Vehicle rolling critical condition prediction method and device

Abstract

The invention discloses a vehicle rolling critical condition prediction method and device. The method comprises the following steps: acquiring n groups of rolling data about vehicle rolling; based on the n groups of vehicle rollover parameters, performing new data synthesis processing to obtain m groups of enhanced rollover parameters corresponding to the vehicle rollover parameters; inputting the enhanced tumbling parameter into the implicit neural network, and outputting a tumbling result corresponding to the enhanced tumbling parameter; training the initial explicit neural network through the enhanced tumbling parameter and a tumbling result corresponding to the enhanced tumbling parameter to obtain a trained explicit neural network; and determining a critical condition for triggering the rolling of the target vehicle through the explicit neural network according to the input actual rolling parameter of the target vehicle. In this way, the implicit neural network is used to learn the rolling discrimination capability from the small-scale real vehicle rolling data; and then a large number of reasonable enhanced rolling parameters are generated to train an explicit neural network so as to output critical conditions of vehicle rolling expressed in a specific parameter threshold form.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle control, in particular to a vehicle rollover critical condition prediction method and device. BACKGROUND

[0002] In road traffic accidents, although the frequency of vehicle rollover accidents is relatively low, the injury rate and fatality rate are significantly higher than those of other types of collision accidents. In order to improve the passenger protection level, automobile manufacturers need to accurately calibrate the airbag point explosion time through real vehicle rollover test. However, the existing rollover test highly depends on the experience of engineers to set the test conditions, which has multiple limitations: on the one hand, due to the complex motion state of the vehicle and the influence of multiple parameter coupling, the test conditions set by manual often cannot trigger the rollover accurately, resulting in test failure; on the other hand, even if the rollover is successful, the obtained conditions are usually far from the critical value, which is difficult to support the development of high-precision control strategy. In addition, the real vehicle test is costly, and if the critical boundary is approached through repeated trial and error, the economic cost is huge; at the same time, the rollover conditions obtained by a specific vehicle model lack generalization ability and are difficult to migrate to other vehicle platforms. Traditional modeling methods also face difficulties: implicit neural networks can realize the binary classification of rollover / non-rollover under small samples, but their black box characteristics cannot provide explicit critical condition thresholds; while explicit models such as decision trees have good interpretability, but it is difficult to train high-precision and high-robustness rules due to the scarcity of real vehicle data. SUMMARY

[0003] In view of the above problems, the present application provides a vehicle rollover critical condition prediction method and device, which accurately obtains the critical condition of vehicle rollover for guiding airbag point explosion calibration.

[0004] According to a first aspect of the present application, a vehicle rollover critical condition prediction method is provided, comprising obtaining n groups of rollover data about vehicle rollover, each group of rollover data including vehicle rollover parameters and rollover results; based on the n groups of vehicle rollover parameters, performing new data synthesis processing to obtain m groups of enhanced rollover parameters corresponding to the n groups of vehicle rollover parameters; wherein m is greater than n, and m and n are both integers greater than 0; inputting the enhanced rollover parameters into an implicit neural network to output the rollover results corresponding to the enhanced rollover parameters; wherein the implicit neural network is trained by the n groups of rollover data; training an initial explicit neural network through the enhanced rollover parameters and the rollover results corresponding to the enhanced rollover parameters to obtain a trained explicit neural network; inputting the actual rollover parameters of the target vehicle into the explicit neural network to determine the critical condition triggering the rollover of the target vehicle through the explicit neural network.

[0005] Optionally, the new data synthesis processing based on the n groups of vehicle rollover parameters is performed to obtain m groups of enhanced rollover parameters corresponding to the n groups of vehicle rollover parameters, including: determining the value range and distribution characteristics of each parameter in the vehicle rollover parameters respectively; generating m groups of enhanced rollover parameters corresponding to the n groups of vehicle rollover parameters according to the value range and distribution characteristics of each parameter in the vehicle rollover parameters; Optionally, the generating m groups of enhanced rollover parameters corresponding to the n groups of vehicle rollover parameters according to the value range and distribution characteristics of each parameter in the vehicle rollover parameters includes: for each parameter in the vehicle rollover parameters, generating an enhanced parameter set corresponding to the parameter through interpolation processing within the value range of the parameter, and the enhanced parameter set includes multiple numerical values of the parameter; constructing a physical constraint relationship according to the distribution characteristics of each parameter in the vehicle rollover parameters; generating m enhanced rollover parameters from the enhanced parameter set corresponding to each parameter based on the physical constraint relationship.

[0006] Optionally, the constructing a physical constraint relationship according to the distribution characteristics of each parameter in the vehicle rollover parameters includes: determining the physical correlation between each vehicle rollover parameter according to the distribution characteristics of each parameter in the vehicle rollover parameters; based on the physical correlation, setting a combination rule for the enhanced parameter set of each parameter generated in the interpolation processing to construct a physical constraint relationship.

[0007] Optionally, the training step of the implicit neural network includes: dividing the n groups of vehicle rollover parameters and corresponding rollover results into a training set, a validation set and a test set; training an initial implicit neural network using the training set, and adjusting network hyperparameters based on the loss function value of the validation set; when the prediction accuracy of the initial implicit neural network on the test set reaches a preset threshold, obtaining the trained implicit neural network.

[0008] Optionally, the training step of the explicit neural network includes: using the m groups of enhanced rollover parameters and the respective corresponding rollover results as a training data set; based on the training data set, using information entropy as a splitting criterion to train an initial explicit neural network until a preset stopping condition is met, and obtaining a trained explicit neural network.

[0009] Optionally, the step of inputting the actual rollover parameters of the target vehicle into the explicit neural network, and determining the critical conditions that trigger the rollover of the target vehicle through the explicit neural network, includes: The actual rollover parameters of the target vehicle are input into the explicit neural network to obtain the corresponding target decision path; Analyze each decision node in the target decision path and extract the parameter thresholds corresponding to the rollover parameters of each vehicle that are related to the rollover result corresponding to the actual rollover parameters. The parameter thresholds corresponding to the rollover parameters of each vehicle are used as the critical conditions for triggering vehicle rollover.

[0010] According to a second aspect of the present invention, a vehicle rollover critical condition prediction device is provided, comprising: The acquisition module acquires n sets of rollover data about the vehicle rollover, and each set of rollover data includes vehicle rollover parameters and rollover result; The synthesis module is used to perform new data synthesis processing based on the n sets of vehicle rollover parameters to obtain m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters; where m is greater than n, and m and n are both integers greater than 0; The first training module is used to input the enhanced roll parameters into the implicit neural network and output the roll result corresponding to the enhanced roll parameters; wherein, the implicit neural network is trained from n sets of roll data; The second training module is used to train the initial explicit neural network using the enhanced roll parameters and the roll results corresponding to the enhanced roll parameters, so as to obtain the trained explicit neural network. The output module is used to input the actual rollover parameters of the target vehicle into the explicit neural network, and through the explicit neural network, determine the critical conditions that trigger the rollover of the target vehicle.

[0011] According to a third aspect of the present invention, an electronic device is provided, the electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned vehicle rollover critical condition prediction method.

[0012] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method for predicting critical conditions for vehicle rollover.

[0013] The above-described one or more technical solutions in the embodiments of this specification have at least the following technical effects: This specification provides a method and apparatus for predicting critical conditions for vehicle rollover. The method involves acquiring n sets of rollover data, including vehicle rollover parameters and rollover results; synthesizing new data based on the n sets of vehicle rollover parameters to obtain m sets of enhanced rollover parameters corresponding to the vehicle rollover parameters; inputting the enhanced rollover parameters into an implicit neural network to output the rollover results corresponding to the enhanced rollover parameters; wherein the implicit neural network is trained from the n sets of rollover data; training an initial explicit neural network using the enhanced rollover parameters and the corresponding rollover results to obtain a trained explicit neural network; and inputting the actual rollover parameters of the target vehicle into the explicit neural network to determine the critical conditions that trigger the target vehicle rollover. Thus, an implicit neural network is first used to learn rollover discrimination capabilities from small-scale real-vehicle rollover data. Then, combining parameter distribution characteristics, a large number of reasonable virtual samples are generated based on physical constraints, and these are labeled using the trained implicit neural network to construct a large number of enhanced rollover parameters. These parameters are then used to train an explicit neural network to output the critical conditions for vehicle rollover expressed in the form of specific parameter thresholds. This entire process overcomes the bottleneck of insufficient real-vehicle data while ensuring the physical rationality and engineering feasibility of the generated rules.

[0014] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference figures denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a vehicle rollover critical condition prediction method according to an embodiment of the present invention is shown.

[0016] Figure 2 A block diagram of a vehicle rollover critical condition prediction device according to an embodiment of the present invention is shown. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0018] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0019] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0020] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0021] Combination Figure 1 As shown, the present invention provides a method for predicting vehicle rollover critical conditions, which includes steps 101 to 105: Step 101: Obtain n sets of rollover data about the vehicle rollover, each set of rollover data including vehicle rollover parameters and rollover result; Step 102: Based on the n sets of vehicle rollover parameters, perform new data synthesis processing to obtain m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters; where m is greater than n, and m and n are both integers greater than 0; Step 103: Input the enhanced roll parameters into the implicit neural network and output the roll result corresponding to the enhanced roll parameters; wherein, the implicit neural network is trained by the n sets of roll data; Step 104: Train the initial explicit neural network using the enhanced roll parameters and the roll results corresponding to the enhanced roll parameters to obtain the trained explicit neural network; Step 105: Input the actual rollover parameters of the target vehicle into the explicit neural network, and determine the critical conditions that trigger the rollover of the target vehicle through the explicit neural network.

[0022] In this embodiment, the implicit neural network is a typical deep neural network (e.g., a DNN), which takes various vehicle rollover parameters (e.g., speed, angle, mass, etc.) as input and outputs a binary classification result: rollover or no rollover.

[0023] This is because the decision-making process within an implicit neural network is opaque. Given a set of parameters, it can tell you very accurately whether the outcome is rollover or not, but it cannot explain why. For example, an implicit neural network will not output a rule like "it rolled over because the slope angle was greater than 48.5 degrees." The knowledge of an implicit neural network is "implicitly" encoded in the network's weights and biases, making it difficult to interpret directly. However, in small sample sizes, its nonlinear fitting capabilities can achieve high prediction accuracy.

[0024] Considering the high cost of real vehicle rollovers and the limited availability of actual sample data, this embodiment utilizes the aforementioned advantages of implicit neural networks to train the implicit neural network using a limited set of n rollover data. The trained implicit neural network then helps to expand the rollover data.

[0025] In this embodiment, n sets of rollover data are first acquired. Each set of data includes multiple vehicle rollover parameters (e.g., vehicle speed, cutting angle, slope angle, wheel mass, axle load distribution, vehicle dimensions, etc.) and the corresponding rollover result (i.e., whether a rollover occurred, represented by 1 or 0). Since actual rollover tests are costly, the amount of real test data n accumulated is usually small, making it difficult to guarantee prediction accuracy and generalization ability when directly used to train an explicit model. Therefore, this invention proposes to first perform new data synthesis processing on the original n sets of data to generate m sets of enhanced rollover parameters, where m is much larger than n. This enhancement process is not simply random generation, but rather based on the value range and distribution characteristics of each parameter in the original data, performing reasonable interpolation and combination under physical constraints to construct virtual samples that both conform to the distribution of real data and are sufficiently numerous. Subsequently, these m sets of enhanced rollover parameters are input into an implicit neural network trained from the original n sets of rollover data, and the network outputs the corresponding rollover result label.

[0026] The implicit neural network acts as a virtual testing platform, leveraging its nonlinear mapping learned under small sample conditions to assign reliable labels to a large number of augmented samples. These labeled augmented samples are then used to train an initial explicit neural network (e.g., a decision tree), ultimately resulting in an explicit model that clearly outputs the threshold values ​​for each parameter. When the actual rollover parameters of the target vehicle are input into this explicit network, its internal decision path can be analyzed, extracting the key parameters triggering the rollover and their critical thresholds, thereby accurately guiding the calibration of airbag deployment timing or the setting of test conditions. This method effectively integrates the data generation capabilities of the implicit model and the interpretability of the explicit model, significantly improving the accuracy and practicality of predicting rollover critical conditions.

[0027] Optionally, the step of performing new data synthesis processing based on the n sets of vehicle rollover parameters to obtain m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters includes: The value range and distribution characteristics of each parameter in the vehicle rollover parameters are determined respectively; Based on the value range and distribution characteristics of each parameter in the vehicle rollover parameters, m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters are generated.

[0028] In this embodiment, during the new data synthesis process, statistical analysis is first performed on each vehicle rollover parameter in the original n sets of rollover data to determine its value range (e.g., minimum and maximum values) and distribution characteristics (e.g., whether it follows a normal distribution, whether there is skewness, etc.). For example, for a continuous variable like vehicle speed, it may exhibit an approximately uniform distribution; while for discrete variables like vehicle type, it is necessary to statistically analyze the frequency of occurrence of each category. Based on this statistical information, new parameter combinations can be scientifically generated while preserving the distribution characteristics of the original data.

[0029] Specifically, for continuous parameters, multiple intermediate values ​​can be generated within their range using linear or nonlinear interpolation methods; for discrete parameters, sampling expansion is performed according to their category proportions. This data augmentation method based on distribution characteristics ensures that the generated m sets of augmented tumbling parameters are not only sufficient in number but also highly consistent with the statistical characteristics of real experimental data, avoiding data distortion problems caused by blind expansion. The resulting augmented dataset is closer to the actual physical scenario, laying a high-quality data foundation for subsequent implicit network label prediction and explicit network training, thereby significantly improving the generalization ability and engineering applicability of the final model.

[0030] Optionally, the step of generating m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters based on the value range and distribution characteristics of each parameter in the vehicle rollover parameters includes: For each parameter in the vehicle rollover parameters, within the range of values ​​for that parameter, an enhancement parameter set corresponding to that parameter is generated through interpolation processing. The enhancement parameter set includes multiple values ​​for that parameter. Based on the distribution characteristics of each parameter in the vehicle rollover parameters, physical constraint relationships are constructed; Based on the physical constraint relationship, m enhanced roll parameters are generated from the enhanced parameter set corresponding to each parameter.

[0031] In this embodiment, to further enhance the physical plausibility of the augmented data, after generating the augmented parameter sets for each parameter through interpolation, physical constraints are introduced for filtering and combination. Specifically, for the complex dynamic process of vehicle rollover, there are inherent physical relationships between different parameters. For example, the mass of the left front wheel is usually similar to that of the right front wheel, the mass of the front axle should be approximately equal to the sum of the masses of the left and right front wheels, and the vehicle height is closely related to the center of gravity. If these constraints are ignored during data synthesis, the generated parameter combinations may not exist in reality (for example, the mass of the left front wheel is much greater than that of the right front wheel, and the sum exceeds the total vehicle mass), thus misleading the model learning. Therefore, this embodiment constructs physical correlation rules between parameters based on the understanding of vehicle structure and kinematics, and sets combination rules accordingly. For example, when extracting values ​​from the augmented sets of each parameter to form a complete sample, the system verifies whether it meets the preset physical constraints (such as mass conservation, geometric consistency, etc.), and only retains valid combinations. In this way, the generated m sets of enhanced roll parameters are not only numerous and reasonably distributed, but also have real physical meaning, which greatly improves the reliability of subsequent implicit neural network prediction labels and provides a guarantee for the explicit model to learn decision boundaries that conform to physical laws. Ultimately, this makes the predicted roll critical conditions more valuable for engineering guidance.

[0032] Optionally, the step of constructing physical constraint relationships based on the distribution characteristics of each parameter in the vehicle rollover parameters includes: Based on the distribution characteristics of each parameter in the vehicle rollover parameters, the physical correlation between each vehicle rollover parameter is determined; Based on the physical correlation, combination rules are set for the enhanced parameter sets of each parameter generated in the interpolation process to construct physical constraint relationships.

[0033] In this embodiment, the construction of physical constraint relationships relies on in-depth analysis of the distribution characteristics of each parameter in the original rollover data. This invention quantifies the strength of physical correlations by analyzing the joint distribution, correlation coefficients, or covariance matrices of different parameters in n sets of original data. For example, if the data shows a high correlation between the front axle mass and the sum of the masses of the left and right front wheels, an equality constraint can be established; if vehicle height is negatively correlated with rollover tendency, then parameter configurations corresponding to higher vehicle height and greater rollover tendency can be prioritized during combination. Based on this correlation analysis, the system sets combination rules for each set of augmented parameters generated by interpolation, such as requiring certain parameters to change synchronously or certain parameter combinations to fall within a specific interval. These rules essentially constitute a soft or hard physical constraint space, within which all generated augmented samples must reside. This data-driven method of constructing physical constraints avoids the subjectivity of relying entirely on expert experience and overcomes the non-physical interpretations that may arise from pure data-driven approaches. It ensures that the augmented data maintains diversity while strictly adhering to the basic principles of vehicle dynamics, thereby significantly improving the physical reliability and robustness of the entire prediction system.

[0034] Optionally, the training steps of the implicit neural network include: The n sets of vehicle rollover parameters and corresponding rollover results are divided into a training set, a validation set, and a test set. The initial implicit neural network is trained using the training set, and the network hyperparameters are adjusted based on the loss function values ​​of the validation set. When the prediction accuracy of the initial implicit neural network on the test set reaches a preset threshold, the trained implicit neural network is obtained.

[0035] In this embodiment, training the implicit neural network is the fundamental step for achieving high-precision rollover discrimination and subsequent data augmentation in this invention. The process first requires systematic preprocessing of the n sets of real-vehicle rollover test data collected. This raw data includes vehicle type, wheel mass (front left, front right, rear left, rear right), front and rear axle mass, vehicle dimensions (length, width, height), and test parameters (such as test speed, vehicle entry angle, and slope angle), labeled as "rollover" or "not rollover". To adapt to the neural network input requirements, textual variables (such as sedan, SUV, MPV, etc. in vehicle type) need to be assigned integer codes. Simultaneously, since the units of each parameter are different and the numerical magnitudes vary significantly (e.g., mass is in kilograms, angle in degrees, speed in km / h), de-unitization and normalization processing must be performed to ensure that all input features are within a similar numerical range, thereby improving the stability and convergence speed of network training. The label values ​​are also converted to binary values: "1" for a rollover result and "0" for no rollover, to ensure compatibility with non-linear activation functions such as the Sigmoid function. After preprocessing, all samples are randomly divided into training, validation, and test sets, regardless of vehicle model, to fully exploit common rollover features across vehicle models and enhance the model's generalization ability. Subsequently, the initial implicit neural network is iteratively trained using the training set. This network structure includes an input layer, one or more intermediate hidden layers, and an output layer. Each neuron in each layer transmits information through weighted summation and an activation function (the output layer uses the Sigmoid function). During training, the validation set is simultaneously input into the network, and the trend of the loss function (such as cross-entropy loss) is monitored in real time. If the validation loss stops decreasing or even increases, it is considered overfitting, requiring timely adjustment of hyperparameters or early termination of training. When the loss function drops below a preset threshold and the validation curve stabilizes, the current network parameters are saved. Finally, the model is evaluated using a test set that was not involved in training or validation; only the prediction error is calculated without updating the network weights. Only when the test error is below the set tolerance is the implicit neural network confirmed to have reliable discriminative ability and retained. Although the trained implicit network cannot directly output critical condition values, it can still accurately determine whether a rollover occurs under any combination of parameters under small sample conditions. This provides a large number of labeled virtual samples for subsequent explicit models, effectively solving the core bottlenecks of high cost and scarce data in real vehicle testing. It is a key support for realizing the explicit-implicit fusion prediction architecture.

[0036] Optionally, the training steps of the explicit neural network include: The m sets of enhanced tumbling parameters and their corresponding tumbling results are used as the training dataset; Based on the training dataset, information entropy is used as the splitting criterion to train the initial explicit neural network until a preset stopping condition is met, thus obtaining the trained explicit neural network.

[0037] In this embodiment, the training of the explicit neural network is the core step in achieving interpretable prediction of rollover critical conditions. This process uses m sets of enhanced rollover parameters generated by the implicit neural network and their corresponding rollover results (output by the implicit network) as the training dataset, constructing a dedicated explicit prediction model for each vehicle type. The reason for training by vehicle type is that different vehicle types (such as sedans, SUVs, MPVs, etc.) have significant differences in structural characteristics such as mass distribution, center of gravity height, wheelbase, and track width, resulting in different rollover mechanisms and sensitive parameters. Modeling by vehicle type improves prediction accuracy and engineering applicability. In terms of model selection, this invention uses a decision tree as the basic architecture of the explicit neural network. This is because it inherently possesses advantages such as readable rules, clear thresholds, and no need for complex post-processing, directly outputting the boundary conditions of various experimental parameters that trigger rollover, meeting the explicit requirement of "critical values" for airbag calibration. Before training begins, all input parameters need to be type-identified: variables with limited values, such as vehicle type, are classified as discrete parameters, while continuously changing variables such as vehicle speed, slope angle, wheel mass, and vehicle height are classified as continuous parameters. During training, the system uses information entropy as the core criterion for node splitting, selecting the optimal splitting parameters and points by maximizing information gain. Specifically, for each sample set contained in a node to be split, the overall entropy value is calculated sequentially for each parameter used as the splitting criterion. The overall entropy value is calculated as follows: if a parameter has V possible values ​​(discrete) or is divided into several intervals (continuous), then the entropy values ​​of each subset are weighted and summed according to the proportion of the number of samples in each subset to the total number of samples. This weighted sum is the overall entropy after splitting using this parameter. Information gain is the difference between the entropy before splitting and the overall entropy after splitting. The larger the gain, the more significantly the splitting improves the purity of the subset categories. For continuous parameters, all possible splitting points t need to be traversed within their value range to find the t value that minimizes the overall entropy (i.e., maximizes the information gain) as the optimal splitting threshold. For example, when determining whether "experimental speed" affects rollover, the system automatically searches for key speed points such as 15.2 km / h and 19.4 km / h, so that samples below or above these values ​​belong to the "rollover" or "non-rollover" category as much as possible. The splitting process is recursive: after selecting the current optimal parameters and completing the split, the entropy calculation and parameter selection process is repeated for each generated child node until a preset stopping condition is met, such as the number of samples in a node being lower than a certain threshold, the information gain being less than a set tolerance, or the tree depth reaching its upper limit. It is worth noting that after a split, the subset of continuous parameters remains a continuous variable; therefore, new split points need to be recalculated in subsequent levels to capture more refined nonlinear relationships. The final trained explicit neural network is presented as one or more decision trees, where each decision node corresponds to a specific vehicle rollover parameter and its critical threshold (e.g., "F3<=19.4" indicates a test speed not exceeding 19.4 km / h), and the leaf nodes directly provide the rollover category (Class=1 or 0).This structure not only clearly reveals the influence path of each parameter on the rollover result, but also directly extracts the comprehensive critical condition combination under the coupling effect of multiple parameters. Compared with black-box prediction relying solely on implicit networks, this explicit model provides experimental engineers with intuitive and operable technical guidance. Compared with the overfitting or underfitting problems caused by insufficient samples in traditional explicit models, this invention significantly improves the generalization ability and robustness of the explicit model by expanding high-quality labeled data through implicit networks, thereby achieving the goal of high-precision, interpretable, and low-cost prediction of vehicle rollover critical conditions.

[0038] Optionally, the step of inputting the actual rollover parameters of the target vehicle into the explicit neural network, and determining the critical conditions that trigger the rollover of the target vehicle through the explicit neural network, includes: The actual rollover parameters of the target vehicle are input into the explicit neural network to obtain the corresponding target decision path; Analyze each decision node in the target decision path and extract the parameter thresholds corresponding to the rollover parameters of each vehicle that are related to the rollover result corresponding to the actual rollover parameters. The parameter thresholds corresponding to the rollover parameters of each vehicle are used as the critical conditions for triggering vehicle rollover.

[0039] In practical applications, when it's necessary to predict the rollover critical conditions of a target vehicle, its actual rollover parameters (such as current vehicle speed, slope angle, load distribution, etc.) are input into a pre-trained explicit neural network. The network will make judgments along a specific decision path, which consists of a series of parameter threshold comparisons. By analyzing this target decision path, the system extracts the vehicle rollover parameters and their corresponding parameter thresholds involved in all judgment nodes along the path. For example, the path might sequentially judge whether "the slope angle is greater than 18 degrees," "the vehicle speed exceeds 50 km / h," and "the rear axle mass is less than 1200 kg," etc. These thresholds together constitute the set of critical conditions that trigger the rollover of the target vehicle. Test engineers can set test conditions based on these explicit values ​​to accurately approximate the rollover boundary, thereby efficiently completing airbag deployment calibration. This process not only avoids the high cost and low efficiency of traditional trial-and-error methods but also, due to the model's generalization ability, can be applied to vehicles of the same or even different types, significantly improving R&D efficiency and safety.

[0040] In summary, the vehicle rollover critical condition prediction method provided in this specification involves acquiring n sets of rollover data, including vehicle rollover parameters and rollover results; performing new data synthesis processing based on the n sets of vehicle rollover parameters to obtain m sets of enhanced rollover parameters corresponding to the vehicle rollover parameters; inputting the enhanced rollover parameters into an implicit neural network to output the rollover results corresponding to the enhanced rollover parameters; wherein the implicit neural network is trained from the n sets of rollover data; training an initial explicit neural network using the enhanced rollover parameters and the corresponding rollover results to obtain a trained explicit neural network; and inputting the actual rollover parameters of the target vehicle into the explicit neural network to determine the critical conditions that trigger the target vehicle rollover. Thus, an implicit neural network is first used to learn rollover discrimination capabilities from small-scale real-vehicle rollover data. Then, combining parameter distribution characteristics, a large number of reasonable virtual samples are generated based on physical constraints, and these are labeled using the trained implicit neural network to construct a large number of enhanced rollover parameters. These parameters are then used to train an explicit neural network to output the critical conditions for vehicle rollover expressed in the form of specific parameter thresholds. This entire process overcomes the bottleneck of insufficient real-vehicle data while ensuring the physical rationality and engineering feasibility of the generated rules.

[0041] Based on the same inventive concept, combined with Figure 2 As shown, embodiments of the present invention also provide a vehicle rollover critical condition prediction device, including... The acquisition module acquires n sets of rollover data about the vehicle rollover, and each set of rollover data includes vehicle rollover parameters and rollover result; The synthesis module is used to perform new data synthesis processing based on the n sets of vehicle rollover parameters to obtain m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters; where m is greater than n, and m and n are both integers greater than 0; The first training module is used to input the enhanced roll parameters into the implicit neural network and output the roll result corresponding to the enhanced roll parameters; wherein, the implicit neural network is trained from n sets of roll data; The second training module is used to train the initial explicit neural network using the enhanced roll parameters and the roll results corresponding to the enhanced roll parameters, so as to obtain the trained explicit neural network. The output module is used to input the actual rollover parameters of the target vehicle into the explicit neural network, and through the explicit neural network, determine the critical conditions that trigger the rollover of the target vehicle.

[0042] Optionally, the synthesis module is also used for: The value range and distribution characteristics of each parameter in the vehicle rollover parameters are determined respectively; Based on the value range and distribution characteristics of each parameter in the vehicle rollover parameters, generate m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters; Optionally, the synthesis module is also used for: For each parameter in the vehicle rollover parameters, within the range of values ​​for that parameter, an enhancement parameter set corresponding to that parameter is generated through interpolation processing. The enhancement parameter set includes multiple values ​​for that parameter. Based on the distribution characteristics of each parameter in the vehicle rollover parameters, physical constraint relationships are constructed; Based on the physical constraint relationship, m enhanced roll parameters are generated from the enhanced parameter set corresponding to each parameter.

[0043] Optionally, the synthesis module is also used for: Based on the distribution characteristics of each parameter in the vehicle rollover parameters, the physical correlation between each vehicle rollover parameter is determined; Based on the physical correlation, combination rules are set for the enhanced parameter sets of each parameter generated in the interpolation process to construct physical constraint relationships.

[0044] Optionally, the training steps of the implicit neural network include: The n sets of vehicle rollover parameters and corresponding rollover results are divided into a training set, a validation set, and a test set. The initial implicit neural network is trained using the training set, and the network hyperparameters are adjusted based on the loss function values ​​of the validation set. When the prediction accuracy of the initial implicit neural network on the test set reaches a preset threshold, the trained implicit neural network is obtained.

[0045] Optionally, the training steps of the explicit neural network include: The m sets of enhanced tumbling parameters and their corresponding tumbling results are used as the training dataset; Based on the training dataset, information entropy is used as the splitting criterion to train the initial explicit neural network until a preset stopping condition is met, thus obtaining the trained explicit neural network.

[0046] Optionally, the output module is also used for: The actual rollover parameters of the target vehicle are input into the explicit neural network to obtain the corresponding target decision path; Analyze each decision node in the target decision path and extract the parameter thresholds corresponding to the rollover parameters of each vehicle that are related to the rollover result corresponding to the actual rollover parameters. The parameter thresholds corresponding to the rollover parameters of each vehicle are used as the critical conditions for triggering vehicle rollover.

[0047] In summary, the vehicle rollover critical condition prediction device provided in this specification acquires n sets of rollover data related to vehicle rollover, including vehicle rollover parameters and rollover results; based on the n sets of vehicle rollover parameters, new data synthesis processing is performed to obtain m sets of enhanced rollover parameters corresponding to the vehicle rollover parameters; the enhanced rollover parameters are input into an implicit neural network, and the rollover results corresponding to the enhanced rollover parameters are output; wherein, the implicit neural network is trained from the n sets of rollover data; an initial explicit neural network is trained using the enhanced rollover parameters and the rollover results corresponding to the enhanced rollover parameters to obtain a trained explicit neural network; the actual rollover parameters of the target vehicle are input into the explicit neural network, and the critical conditions that trigger the rollover of the target vehicle are determined through the explicit neural network. Thus, an implicit neural network is first used to learn rollover discrimination capabilities from small-scale real-vehicle rollover data. Then, combining parameter distribution characteristics, a large number of reasonable virtual samples are generated based on physical constraints, and these are labeled using the trained implicit neural network to construct a large number of enhanced rollover parameters. These parameters are then used to train an explicit neural network to output the critical conditions for vehicle rollover expressed in the form of specific parameter thresholds. This entire process overcomes the bottleneck of insufficient real-vehicle data while ensuring the physical rationality and engineering feasibility of the generated rules.

[0048] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the vehicle rollover critical condition prediction device described above can be referred to the corresponding process in the aforementioned method, and will not be elaborated further here.

[0049] Based on the same inventive concept, this embodiment provides an electronic device including a vehicle rollover critical condition prediction device, a memory, a processor, and a communication unit. The memory stores machine-readable instructions that can be executed by the processor. When the electronic device is running, the processor and the memory communicate through a bus. The processor executes the machine-readable instructions and performs the vehicle rollover critical condition prediction method.

[0050] The memory, processor, and communication unit are electrically connected directly or indirectly to achieve signal transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines. The vehicle rollover critical condition prediction device includes at least one software functional module that can be stored in the memory in the form of software or firmware. The processor is used to execute the executable module stored in the memory (e.g., the software functional module or computer program included in the vehicle rollover critical condition prediction device).

[0051] The memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc.

[0052] In some embodiments, the processor is used to perform one or more functions described in this embodiment. In some embodiments, the processor may include one or more processing cores (e.g., a single-core processor (S) or a multi-core processor (S)). By way of example only, the processor may include a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), an Application Specific Instruction-set Processor (ASIP), a Graphics Processing Unit (GPU), a Physics Processing Unit (PPU), a Digital Signal Processor (DSP), a Field Programmable Gate Array (FPGA), a Programmable Logic Device (PLD), a controller, a microcontroller unit, a Reduced Instruction Set Computing (RISC) computer, or a microprocessor, or any combination thereof.

[0053] For ease of explanation, only one processor is described in the electronic device. However, it should be noted that the electronic device in this embodiment may also include multiple processors, and therefore the steps performed by one processor as described in this embodiment may also be performed jointly or individually by multiple processors. For example, if the server's processor performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, one processor performs step A, and a second processor performs step B, or the first and second processors jointly perform steps A and B.

[0054] In this embodiment, the memory is used to store the program, and the processor is used to execute the program after receiving the execution instruction. The process definition method disclosed in any implementation of this embodiment can be applied to the processor, or implemented by the processor.

[0055] The communication unit is used to establish communication connections between electronic devices and other devices via a network, and to send and receive data via the network.

[0056] In some implementations, the network can be any type of wired or wireless network, or a combination thereof. By way of example only, the network may include wired networks, wireless networks, fiber optic networks, telecommunications networks, intranets, the Internet, local area networks (LANs), wide area networks (WANs), wireless local area networks (WLANs), metropolitan area networks (MANs), public switched telephone networks (PSTNs), Bluetooth networks, ZigBee networks, or near field communication (NFC) networks, or any combination thereof.

[0057] In this embodiment, the electronic device may be, but is not limited to, a laptop, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), or other electronic devices. This embodiment does not impose any restrictions on the specific type of electronic device.

[0058] Based on the above, this embodiment provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the vehicle rollover critical condition prediction method of any of the aforementioned embodiments.

[0059] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the readable storage medium described above can be referred to the corresponding process in the aforementioned method, and will not be elaborated further here.

[0060] The above are merely various embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting critical conditions for vehicle rollover, characterized in that, include: Obtain n sets of rollover data related to vehicle rollover, each set of rollover data including vehicle rollover parameters and rollover result; Based on the n sets of vehicle rollover parameters, new data synthesis processing is performed to obtain m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters; where m is greater than n, and m and n are both integers greater than 0; The enhanced roll parameters are input into an implicit neural network, which outputs a roll result corresponding to the enhanced roll parameters; wherein, the implicit neural network is trained from the n sets of roll data; The initial explicit neural network is trained using the enhanced roll parameters and the roll results corresponding to the enhanced roll parameters to obtain the trained explicit neural network. The actual rollover parameters of the target vehicle are input into the explicit neural network, and the critical conditions that trigger the rollover of the target vehicle are determined through the explicit neural network.

2. The method according to claim 1, characterized in that, The process involves synthesizing new data based on n sets of vehicle rollover parameters to obtain m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters, including: The value range and distribution characteristics of each parameter in the vehicle rollover parameters are determined respectively; Based on the value range and distribution characteristics of each parameter in the vehicle rollover parameters, m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters are generated.

3. The method according to claim 2, characterized in that, The step of generating m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters, based on the value range and distribution characteristics of each parameter in the vehicle rollover parameters, includes: For each parameter in the vehicle rollover parameters, within the range of values ​​for that parameter, an enhancement parameter set corresponding to that parameter is generated through interpolation processing. The enhancement parameter set includes multiple values ​​for that parameter. Based on the distribution characteristics of each parameter in the vehicle rollover parameters, physical constraint relationships are constructed; Based on the physical constraint relationship, m enhanced roll parameters are generated from the enhanced parameter set corresponding to each parameter.

4. The method according to claim 3, characterized in that, The step of constructing physical constraint relationships based on the distribution characteristics of each parameter in the vehicle rollover parameters includes: Based on the distribution characteristics of each parameter in the vehicle rollover parameters, the physical correlation between each vehicle rollover parameter is determined; Based on the physical correlation, combination rules are set for the enhanced parameter sets of each parameter generated in the interpolation process to construct physical constraint relationships.

5. The method according to claim 1, characterized in that, The training steps of the implicit neural network include: The n sets of vehicle rollover parameters and corresponding rollover results are divided into a training set, a validation set, and a test set. The initial implicit neural network is trained using the training set, and the network hyperparameters are adjusted based on the loss function values ​​of the validation set. When the prediction accuracy of the initial implicit neural network on the test set reaches a preset threshold, the trained implicit neural network is obtained.

6. The method according to claim 1, characterized in that, The training steps of the explicit neural network include: The m sets of enhanced tumbling parameters and their corresponding tumbling results are used as the training dataset; Based on the training dataset, information entropy is used as the splitting criterion to train the initial explicit neural network until a preset stopping condition is met, thus obtaining the trained explicit neural network.

7. The method according to claim 1, characterized in that, The step of inputting the actual rollover parameters of the target vehicle into the explicit neural network, and determining the critical conditions that trigger the rollover of the target vehicle through the explicit neural network, includes: The actual rollover parameters of the target vehicle are input into the explicit neural network to obtain the corresponding target decision path; Analyze each decision node in the target decision path and extract the parameter thresholds corresponding to the rollover parameters of each vehicle that are related to the rollover result corresponding to the actual rollover parameters. The parameter thresholds corresponding to the rollover parameters of each vehicle are used as the critical conditions for triggering vehicle rollover.

8. A vehicle rollover critical condition prediction device, characterized in that, include The acquisition module acquires n sets of rollover data about the vehicle rollover, and each set of rollover data includes vehicle rollover parameters and rollover result; The synthesis module is used to perform new data synthesis processing based on the n sets of vehicle rollover parameters to obtain m sets of enhanced rollover parameters corresponding to the n sets of vehicle rollover parameters; where m is greater than n, and m and n are both integers greater than 0; The first training module is used to input the enhanced roll parameters into the implicit neural network and output the roll result corresponding to the enhanced roll parameters; wherein, the implicit neural network is trained from n sets of roll data; The second training module is used to train the initial explicit neural network using the enhanced roll parameters and the roll results corresponding to the enhanced roll parameters, so as to obtain the trained explicit neural network. The output module is used to input the actual rollover parameters of the target vehicle into the explicit neural network, and through the explicit neural network, determine the critical conditions that trigger the rollover of the target vehicle.

9. An electronic device, characterized in that, The electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the vehicle rollover critical condition prediction method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the vehicle rollover critical condition prediction method as described in any one of claims 1-7.

Images