Collision waveform prediction method and device, vehicle and storage medium

Abstract

The invention relates to a collision waveform prediction method and device, a vehicle and a storage medium, and the method comprises the steps: obtaining a plurality of groups of collision state data of a plurality of sample vehicles under different collision working conditions and a collision waveform signal corresponding to each group of collision state data, and constructing a sample set through the plurality of groups of collision state data and the collision waveform signals; performing data processing on the sample set to obtain a principal component of a collision waveform meeting a preset screening condition; constructing a model training set by using the principal component and the sample set, and calculating a waveform mean value and a principal component waveform of the sample set; and training a pre-constructed initial collision waveform prediction model by using the model training set to obtain a final collision waveform prediction model, and predicting the collision waveform of the target vehicle model under the target collision working condition by using the final collision waveform prediction model, the waveform mean value and the principal component waveform. Therefore, the technical problem that the accuracy of the waveform data obtained by the model obtained by training is insufficient due to the fact that the acquisition cost of the high-quality collision waveform data is relatively high and the data capable of being used for model training is relatively small in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and particularly relates to a method and device for predicting a collision waveform, a vehicle, and a storage medium. Background Art

[0002] Driven by the trends of electrification, intelligence, and networking in the development of automobiles, traditional passive safety technologies are also transforming towards intelligence to provide more comprehensive, multi-scenario, and adaptive protection for the drivers and passengers of autonomous vehicles. To achieve this goal, it is necessary to establish an accurate and real-time pre-collision perception model and occupant injury prediction technology, and form a collision mitigation strategy before a collision. The occupant injury prediction technology refers to extracting information related to the severity of a collision accident based on the in-vehicle perception system of an intelligent vehicle to accurately evaluate the severity of the injuries of the occupants in the cabin during the accident. The variables that have a key impact on the occupant injury situation can be divided into two aspects: inside and outside the cabin, specifically including: information inside the cabin - the physiological characteristics of the occupants (such as age, gender, height, weight, etc.), sitting postures (seat orientation, backrest angle, body posture, etc.), and restraint system information (seat belts, airbags, etc.); information outside the cabin - vehicle physical parameters (mass, size, stiffness, etc.), obstacle types (cars, pedestrians, buildings, etc.), collision speed, and collision location, etc. The collision waveform prediction technology can be used as a pre-process for occupant injury prediction, using information outside the cabin to predict the collision signal / waveform generated by inertial force inside the cabin after a collision, and combining the collision signal with information inside the cabin can predict the injuries generated in different parts of the occupants.

[0003] Machine learning methods have been widely applied to the task of predicting collision waveforms. To establish a reliable waveform prediction model, it is necessary to use high-quality data for training as much as possible. The highest-quality data is real vehicle collision data, but the cost of obtaining real vehicle collision data is extremely high, and a high-speed collision experiment will consume a prototype vehicle. Collision data of slightly lower quality is data obtained by finite element simulation. By increasing the finite element mesh and shortening the calculation step size, the accuracy of the simulation can be improved to make the simulation data closer to real collision data. However, the price is to consume more computing resources. To obtain an example of high-quality simulation data, even when simulating on a high-performance computing cluster, it takes several hours.

[0004] Therefore, the cost of obtaining high-quality collision waveform data is relatively high, and there is less data available for model training, which urgently needs to be improved. Summary of the Invention

[0005] The present application provides a method and device for predicting a collision waveform, a vehicle, and a storage medium to solve the technical problem in the related art that the cost of obtaining high-quality collision waveform data is relatively high, there is less data available for model training, and the accuracy of the waveform data obtained by the trained model is insufficient.

[0006] An embodiment of the first aspect of the present application provides a method for predicting a collision waveform, which is applied to the model construction stage. The method includes the following steps: obtaining multiple groups of collision state data of multiple sample vehicles under different collision conditions and the corresponding collision waveform signals for each group of collision state data, and constructing a sample set by using the multiple groups of collision state data and the collision waveform signals; processing the data of the sample set to obtain the principal components of the collision waveform that meet the preset screening conditions; constructing a model training set by using the principal components and the sample set, and calculating the waveform mean value and the principal component waveforms of the sample set; training a pre-constructed initial collision waveform prediction model by using the model training set to obtain a final collision waveform prediction model, so as to predict the collision waveform of a target vehicle model under a target collision condition by using the final collision waveform prediction model, the waveform mean value and the principal component waveforms.

[0007] According to the above technical means, the embodiments of the present application can combine the principal component analysis method to reduce the difficulty of model training, reduce the algorithm resource consumption, and combine the Gaussian process regression to reduce the demand for training samples, so as to achieve a more low-cost and more efficient collision waveform prediction.

[0008] Optionally, in an embodiment of the present application, the processing the data of the sample set to obtain the principal components of the collision waveform that meet the preset screening conditions includes: constructing a collision signal matrix by using all the collision waveform signals in the sample set; decomposing the collision signal matrix to obtain the principal component variances and the principal component vector matrix; screening in the principal component vector matrix based on the principal component variances to obtain the principal components that meet the preset screening conditions.

[0009] According to the above technical means, the embodiments of the present application can perform principal component analysis on the data in the sample set, so as to use the principal component analysis method to reduce the difficulty of model training and the algorithm resource consumption, and improve the response speed of the prediction model.

[0010] Optionally, in an embodiment of the present application, the constructing a model training set by using the principal components and the sample set, and calculating the waveform mean value and the principal component waveforms of the sample set includes: calculating the component coefficients of all samples in the sample set on different principal components; constructing a model training set based on the component coefficients and the sample set.

[0011] According to the above technical means, the embodiments of the present application can use the principal component analysis method to construct a model training set, so as to reduce the difficulty of model training and the algorithm resource consumption, and improve the response speed of the prediction model.

[0012] The second aspect of the embodiments of the present application provides a method for predicting a collision waveform, which is applied to the model usage stage. The method includes the following steps: obtaining the collision state data of a target vehicle model under an expected collision condition; inputting the collision state data into a pre-constructed final collision waveform prediction model to predict the principal component components in the form of a waveform, where the final collision waveform prediction model is trained by a model training set; reconstructing the waveform based on the principal component components, the waveform mean calculated from a preset sample set, and the principal component waveform to obtain a waveform reconstruction result, and obtaining the predicted collision waveform of the target vehicle model under the expected collision condition according to the waveform reconstruction result, where the preset sample set is constructed from multiple groups of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each group of collision state data.

[0013] According to the above technical means, the embodiments of the present application can use the trained model to predict the collision waveform and reconstruct the waveform based on the data of principal component analysis in the sample set, so as to obtain a more accurate prediction result.

[0014] The third aspect of the embodiments of the present application provides a device for predicting a collision waveform, which is applied to the model construction stage. The device includes: an acquisition module, configured to acquire multiple groups of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each group of collision state data, and construct a sample set by using the multiple groups of collision state data and the collision waveform signals; a processing module, configured to perform data processing on the sample set to obtain the principal components of the collision waveform that meet the preset screening conditions; a calculation module, configured to construct a model training set by using the principal components and the sample set, and calculate the waveform mean and the principal component waveform of the sample set; a training module, configured to train a pre-constructed initial collision waveform prediction model by using the model training set to obtain a final collision waveform prediction model, so as to predict the collision waveform of a target vehicle model under a target collision condition by using the final collision waveform prediction model, the waveform mean, and the principal component waveform.

[0015] Optionally, in an embodiment of the present application, the processing module includes: a first construction unit, configured to construct a collision signal matrix by using all the collision waveform signals in the sample set; a decomposition unit, configured to decompose the collision signal matrix to obtain the principal component variance and the principal component vector matrix; a screening unit, configured to perform screening in the principal component vector matrix based on the principal component variance to obtain the principal components that meet the preset screening conditions.

[0016] Optionally, in an embodiment of the present application, the calculation module includes: a calculation unit, configured to calculate the component coefficients of all samples in the sample set on different principal components; a second construction unit, configured to construct a model training set based on the component coefficients and the sample set.

[0017] In the fourth aspect of the embodiments of the present application, a prediction device for collision waveforms is provided, which is applied in the model usage stage. The device includes: an acquisition module for acquiring collision state data of a target vehicle model under an expected collision condition; a prediction module for inputting the collision state data into a pre-constructed final collision waveform prediction model to predict the principal component components in the form of waveforms, where the final collision waveform prediction model is trained by a model training set; a reconstruction module for reconstructing waveforms based on the principal component components, the waveform mean calculated from a preset sample set, and the principal component waveforms to obtain a waveform reconstruction result, and obtaining a predicted collision waveform of the target vehicle model under the expected collision condition according to the waveform reconstruction result, where the preset sample set is constructed from multiple sets of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each set of collision state data.

[0018] In the fifth aspect of the embodiments of the present application, a vehicle is provided, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the program to implement the collision waveform prediction method as described in the above embodiments.

[0019] In the sixth aspect of the embodiments of the present application, a computer-readable storage medium is provided, where the computer-readable storage medium stores computer instructions for causing the computer to execute the collision waveform prediction method as described in the above embodiments.

[0020] In the seventh aspect of the embodiments of the present application, a computer program product is provided, including a computer program, which when executed, is used to implement the collision waveform prediction method as above.

[0021] The embodiments of the present application can construct a sample set from multiple sets of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each set of collision state data, and obtain the principal components of the collision waveforms that meet the preset screening conditions, and then construct a model training set, and calculate the waveform mean and principal component waveforms of the sample set, and use the model training set to train a pre-constructed initial collision waveform prediction model to obtain a final collision waveform prediction model, so as to predict the collision waveform of the target vehicle model under the target collision condition by using the final collision waveform prediction model, the waveform mean, and the principal component waveforms, effectively utilize real vehicle and simulation collision data, the training process is more stable, and the training result is less affected by model parameters. While reducing costs, waveform prediction can be carried out more efficiently. Thus, the technical problem in the related art that the acquisition cost of high-quality collision waveform data is relatively high, the data available for model training is less, and the accuracy of the waveform data obtained by the trained model is insufficient is solved.

[0022] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present application. Brief Description of the Drawings

[0023] The above-mentioned and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of embodiments in conjunction with the accompanying drawings, where:

[0024] Figure 1 is a flowchart of a method for predicting a collision waveform provided according to an embodiment of the present application;

[0025] Figure 2 is a schematic diagram of the principle of Gaussian process training according to an embodiment of the present application;

[0026] Figure 3 is a schematic diagram of the principle of a method for predicting a collision waveform according to an embodiment of the present application;

[0027] Figure 4 is a schematic structural diagram of a device for predicting a collision waveform provided according to an embodiment of the present application;

[0028] Figure 5 is a flowchart of another method for predicting a collision waveform provided according to an embodiment of the present application;

[0029] Figure 6 is a schematic structural diagram of another device for predicting a collision waveform provided according to an embodiment of the present application;

[0030] Figure 7 is a schematic structural diagram of a vehicle provided according to an embodiment of the present application. Detailed Description of the Embodiments

[0031] Embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present application and should not be construed as limiting the present application.

[0032] The following describes a method, apparatus, vehicle, and storage medium for predicting a collision waveform according to an embodiment of the present application. In view of the technical problem in the related art mentioned in the above background art that the acquisition cost of high-quality collision waveform data is relatively high, and there is less data available for model training, resulting in insufficient accuracy of the waveform data obtained by the trained model, the present application provides a method for predicting a collision waveform. In this method, a sample set can be constructed with multiple sets of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each set of collision state data, and the principal components of the collision waveforms that meet the preset screening conditions can be obtained. Then, a model training set can be constructed, and the waveform mean value of the sample set and the principal component waveforms can be calculated. The initially constructed initial collision waveform prediction model can be trained using the model training set to obtain a final collision waveform prediction model. Thus, the collision waveform of the target vehicle model under the target collision condition can be predicted using the final collision waveform prediction model, the waveform mean value, and the principal component waveforms, effectively utilizing real vehicle and simulation collision data. The training process is more stable, and the training result is less affected by model parameters. While reducing costs, waveform prediction can be performed more efficiently. Thereby, the technical problem in the related art that the acquisition cost of high-quality collision waveform data is relatively high, and there is less data available for model training, resulting in insufficient accuracy of the waveform data obtained by the trained model is solved.

[0033] Specifically, Figure 1 FIG. is a schematic flowchart of a method for predicting a collision waveform provided by an embodiment of the present application.

[0034] As Figure 1 shown, the method for predicting a collision waveform is applied to the model construction stage, and the method includes the following steps:

[0035] In step S101, multiple sets of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each set of collision state data are obtained, and a sample set is constructed using the multiple sets of collision state data and the collision waveform signals.

[0036] It can be understood that in order to establish a reliable waveform prediction model, high-quality data needs to be used for training, and low-precision simulation data should be avoided. Considering the characteristic of less training data, the embodiment of the present application can implement the construction of a collision waveform prediction model based on Gaussian process regression and principal component analysis.

[0037] Among them, Gaussian process regression is a typical non-parametric model. Different from neural network which is a parametric model, it does not construct the mapping relationship between input and output variables, but constructs the correlation relationship between samples. Gaussian process regression can be understood as a generalized interpolation method. Different from algorithms such as K-nearest neighbor or radial basis function interpolation that use fixed parameters, the parameters in the Gaussian process model can be learned from training samples. Although Gaussian process has advantages such as high stability and low overfitting risk for tasks with few samples, since it is a non-parametric model, all training samples need to be called during prediction. Due to the high time dimension of collision acceleration data, this results in consuming more memory resources each time prediction is made, and it is also difficult to apply to real-time prediction. For tasks with high timeliness, a solution with lower consumption of computing resources and storage resources is required. PCA (Principal Component Analysis) is a widely used algorithm for data compression, which decomposes data into a linear combination of principal components. During prediction, only a fixed number of principal components need to be stored in memory, rather than all training data, which can save computing resources and storage resources.

[0038] On this basis, the embodiments of the present application can realize collision waveform prediction based on Gaussian process regression and principal component analysis.

[0039] First, the embodiments of the present application can construct a sample set.

[0040] For example, the sample set can be expressed as (x 1 , S 1 ), …, (x N , S N ), where N represents the number of training samples, is a vector representing the collision state, including collision speed, angle, overlap rate, etc., and D represents the number of states. is the collision waveform signal, and T is the dimension of the waveform signal. On this basis, the embodiments of the present application can also calculate the waveform mean of the sample set, and the waveform mean is the average value of all collision waveforms as

[0041] In step S102, the sample set is processed to obtain the principal components of the collision waveforms that meet the preset screening conditions.

[0042] Furthermore, the embodiments of the present application can process the sample set, screen the processed data so that the screened data can be used for model training, and use PCA to decompose the data into a linear combination of principal components to save computing resources and storage resources.

[0043] Optionally, in an embodiment of the present application, data processing is performed on the sample set to obtain the principal components of the collision waveforms that meet the preset screening conditions, including: constructing a collision signal matrix using all the collision waveform signals in the sample set; decomposing the collision signal matrix to obtain the principal component variances and the principal component vector matrix; and screening in the principal component vector matrix based on the principal component variances to obtain the principal components that meet the preset screening conditions.

[0044] The embodiment of the present application can denote the matrix as the collision waveform matrix that has been centralized, and the average value of each column of this matrix is 0. Using Singular Value Decomposition (SVD), the collision signal matrix S is decomposed into the product of three parts: U, D, and V:

[0045] S = UDV T ,

[0046] where is the principal component vector matrix, each column represents a principal component, D is a diagonal matrix, and the diagonal elements represent the principal component variances. The principal components with larger variances can better represent the important features of the training data.

[0047] The principal component vectors are screened according to the principal component variances indicating that the number of screened principal components is K. Those skilled in the art can give the number of principal components, and then retain the K principal components with the largest principal component variances; or those skilled in the art give a threshold ε, and then retain all the principal components with principal component variances greater than the threshold.

[0048] In step S103, a model training set is constructed using the principal components and the sample set, and the waveform mean value of the sample set and the principal component waveforms are calculated.

[0049] In the actual execution process, the embodiment of the present application can construct a model training set using the screened principal components and the sample set, and calculate the waveform mean value of the sample set and the principal component waveforms to output the features for model training Output the waveform mean value for waveform reconstruction and the principal component waveforms

[0050] Optionally, in an embodiment of the present application, a model training set is constructed using the principal components and the sample set, and the waveform mean value of the sample set and the principal component waveforms are calculated, including: calculating the component coefficients of all samples in the sample set on different principal components; and constructing a model training set based on the component coefficients and the sample set.

[0051] Using the formula to calculate the component coefficients of all samples on different principal components, where the component matrix can be decomposed by rows and columns as follows: represents the components of the nth sample on all principal components, and represents the components of all samples on the kth principal component.

[0052] In step S104, the pre-constructed initial collision waveform prediction model is trained using the model training set to obtain the final collision waveform prediction model, so as to predict the collision waveform of the target vehicle model under the target collision condition using the final collision waveform prediction model, the waveform mean, and the principal component waveform.

[0053] As a possible implementation manner, the embodiment of the present application can be as Figure 2 shown, initialize the objective function and the model parameter θ 0 , β, where θ 0 > 0, η i > 0, β > 0.

[0054] The embodiment of the present application can calculate the covariance C(x , x ′ ) between different working conditions according to the input feature n,j , x n,k ), kernel function k(x, x

[0055]

[0056] C(x n , x m ) = k(x n , x m ) + β -1 δ nm ,

[0057] where δ nm is an indicator function, which takes the value of 1 if and only if m = n, otherwise 0. Denote the covariance matrix C N as an N-row and N-column symmetric matrix, and the element in the mth row and nth column is C(x n , x m ).

[0058] According to the covariance matrix C N and calculate the objective function

[0059]

[0060] where ω k is the kth column vector of W, and θ generally refers to all undetermined parameters.

[0061] Judge the relative change amount of the objective function Whether it is less than a certain threshold. When satisfied, output the current model parameters and terminate the process. Otherwise, calculate the objective function For the parameter θ i partial derivative:

[0062]

[0063] where θ i generally refers to all undetermined parameters. Denote the gradient is composed of partial derivatives with respect to all undetermined parameters.

[0064] Finally, the embodiment of the present application can update the model parameter θ according to the gradient using the BFGS method. Among them, BFGS is a gradient-based quasi-Newton method, which is widely used to solve optimization problems. And after completion, jump to the calculation of the kernel function and covariance until the model parameters are determined to obtain the final collision waveform prediction model, and combine the waveform mean and the principal component waveform to realize waveform prediction and reconstruction.

[0065] Combined with Figure 2 and Figure 3 As shown, the working principle of the collision waveform prediction method of the embodiment of the present application is elaborated in detail with an embodiment.

[0066] As Figure 3 shown, the embodiment of the present application can include two parts: model training and prediction. Specifically, it can include the following steps:

[0067] Step S1: Principal component analysis.

[0068] The embodiment of the present application can construct a sample set. Among them, in the sample set, the training data for model training can be (x 1 , S 1 ),…,(x N , S H ), where N represents the number of training samples, is a vector representing the collision state, including collision speed, angle, overlap rate, etc., and D represents the number of states. is the collision waveform signal, and T is the dimension of the waveform signal. The waveform mean is the average of all collision waveforms as

[0069] Furthermore, the embodiment of the present application can denote the matrix represents the collision waveform matrix that has been centered, and the average value of each column of this matrix is 0. Using singular value decomposition, decompose the collision signal matrix S into the product of three parts: U, D, and V:

[0070] S = UDV T

[0071] Among them, is the principal component vector matrix, and each column represents a principal component. D is a diagonal matrix, and the diagonal elements represent the variances of the principal components. The principal component with a larger variance can better represent the important features of the training data.

[0072] The embodiments of the present application can screen out the principal component vectors according to the variances of the principal components indicates that the number of selected principal components is K. Those skilled in the art can give the number of principal components, and then retain the K principal components with the largest variances of the principal components; or those skilled in the art give a threshold ε, and then retain all the principal components with variances of the principal components greater than the threshold.

[0073] The embodiments of the present application can use the formula to calculate the component coefficients of all training samples on different principal components. The component matrix can be decomposed as follows by rows and columns: represents the components of the nth sample on all principal components, represents the components of all training samples on the kth principal component.

[0074] The embodiments of the present application can output for training the Gaussian process model in step S2; and output the waveform mean and the principal component waveforms for waveform reconstruction in step S4.

[0075] Step S2: Gaussian process training.

[0076] The embodiments of the present application can, as Figure 2 shown, initialize the objective function and the model parameters θ 0 , β, which need to satisfy θ 0 > 0, η i > 0, β > 0.

[0077] The embodiments of the present application can, according to the input features kernel function k(x, x ′ ) and the model parameters, calculate the covariance C(x n,j , x n,j ) between different working conditions. The calculation methods of the kernel function and the covariance are as follows:

[0078]

[0079] C(x n , x m ) = k(x n , x m ) + β-1 δ nm ,

[0080] Among them, δ nm is an indicator function that takes the value of 1 if and only if m = n, and 0 otherwise. Denote the covariance matrix C N as an N-by-N symmetric matrix, and the element in the m-th row and n-th column is C(x n , x m ).

[0081] According to the covariance matrix C N and calculate the objective function

[0082]

[0083] Among them, ω k is the k-th column vector of W, and θ generally refers to all undetermined parameters.

[0084] Judge whether the relative change of the objective function is less than a certain threshold. When it is satisfied, output the current model parameters and terminate the process. Otherwise, calculate the objective function with respect to the parameter θ i partial derivative:

[0085]

[0086] Among them, θ i generally refers to all undetermined parameters. Denote the gradient which is composed of the partial derivatives with respect to all undetermined parameters.

[0087] Finally, the embodiment of the present application can update the model parameter θ according to the gradient using the BFGS method. Among them, BFGS is a gradient-based quasi-Newton method, which is widely used to solve optimization problems. And after completion, jump to the calculation of the kernel function and covariance until the model parameters are determined, obtain the final collision waveform prediction model, and combine the waveform mean and the principal component waveform to achieve waveform prediction and reconstruction.

[0088] Step S3: Gaussian process prediction.

[0089] The embodiment of the present application can input the collision state to be predicted that is, the collision state data of the target vehicle model under the expected collision condition, and the collision state of the training samples According to the kernel function k(x, x ′ ) and the already trained model parameter θ 0 , β, calculate all for n = 1, …, N, and denote the column vector

[0090] The embodiments of the present application can be based on the mean value of the collision signal obtained by principal component analysis the coefficient matrix W, the sample covariance matrix C N , the principal component The predicted collision signal can be calculated using the following formula

[0091]

[0092] Output the predicted principal component components

[0093] Step S4: Waveform reconstruction.

[0094] The embodiments of the present application can be based on the predicted principal component components the mean value of the collision signal output previously and the principal component to calculate the collision signal

[0095]

[0096] Output Complete the prediction process.

[0097] In summary, the embodiments of the present application combine the Gaussian process regression and the principal component analysis method to effectively address the problem of fewer high-quality training samples using the Gaussian process regression method. Compared with using a neural network for prediction, this model has a more stable training process, and the training results are less affected by model parameters; and the principal component analysis method is used to reduce the difficulty of model training, as well as the storage resources and computing resources consumed by the algorithm, enabling the waveform prediction model to be installed on a real vehicle and respond in real time, rather than just being used as an auxiliary tool in the R & D process. The model obtained by the embodiments of the present application can be used in multiple aspects:

[0098] (1) Using the predicted waveform and the in-cabin information to obtain the prediction of occupant injury;

[0099] (2) Using the predicted waveform to calculate the injury-related metrics as part of the objective function in the vehicle path planning problem;

[0100] (3) Using the waveform prediction model to improve the efficiency of daily simulation analysis;

[0101] (4) The predicted waveform can be used for the risk assessment of airbag deployment.

[0102] According to the collision waveform prediction method proposed by the embodiments of the present application, a sample set can be constructed with multiple sets of collision state data of various sample vehicles under different collision conditions and the corresponding collision waveform signals for each set of collision state data, and the principal components of the collision waveforms that meet the preset screening conditions can be obtained. Furthermore, a model training set can be constructed, and the waveform mean value and principal component waveforms of the sample set can be calculated. The initially constructed initial collision waveform prediction model can be trained using the model training set to obtain the final collision waveform prediction model, so as to predict the collision waveform of the target vehicle model under the target collision condition by using the final collision waveform prediction model, the waveform mean value, and the principal component waveforms. The real vehicle and simulation collision data can be effectively utilized, the training process is more stable, and the training result is less affected by model parameters. While reducing costs, waveform prediction can be carried out more efficiently. Thus, the technical problem in the related art that the acquisition cost of high-quality collision waveform data is relatively high, the data available for model training is scarce, and the waveform data accuracy obtained by the trained model is insufficient is solved.

[0103] Next, a collision waveform prediction device proposed according to the embodiments of the present application will be described with reference to the accompanying drawings.

[0104] Figure 4 It is a block diagram of the collision waveform prediction device according to the embodiments of the present application.

[0105] As Figure 4 shown, the collision waveform prediction device 10 is applied to the model construction stage. Among them, the device 10 includes: an acquisition module 101, a processing module 102, a calculation module 103, and a training module 104.

[0106] Specifically, the acquisition module 101 is configured to acquire multiple sets of collision state data of various sample vehicles under different collision conditions and the corresponding collision waveform signals for each set of collision state data, and construct a sample set by using the multiple sets of collision state data and the collision waveform signals.

[0107] The processing module 102 is configured to perform data processing on the sample set to obtain the principal components of the collision waveforms that meet the preset screening conditions.

[0108] The calculation module 103 is configured to construct a model training set by using the principal components and the sample set, and calculate the waveform mean value and the principal component waveforms of the sample set.

[0109] The training module 104 is configured to train the initially constructed initial collision waveform prediction model by using the model training set to obtain the final collision waveform prediction model, so as to predict the collision waveform of the target vehicle model under the target collision condition by using the final collision waveform prediction model, the waveform mean value, and the principal component waveforms.

[0110] Optionally, in an embodiment of the present application, the processing module 102 includes: a first construction unit, a decomposition unit, and a screening unit.

[0111] Among them, the first construction unit is used to construct a collision signal matrix by using all the collision waveform signals in the sample set.

[0112] The decomposition unit is used to decompose the collision signal matrix to obtain the principal component variances and the principal component vector matrix.

[0113] The screening unit is used to screen in the principal component vector matrix based on the principal component variances to obtain the principal components that meet the preset screening conditions.

[0114] Optionally, in an embodiment of the present application, the calculation module 103 includes: a calculation unit and a second construction unit.

[0115] Among them, the calculation unit is used to calculate the component coefficients of all samples in the sample set on different principal components.

[0116] The second construction unit is used to construct a model training set based on the component coefficients and the sample set.

[0117] It should be noted that the foregoing explanation of the embodiment of the collision waveform prediction method also applies to the collision waveform prediction device of this embodiment, and will not be elaborated here.

[0118] According to the collision waveform prediction device provided by the embodiment of the present application, a sample set can be constructed with multiple sets of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each set of collision state data, and the principal components of the collision waveforms that meet the preset screening conditions can be obtained. Furthermore, a model training set can be constructed, and the waveform mean value and the principal component waveforms of the sample set can be calculated. The initially constructed initial collision waveform prediction model is trained by using the model training set to obtain a final collision waveform prediction model. Thus, the collision waveform of the target vehicle model under the target collision condition can be predicted by using the final collision waveform prediction model, the waveform mean value and the principal component waveforms, effectively utilizing the real vehicle and simulation collision data. The training process is more stable, and the training result is less affected by model parameters. While reducing costs, waveform prediction can be carried out more efficiently. Thereby, the technical problem in the related art that the acquisition cost of high-quality collision waveform data is relatively high, and the data available for model training is less, resulting in insufficient accuracy of the waveform data obtained by the trained model is solved.

[0119] The above is the elaboration of the embodiment of the present application in the model construction stage. The following is the elaboration of the embodiment of the present application in the model usage stage.

[0120] Specifically, Figure 5 is a schematic flowchart of a collision waveform prediction method provided by an embodiment of the present application.

[0121] As Figure 5As shown, the method for predicting the collision waveform is applied in the model usage phase, and the method includes the following steps:

[0122] In step S501, obtain the collision state data of the target vehicle model under the desired collision condition.

[0123] In step S502, input the collision state data into the pre-constructed final collision waveform prediction model to predict the principal component components in the form of a waveform, where the final collision waveform prediction model is obtained by training with a model training set.

[0124] In step S503, based on the principal component components and the waveform mean and principal component waveforms calculated from the preset sample set, perform waveform reconstruction to obtain the waveform reconstruction result, and obtain the predicted collision waveform of the target vehicle model under the desired collision condition according to the waveform reconstruction result, where the preset sample set is constructed from multiple sets of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each set of collision state data.

[0125] According to the method for predicting the collision waveform proposed in the embodiment of the present application, a sample set can be constructed from multiple sets of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each set of collision state data, and the principal components of the collision waveforms that meet the preset screening conditions can be obtained, and then a model training set can be constructed, and the waveform mean and principal component waveforms of the sample set can be calculated, and the pre-constructed initial collision waveform prediction model can be trained with the model training set to obtain the final collision waveform prediction model, so as to use the final collision waveform prediction model, waveform mean and principal component waveforms to predict the collision waveform of the target vehicle model under the target collision condition, effectively utilize the real vehicle and simulation collision data, the training process is more stable, and the training result is less affected by model parameters. While reducing costs, waveform prediction can be carried out more efficiently. Thus, the technical problem in the related art that the acquisition cost of high-quality collision waveform data is relatively high, the data available for model training is relatively small, and the accuracy of the waveform data obtained by the trained model is insufficient is solved.

[0126] Next, describe the collision waveform prediction device proposed in the embodiment of the present application with reference to the accompanying drawings.

[0127] Figure 6 It is a block diagram of the collision waveform prediction device according to the embodiment of the present application.

[0128] As Figure 6 shown, the collision waveform prediction device 20 is applied in the model usage phase, and the device 20 includes: an acquisition module 201, a prediction module 202, and a reconstruction module 203.

[0129] Specifically, the acquisition module 201 is used to obtain the collision state data of the target vehicle model under the desired collision condition.

[0130] The prediction module 202 is configured to input the collision state data into a pre-constructed final collision waveform prediction model to predict the principal component components in the form of a waveform, where the final collision waveform prediction model is obtained by training with a model training set.

[0131] The reconstruction module 203 is configured to perform waveform reconstruction based on the principal component components, the waveform mean calculated from a preset sample set, and the principal component waveforms to obtain a waveform reconstruction result, and obtain the predicted collision waveform of the target vehicle under the desired collision condition according to the waveform reconstruction result, where the preset sample set is constructed from multiple groups of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each group of collision state data.

[0132] It should be noted that the foregoing explanation of the embodiments of the collision waveform prediction method also applies to the collision waveform prediction device of this embodiment, and will not be elaborated here.

[0133] According to the collision waveform prediction device provided by the embodiments of the present application, a sample set can be constructed from multiple groups of collision state data of multiple sample vehicles under different collision conditions and the collision waveform signals corresponding to each group of collision state data, and the principal components of the collision waveforms that meet the preset screening conditions can be obtained. Furthermore, a model training set can be constructed, the waveform mean and principal component waveforms of the sample set can be calculated, and the pre-constructed initial collision waveform prediction model can be trained with the model training set to obtain the final collision waveform prediction model, so as to use the final collision waveform prediction model, the waveform mean, and the principal component waveforms to predict the collision waveform of the target vehicle under the target collision condition, effectively using real vehicle and simulation collision data, with a more stable training process and the training result being less affected by model parameters. While reducing costs, waveform prediction can be performed more efficiently. Thus, the technical problem in the related art that the acquisition cost of high-quality collision waveform data is relatively high, and the data available for model training is relatively small, resulting in insufficient accuracy of the waveform data obtained by the trained model, is solved.

[0134] Figure 7 The structure diagram of the vehicle provided by the embodiments of the present application. The vehicle may include:

[0135] A memory 701, a processor 702, and a computer program stored on the memory 701 and executable on the processor 702.

[0136] When the processor 702 executes the program, it implements the collision waveform prediction method provided in the above embodiments.

[0137] Further, the vehicle further includes:

[0138] A communication interface 703 for communication between the memory 701 and the processor 702.

[0139] A memory 701 for storing a computer program that can run on a processor 702.

[0140] The memory 701 may include a high-speed RAM memory and may also include non-volatile memory, such as at least one disk memory.

[0141] If the memory 701, the processor 702, and the communication interface 703 are implemented independently, the communication interface 703, the memory 701, and the processor 702 can be interconnected via a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 only a thick line is shown in the figure, but it does not mean that there is only one bus or one type of bus.

[0142] Optionally, in a specific implementation, if the memory 701, the processor 702, and the communication interface 703 are integrated on a single chip, the memory 701, the processor 702, and the communication interface 703 can communicate with each other through an internal interface.

[0143] The processor 702 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0144] This embodiment also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the method for predicting a collision waveform as described above is implemented.

[0145] The embodiments of the present application also provide a computer program product, including a computer program, and when the computer program is executed by a processor, the method for predicting a collision waveform provided by the embodiments of the present invention is implemented.

[0146] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0147] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of this application, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0148] Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment, or portion of code including one or N executable instructions for implementing a customized logic function or process, and the scope of the preferred embodiments of this application includes additional implementations, where the functions can be executed in a manner that is not in the order shown or discussed, including in a substantially simultaneous manner according to the functions involved or in a reverse order, which should be understood by those skilled in the art to which the embodiments of this application pertain.

[0149] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection part (electronic device) having one or N wirings, a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically by optically scanning the paper or other media, followed by editing, interpretation, or otherwise processing as appropriate, and then stored in a computer memory.

[0150] It should be understood that various parts of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0151] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0152] In addition, in each embodiment of the present application, each functional unit can be integrated into one processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0153] The above-mentioned storage medium can be a read-only memory, a magnetic disk, an optical disc, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for predicting a collision waveform, characterized in that: Applied to the model building stage, wherein the method comprises the following steps: Acquire multiple groups of collision state data of multiple sample vehicles under different collision conditions and collision waveform signals corresponding to each group of collision state data, and construct a sample set using the multiple groups of collision state data and the collision waveform signals; Performing data processing on the sample set to obtain the principal component of the collision waveform that meets the preset screening conditions; Constructing a model training set using the principal component and the sample set, and calculating the waveform mean and principal component waveform of the sample set; The model training set is used to train a pre-built initial collision waveform prediction model to obtain a final collision waveform prediction model, so as to use the final collision waveform prediction model, the waveform mean and the principal component waveform to predict the collision waveform of the target vehicle model under the target collision condition.

2. The method according to claim 1, characterized in that: The data processing of the sample set to obtain the principal component of the collision waveform that meets the preset screening condition includes: Constructing a collision signal matrix using all collision waveform signals in the sample set; Decomposing the collision signal matrix to obtain a principal component variance and a principal component vector matrix; The principal component vector matrix is ​​screened based on the principal component variance to obtain the principal component that meets the preset screening condition.

3. The method according to claim 1, characterized in that The method of constructing a model training set by using the principal component and the sample set, and calculating the waveform mean and the principal component waveform of the sample set, includes: Calculate the component coefficients of all samples in the sample set on different principal components; A model training set is constructed based on the component coefficients and the sample set.

4. A method for predicting a collision waveform, characterized in that: Applied to the model use phase, wherein the method comprises the following steps: Obtain collision status data of the target vehicle model under expected collision conditions; Inputting the collision state data into a pre-built final collision waveform prediction model to obtain principal components by waveform prediction, wherein the final collision waveform prediction model is trained by a model training set; A waveform reconstruction is performed based on the principal component and the waveform mean and the principal component waveform calculated by a preset sample set to obtain a waveform reconstruction result, and a predicted collision waveform of the target vehicle model under the expected collision condition is obtained according to the waveform reconstruction result, wherein the preset sample set is constructed by multiple groups of collision state data of multiple sample vehicles under different collision conditions and collision waveform signals corresponding to each group of collision state data.

5. A collision waveform prediction device, characterized in that: Applied to the model building stage, wherein the device comprises: An acquisition module, used to acquire multiple groups of collision state data of multiple sample vehicles under different collision conditions and collision waveform signals corresponding to each group of collision state data, and construct a sample set using the multiple groups of collision state data and the collision waveform signals; A processing module, used for performing data processing on the sample set to obtain a principal component of a collision waveform that meets a preset screening condition; A calculation module, used to construct a model training set using the principal component and the sample set, and calculate the waveform mean and principal component waveform of the sample set; The training module is used to train a pre-built initial collision waveform prediction model using the model training set to obtain a final collision waveform prediction model, so as to use the final collision waveform prediction model, the waveform mean and the principal component waveform to predict the collision waveform of the target vehicle model under the target collision condition.

6. The device according to claim 5, characterized in that The processing module comprises: A first construction unit, configured to construct a collision signal matrix using all collision waveform signals in the sample set; A decomposition unit, used for decomposing the collision signal matrix to obtain a principal component variance and a principal component vector matrix; A screening unit is used to screen the principal component vector matrix based on the principal component variance to obtain the principal component that meets the preset screening condition.

7. The device according to claim 5, characterized in that The calculation module comprises: A calculation unit, used for calculating the component coefficients of all samples in the sample set on different principal components; The second construction unit is used to construct a model training set based on the component coefficients and the sample set.

8. A collision waveform prediction device, characterized in that: Applied to the model use stage, wherein the device comprises: An acquisition module is used to acquire collision state data of a target vehicle model under expected collision conditions; A prediction module, used for inputting the collision state data into a pre-built final collision waveform prediction model to obtain principal components by waveform prediction, wherein the final collision waveform prediction model is obtained by training a model training set; A reconstruction module is used to perform waveform reconstruction based on the principal component and the waveform mean and principal component waveform calculated by a preset sample set to obtain a waveform reconstruction result, and obtain a predicted collision waveform of the target vehicle model under the expected collision condition according to the waveform reconstruction result, wherein the preset sample set is constructed by multiple groups of collision state data of multiple sample vehicles under different collision conditions and a collision waveform signal corresponding to each group of collision state data.

9. A vehicle, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the collision waveform prediction method according to any one of claims 1 to 3 or 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the collision waveform prediction method according to any one of claims 1-3 or 4.

Images