() function, that is, the initial feature fields. View the field data quality of the initial feature fields through missingno.matrix(). Determine the intermediate feature fields based on the field data quality and actual vehicle experience. For example, the intermediate feature fields are instrument vehicle speed, engine speed, actual engine torque, driver demand torque, and accelerator pedal opening. Draw a heat map through the sns.heatmap() function to obtain the correlation coefficients between the intermediate feature fields. By screening out the features with higher correlation coefficients, the target feature fields are obtained. For example, the target feature fields are instrument vehicle speed and demand torque.

[0078] S22. Based on the target feature fields and vehicle driving data, construct driving sample data.

[0079] Specifically, first fill in the missing values of the target feature fields by the forward filling method, and based on the target feature fields and vehicle driving data, construct driving sample data. Respectively construct driving duration, maximum speed, idle time ratio, average speed, average demand torque, maximum demand torque, and acceleration time ratio. And take the driving segment from the last stop to the next stop in the vehicle driving data as a kinematic segment, and calculate the feature values of each kinematic segment to form target sample data. And take 80% of the target sample data as the training set (driving sample data) and 20% as the test set (test sample data).

[0080] S23. Perform clustering analysis on the driving sample data to determine the standard driving condition categories of the driving sample data.

[0081] Specifically, according to the driving sample data, determine the condition category values; based on the condition category values, perform clustering analysis on the driving sample data to obtain at least one set of clustering sets; where the number of clustering sets is the same as the condition category values; for any clustering set, use the set category of the clustering set as the standard driving condition category of each driving sample data under the clustering set.

[0082] Exemplarily, perform clustering analysis on the driving sample data, use the K-means clustering algorithm, and apply the elbow method to determine that the optimal k value is 3. According to the eigenvalue of the cluster center, define three driving conditions, namely urban, provincial road, and highway. And according to the clustering result, label each driving sample data. Through correlation analysis and variance analysis, it is found that the constructed features still have a high degree of correlation. Therefore, the average speed, idle time ratio, average demand torque, and acceleration time ratio are finally selected as the final sample features.

[0083] S24. Determine the current model hyperparameters of the initial network model SVM in the current iteration cycle, including the penalty coefficient C and the kernel density width g.

[0084] Specifically, first use the grid search method for rough search to determine the search step size and search range, and determine the optimal C and g values during the rough search through the cross-validation method, which is used as the basis for determining the optimization range of the next genetic algorithm. This solution can reduce the iteration speed of the genetic algorithm and save the convergence time. Determine the optimization range of the genetic algorithm, which is determined by the result of the rough search, and perform binary encoding. Randomly generate M initial individuals. Calculate the classification accuracy through the SVM cross-validation method, which is used as the fitness evaluation criterion for each individual. Continuously evolve the individuals through selection, crossover, and mutation operations to obtain the current individual (the current model hyperparameters (C, g)) in the current iteration cycle.

[0085] S25. Based on the driving sample data and its corresponding standard driving condition categories, and based on the current model hyperparameters of the initial network model, train the initial network model to obtain the current driving condition model that has been trained in the current iteration cycle.

[0086] S26. Test the current driving condition model to determine whether the current driving condition model meets the preset model test conditions; if so, it means that the current driving condition model is better, and execute S27; if not, loop to execute S24 - S26.

[0087] S27. Determine the current driving condition model as the vehicle driving condition model.

[0088] Embodiment III

[0089] Figure 3 It is a schematic structural diagram of a vehicle driving condition model training device provided in Embodiment III of the present invention. A vehicle driving condition model training device provided in an embodiment of the present invention is applicable to the situation of identifying road conditions during the driving process of commercial vehicles. The vehicle driving condition model training device can be implemented in the form of hardware and / or software, such as Figure 3As shown in the figure, the device specifically includes: a working condition category acquisition module 301, a hyperparameter determination module 302, a model training module 303, and a model testing module 304. Among them,

[0090] The working condition category acquisition module 301 is used to acquire driving sample data and its corresponding standard driving working condition category;

[0091] The hyperparameter determination module 302 is used to determine the current model hyperparameters of the initial network model in the current iteration cycle;

[0092] The model training module 303 is used to perform model training on the initial network model based on the driving sample data and its corresponding standard driving working condition category and the current model hyperparameters of the initial network model, so as to obtain the current driving working condition model that has completed training in the current iteration cycle;

[0093] The model testing module 304 is used to perform model testing on the current driving working condition model. If the model testing result meets the preset model testing conditions, the current driving working condition model is determined as the vehicle driving working condition model; the vehicle driving working condition model is used to predict the driving road conditions during the vehicle driving process.

[0094] The technical solution of the embodiment of the present invention determines the current model hyperparameters of the initial network model in the current iteration cycle, and based on the driving sample data and its corresponding standard driving working condition category, performs model training on the initial network model based on the current model hyperparameters of the initial network model until the preset model testing conditions are met, and obtains the vehicle driving working condition model. The above technical solution continuously iteratively optimizes the model hyperparameters during the model training process of the vehicle driving working condition model, thereby training the vehicle driving working model based on the continuously optimized model hyperparameters, making the model training more accurate, significantly improving the model performance, efficiency and robustness. The method of performing model training based on the continuously optimized model hyperparameters and vehicle driving working condition data improves the training accuracy of the vehicle driving working condition model, and further improves the accuracy of identifying the road conditions during the vehicle driving process.

[0095] Optionally, the initial network model is a support vector machine SVM; the model hyperparameters of the SVM include a penalty coefficient and a kernel density width;

[0096] Correspondingly, the hyperparameter determination module 302 includes:

[0097] An optimization range determination unit, which is used to determine the parameter optimization range of the model hyperparameters;

[0098] An initial individual generation unit, which is used to generate a number of initial individuals corresponding to the model hyperparameters based on the parameter optimization range;

[0099] The optimal individual determination unit is configured to determine the optimal individual among the initial individuals based on a genetic algorithm, and use the optimal individual as the current model hyperparameters in the current iteration cycle.

[0100] Optionally, the optimization range determination unit is specifically configured to:

[0101] Determine the search range and search step size of the model hyperparameters through a grid search method;

[0102] Generate at least one set of reference hyperparameters according to the search range and the search step size;

[0103] Select the optimal parameters for rough search from the reference hyperparameters based on a cross-validation algorithm;

[0104] Determine the parameter optimization range according to the optimal parameters for rough search.

[0105] Optionally, the device further includes:

[0106] The feature field processing module is configured to obtain vehicle driving data and perform feature field processing on the vehicle driving data to determine target feature fields before obtaining the driving sample data and its corresponding standard driving condition categories.

[0107] The sample construction module is configured to construct driving sample data based on the vehicle driving data according to the target feature fields.

[0108] The category determination module is configured to determine the standard driving condition category corresponding to the driving sample data.

[0109] Optionally, the feature field processing module is specifically configured to:

[0110] Perform feature field processing on the vehicle driving data to determine initial feature fields;

[0111] Determine the field data quality of the initial feature fields, and determine intermediate feature fields according to the field data quality;

[0112] Determine target feature fields according to the field correlation between the intermediate feature fields.

[0113] Optionally, the category determination module is specifically configured to:

[0114] Determine the condition category value according to the driving sample data;

[0115] Perform cluster analysis on the driving sample data based on the condition category value to obtain at least one set of cluster sets; the number of sets of the cluster sets is the same as the condition category value;

[0116] For any clustering set, the set category of the clustering set is used as the standard driving condition category of each driving sample data under the clustering set.

[0117] The vehicle driving condition model training device provided by the embodiments of the present invention can execute the vehicle driving condition model training method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0118] Embodiment 4

[0119] Figure 4 FIG. shows a schematic structural diagram of an electronic device 40 that can be used to implement the embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as, for example, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, for example, a personal digital processor, a cellular phone, a smart phone, a wearable device (such as a helmet, glasses, a watch, etc.) 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 present invention described and / or claimed herein.

[0120] As Figure 4 shown, the electronic device 40 includes at least one processor 41, and a memory communicatively connected to the at least one processor 41, such as a read-only memory (ROM) 42, a random access memory (RAM) 43, etc. Among them, the memory stores a computer program executable by the at least one processor. The processor 41 can execute various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 42 or the computer program loaded from the storage unit 48 into the random access memory (RAM) 43. In the RAM 43, various programs and data required for the operation of the electronic device 40 can also be stored. The processor 41, the ROM 42, and the RAM 43 are connected to each other through a bus 44. The input / output (I / O) interface 45 is also connected to the bus 44.

[0121] A plurality of components in the electronic device 40 are connected to the I / O interface 45, including: an input unit 46, such as a keyboard, a mouse, etc.; an output unit 47, such as various types of displays, speakers, etc.; a storage unit 48, such as a magnetic disk, an optical disk, etc.; and a communication unit 49, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 49 allows the electronic device 40 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0122] The processor 41 may be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 41 executes the various methods and processes described above, such as the vehicle driving condition model training method.

[0123] In some embodiments, the vehicle driving condition model training method may be implemented as a computer program, which is tangibly included in a computer-readable storage medium, such as the storage unit 48. In some embodiments, part or all of the computer program may be loaded and / or installed onto the electronic device 40 via the ROM 42 and / or the communication unit 49. When the computer program is loaded into the RAM 43 and executed by the processor 41, one or more steps of the vehicle driving condition model training method described above may be executed. Alternatively, in other embodiments, the processor 41 may be configured to execute the vehicle driving condition model training method by any other suitable means (e.g., by means of firmware).

[0124] 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 (FPGA), application-specific integrated circuits (ASIC), application-specific standard products (ASSP), system-on-chip systems (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: implemented in one or more computer programs, the one or more computer programs can be executed and / or interpreted on a programmable system including at least one programmable processor, the programmable processor can be a dedicated or general-purpose programmable processor, can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0125] The computer program for implementing the method of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowchart and / or block diagram are implemented. The computer program can be executed entirely on the machine, partially on the machine, as an independent software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0126] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0127] To provide for interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can also be used to provide for 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 acoustic input, voice input, or tactile input).

[0128] The systems and techniques described herein can be implemented in a computing system that includes backend components (such as, for example, a data server), or a computing system that includes middleware components (such as, for example, an application server), or a computing system that includes frontend components (such as, for example, a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (such as, for example, a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0129] A computing system may include a client and a server. The client and the server are generally far from each other and usually interact via a communication network. The relationship between the client and the server is created by computer programs running on respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, solving the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0130] It should be understood that various forms of processes shown above can be used, steps can be reordered, added or deleted. For example, the steps described in the present invention can be executed in parallel, sequentially or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is made herein.

[0131] The above specific embodiments do not constitute a limitation to the protection scope of the present invention. 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 principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A vehicle driving condition model training method, characterized in that: include: Obtain driving sample data and its corresponding standard driving condition category; Determine the current model hyperparameters of the initial network model in the current iteration cycle; According to the driving sample data and the corresponding standard driving condition categories, based on the current model hyperparameters of the initial network model, the initial network model is trained to obtain a current driving condition model that has completed training in the current iteration cycle; A model test is performed on the current driving condition model. If the model test result meets the preset model test conditions, the current driving condition model is determined as the vehicle driving condition model; the vehicle driving condition model is used to predict the driving road condition of the vehicle during driving.

2. The method according to claim 1, characterized in that The initial network model is a support vector machine (SVM); the model hyperparameters of the SVM include a penalty coefficient and a kernel density width; Accordingly, determining the current model hyperparameters of the initial network model in the current iteration cycle includes: Determining a parameter optimization range for the model hyperparameters; Based on the parameter optimization range, generating a number of initial individuals corresponding to the model hyperparameters; Based on the genetic algorithm, the best individual among the initial individuals is determined, and the best individual is used as the current model hyperparameter in the current iteration cycle.

3. The method according to claim 2, characterized in that The step of determining the parameter optimization range of the model hyperparameters includes: Determine the search range and search step size of the model hyperparameters by grid search method; Generate at least one set of reference hyperparameters according to the search range and the search step size; Based on a cross-validation algorithm, a rough search optimal parameter is selected from each of the reference hyperparameters; According to the rough search for the optimal parameter, a parameter optimization range is determined.

4. The method according to claim 1, characterized in that Before obtaining the driving sample data and the corresponding standard driving condition category, the method further includes: Acquire vehicle driving data, and perform feature field processing on the vehicle driving data to determine a target feature field; According to the target feature field, based on the vehicle driving data, construct driving sample data; Determine the standard driving condition category corresponding to the driving sample data.

5. The method according to claim 4, characterized in that The performing feature field processing on the vehicle driving data to determine the target feature field includes: Performing feature field processing on the vehicle driving data to determine an initial feature field; Determining the field data quality of the initial feature field, and determining the intermediate feature field based on the field data quality; The target feature field is determined according to the field correlation between the intermediate feature fields.

6. The method according to claim 4, characterized in that The determining of the standard driving condition category corresponding to the driving sample data includes: Determining a working condition category value according to the driving sample data; Based on the working condition category value, cluster analysis is performed on the driving sample data to obtain at least one set of cluster sets; the number of cluster sets is the same as the working condition category value; For any cluster set, the set category of the cluster set is used as the standard driving condition category of each driving sample data under the cluster set.

7. A vehicle driving condition model training device, characterized in that: include: A working condition category acquisition module is used to obtain driving sample data and its corresponding standard driving condition category; A hyperparameter determination module, used to determine the current model hyperparameters of the initial network model in the current iteration cycle; A model training module, used for performing model training on the initial network model according to the driving sample data and the corresponding standard driving condition category and based on the current model hyperparameters of the initial network model, to obtain a current driving condition model that has completed training in the current iteration cycle; The model testing module is used to perform a model test on the current driving condition model. If the model testing result meets the preset model testing conditions, the current driving condition model is determined as the vehicle driving condition model; the vehicle driving condition model is used to predict the driving road conditions during the vehicle driving process.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the vehicle driving condition model training method described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the vehicle driving condition model training method according to any one of claims 1 to 6 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the vehicle driving condition model training method according to any one of claims 1 to 6 is implemented.

Images