Vehicle suspension control method and device, computer device and readable storage medium

Abstract

This application relates to a vehicle suspension control method, apparatus, computer device, and readable storage medium. The method includes: combining acquired vehicle dynamics time-series data, vehicle handling time-series data, and environmental perception time-series data during vehicle operation into multi-channel time-series data; determining the adaptive mode number for the time-series data in each channel; performing mode decomposition on the time-series data based on the adaptive mode number to extract frequency domain features; integrating the frequency domain features of the time-series data from multiple channels and inputting them into a trained suspension control parameter prediction model to obtain target time-series features for predicting suspension control parameters; predicting vehicle suspension control parameters based on the target time-series features; and controlling the vehicle's suspension system based on the vehicle suspension control parameters. This method can significantly reduce the risk of vehicle rollover.

Description

Technical Field

[0001] This application relates to the field of vehicle control technology, and in particular to a vehicle suspension control method, device, vehicle, computer equipment, computer-readable storage medium, and computer program product. Background Technology

[0002] With the rapid development of intelligent vehicles, autonomous driving systems, and active safety control technologies, ensuring the operational stability of vehicles in complex dynamic environments has become a research hotspot and a technical challenge. Especially under extreme driving conditions such as high-speed cornering, emergency obstacle avoidance, slippery road surfaces, and asymmetric loads, the vehicle's center of gravity shifts drastically, significantly increasing the risk of rollover and placing higher demands on the dynamic response capabilities of the suspension system.

[0003] However, traditional suspension systems rely solely on mechanical structures and preset damping for passive response, resulting in problems such as response lag and limited adjustment capabilities, making it difficult to effectively suppress the vehicle's tendency to roll or over in transient dynamics.

[0004] Therefore, there is a need to provide a solution that can effectively suppress the risk of vehicle rollover during transient dynamics. Summary of the Invention

[0005] Therefore, it is necessary to provide a vehicle suspension control method, device, vehicle, computer equipment, computer-readable storage medium, and computer program product that can significantly reduce the risk of vehicle rollover, in order to address the aforementioned technical problems.

[0006] In a first aspect, this application provides a vehicle suspension control method, including:

[0007] Acquire vehicle dynamics timing data, vehicle handling timing data, and environmental perception timing data during vehicle operation;

[0008] The vehicle dynamics timing data, the vehicle handling timing data, and the environmental perception timing data are combined into multi-channel timing data.

[0009] For the time-series data under each channel, the adaptive mode number of the time-series data under the channel is determined, and the time-series data is decomposed based on the adaptive mode number to extract the frequency domain features of the time-series data.

[0010] The frequency domain features of time-series data from multiple channels are integrated. Using the integrated multimodal frequency domain features as input, a trained suspension control parameter prediction model is called to extract the time-series features from the multimodal frequency domain features. The extracted time-series features are weighted and fused to obtain the target time-series features for predicting suspension control parameters. Based on the target time-series features, the vehicle suspension control parameters are predicted. The suspension control parameter prediction model is trained on a long short-term memory network containing a self-attention mechanism based on the frequency domain features of historical multimodal time-series data under various road conditions.

[0011] The vehicle's suspension system is controlled based on the vehicle suspension control parameters.

[0012] In one embodiment, for time-series data under each channel, the spectral entropy of the power spectrum of the time-series data of the channel is determined;

[0013] Based on the spectral entropy, the number of adaptive modes of the time-series data of the channel is determined.

[0014] In one embodiment, determining the adaptive mode number of the time-series data of the channel based on the spectral entropy includes:

[0015] Determine the signal change rate of the time-series data of the channel;

[0016] Based on the signal change rate and the preset initial differential factor, an adaptive penalty factor is determined;

[0017] Based on the adaptive penalty factor and the spectral entropy, the adaptive mode number of the time-series data of the channel is determined.

[0018] In one embodiment, the step of performing mode decomposition on the time-series data based on the adaptive mode number to extract the frequency domain features of the time-series data includes:

[0019] Based on the adaptive mode number, the time series data is decomposed into multiple intrinsic mode function components, and the intrinsic mode function components are matched with the adaptive mode number;

[0020] Determine the variance of each intrinsic mode function component, and based on the variance, determine the weight of each intrinsic mode function component;

[0021] Based on the weights of each intrinsic mode function component, the intrinsic mode function components are weighted and fused to obtain the frequency domain characteristics of the time-series data of the channel.

[0022] In one embodiment, the suspension control parameter prediction model includes a bidirectional LSTM layer, a fully connected layer, and a convolutional neural network layer. The extraction of temporal features from the frequency domain features of each modality includes:

[0023] The temporal features of the vehicle dynamics time series data are extracted through the bidirectional LSTM layer;

[0024] The vehicle control timing data is mapped to a high-dimensional space through the fully connected layer to obtain the timing features of the vehicle control timing data.

[0025] The convolutional neural network is used to extract temporal features from environmentally perceptual time-series data.

[0026] In one embodiment, the weighted fusion of the extracted temporal features to obtain target temporal features for predicting suspension control parameters includes:

[0027] The interaction weights between each temporal feature and the suspension control task are determined by using the multi-head self-attention mechanism in the suspension control parameter prediction model.

[0028] Based on the interaction weights, the time-series features are weighted and fused to obtain target time-series features for predicting suspension control parameters.

[0029] In one embodiment, before acquiring the vehicle dynamics timing data, vehicle handling timing data, and environmental perception timing data during vehicle operation, the method further includes:

[0030] Acquire vehicle body dynamics time-series data, vehicle handling time-series data, and environmental perception time-series data under different road conditions to obtain multimodal time-series data;

[0031] Feature alignment of the multimodal time-series data is performed using a gradient inversion layer;

[0032] Based on the feature-aligned multimodal time-series data, the pre-constructed initial suspension control parameter prediction model is trained using an adversarial training strategy to obtain the trained suspension control parameter prediction model.

[0033] In one embodiment, after obtaining the trained suspension control parameter prediction model, the method further includes:

[0034] By using channel pruning, redundant convolutional kernels in the trained suspension control parameter prediction model are removed to obtain a compressed suspension control parameter prediction model.

[0035] The compressed suspension control parameter prediction model is deployed on the vehicle.

[0036] Secondly, this application also provides a vehicle suspension control device, comprising:

[0037] The data acquisition module is used to acquire vehicle dynamics time-series data, vehicle handling time-series data, and environmental perception time-series data during vehicle operation.

[0038] The data combination module is used to combine the vehicle dynamics time-series data, the vehicle handling time-series data, and the environmental perception time-series data into multi-channel time-series data.

[0039] The feature extraction module is used to determine the adaptive mode number of the time series data under each channel, and perform mode decomposition on the time series data based on the adaptive mode number to extract the frequency domain features of the time series data.

[0040] The suspension control parameter prediction module integrates the frequency domain features of time-series data from multiple channels. Taking the integrated multimodal frequency domain features as input, it calls a trained suspension control parameter prediction model to extract the time-series features from the multimodal frequency domain features. The extracted time-series features are then weighted and fused to obtain target time-series features for predicting suspension control parameters. Based on the target time-series features, the vehicle suspension control parameters are predicted. The suspension control parameter prediction model is trained on a long short-term memory network containing a self-attention mechanism based on the frequency domain features of historical multimodal time-series data under various road conditions.

[0041] A suspension control module is used to control the vehicle's suspension system based on the vehicle's suspension control parameters.

[0042] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in any of the above-described vehicle suspension control method embodiments.

[0043] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in any of the vehicle suspension control method embodiments.

[0044] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps in any of the vehicle suspension control method embodiments.

[0045] Sixthly, this application also provides a vehicle, including a vehicle body, a memory, and a controller, wherein the memory stores a computer program, and the controller executes the computer program to implement the steps in any of the vehicle suspension control method embodiments.

[0046] The aforementioned vehicle suspension control method, device, computer equipment, computer-readable storage medium, computer program product, and vehicle, by acquiring vehicle dynamics time-series data, vehicle handling time-series data, and environmental perception time-series data during vehicle operation, can comprehensively understand the current driving scenario. Subsequently, by determining the adaptive modality number of the multimodal time-series data and performing variational modal decomposition on the time-series data based on the adaptive modality number, the frequency domain features of the time-series data are extracted. This allows for automatic adjustment of the decomposition intensity for different road conditions, extracting more representative key frequency domain features and providing more accurate input for subsequent suspension control parameter prediction. Next, by inputting the extracted frequency domain feature data into the suspension control parameter prediction model, this model, through a long short-term memory network and attention mechanism, can capture long-term dependencies in the input frequency domain feature data and filter out target time-series features more suitable for predicting suspension control parameters. Subsequently, more accurate vehicle suspension control parameters are predicted based on the target time-series features. Controlling the vehicle suspension based on these vehicle suspension control parameters can identify risks in advance and adjust the suspension system in a timely manner, significantly reducing the probability of vehicle rollover. The entire solution achieves comprehensive perception, accurate prediction, and rapid response of vehicle status through multimodal perception of driving scenarios, fine frequency domain feature extraction, deep learning modeling, and active control strategies. In particular, when facing complex working conditions that may lead to rollover, it can identify risks in advance and actively adjust the suspension system, significantly reducing the probability of vehicle rollover. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart illustrating a vehicle suspension control method in one embodiment;

[0049] Figure 2 This is a flowchart illustrating the adaptive mode number step in one embodiment;

[0050] Figure 3 This is a flowchart illustrating the adaptive mode number step in another embodiment;

[0051] Figure 4 This is a flowchart illustrating the frequency domain feature extraction steps in one embodiment;

[0052] Figure 5 This is a flowchart illustrating the training process of a suspension control parameter prediction model in one embodiment;

[0053] Figure 6 This is a flowchart illustrating the training and deployment process of a suspension control parameter prediction model in one embodiment;

[0054] Figure 7 This is a structural block diagram of a vehicle suspension control device in one embodiment;

[0055] Figure 8 This is a structural block diagram of the vehicle suspension control device in another embodiment;

[0056] Figure 9 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0058] It should be noted that the terms "comprising" and "having," and any variations thereof, as used in this application, are intended to cover non-exclusive inclusion. The term "multiple" as used in this application refers to two or more. The term "and / or" as used in this application refers to one of the solutions, or any combination of multiple solutions.

[0059] In one embodiment, such as Figure 1 As shown, a vehicle suspension control method is provided. This embodiment illustrates the application of this method to a vehicle. It is understood that this method can also be applied to a server, and further to a system including both a vehicle and a server, and implemented through interaction between the vehicle and the server. In this embodiment, the method includes the following steps (hereinafter referred to as S):

[0060] S100 acquires vehicle dynamics timing data, vehicle handling timing data, and environmental perception timing data during vehicle operation.

[0061] Vehicle dynamics time-series data refers to data on the vehicle's motion state that changes over time, including but not limited to lateral acceleration, displacement and velocity of each wheel suspension, vehicle pitch angle, and roll angle. Vehicle handling time-series data refers to vehicle handling parameters that change over time, including but not limited to throttle data, braking data, and steering wheel angle data. Vehicle handling time-series data can be used to reflect the driver's operating intentions. Environmental perception time-series data refers to data on the vehicle's external environmental state that changes over time, including but not limited to road surface adhesion coefficient, weather condition coding, and changes in road surface height.

[0062] In practical implementation, during vehicle operation, sensors such as suspension sensors, body accelerometers, and inertial measurement units collect data in real time. The vehicle controller (hereinafter referred to as the controller) receives dynamic data from multiple sensors, including suspension sensors, body accelerometers, and inertial measurement units, and performs a series of data processing steps to obtain vehicle dynamics time-series data. Throttle pedal sensors, brake pedal sensors, and steering angle sensors collect vehicle operating parameters in real time. The controller receives vehicle operating data from multiple sensors to obtain vehicle control time-series data. Wheel speed sensors, radar, temperature sensors, and cameras perceive the vehicle's external environment in real time and send the collected data to the controller. The controller processes the received environmental perception data to obtain environmental perception time-series data. It can be understood that after receiving raw sensor data from multiple sensors, the controller may perform filtering, noise reduction, and normalization processing on the raw sensor data, align data from different sources using timestamps, and store the data as time-series data according to channel classification.

[0063] The S200 combines vehicle dynamics timing data, vehicle handling timing data, and environmental perception timing data into multi-channel timing data.

[0064] Multi-channel time-series data refers to data collected simultaneously from multiple sensors or measurement points at different time points. This data comes from different sensors, and each channel records data on different physical quantities or from different sources; the data from each channel is independent.

[0065] For example, taking vehicle dynamics time-series data including suspension displacement and lateral acceleration, vehicle handling time-series data including throttle, brake, and steering wheel angle, and environmental perception time-series data including road adhesion coefficient and weather condition coding as examples, the above-mentioned vehicle dynamics time-series data, vehicle handling time-series data, and environmental perception time-series data are respectively referred to as , and Suspension displacement refers to the vertical movement of the vehicle's suspension system during driving. Lateral acceleration is the acceleration perpendicular to the vehicle's direction of motion, manifesting as a centrifugal effect when the vehicle is turning. The coefficient of adhesion (COP) refers to the tire's ability to grip different road surfaces. The COP value is primarily determined by factors such as road material, road surface condition, tire structure, tread pattern, and vehicle speed. Weather condition codes are numerical or alphanumeric codes derived from various weather phenomena (such as sunny, rainy, snowy, and windy). Specifically, this can be achieved by vehicles obtaining real-time weather updates, including weather condition codes, via V2X communication technology.

[0066] By combination , and This yields multi-channel input data, i.e., multi-channel timing data. The system has 7 channels, each corresponding to a type of sensor data or a data category. For example, suspension displacement is one channel, and lateral acceleration is another.

[0067] S300 determines the adaptive mode number of the time-series data for each channel, and performs mode decomposition on the time-series data based on the adaptive mode number to extract the frequency domain features of the time-series data.

[0068] The adaptive mode number is used to characterize how many intrinsic mode functions (IMFs) a signal can be decomposed into. In this embodiment, the adaptive mode number refers to the optimal number of modes automatically determined through an adaptive method. Typically, different types of signals have varying degrees of complexity, and therefore require different numbers of modes. Frequency domain features are data used to describe the energy distribution characteristics of a signal at different frequencies, including but not limited to the dominant frequency, energy percentage, and power spectral density.

[0069] In practice, for each channel's time series data, an adaptive method such as energy entropy, mutual information, or K-means clustering can be used to estimate the optimal number of modes K for that channel. Then, the time series data of that channel is decomposed into K modal signals. Next, FFT (Fast Fourier Transform) or wavelet transform is performed on each modal signal to extract frequency domain features. By performing the above processing on the time series data of each channel, the frequency domain feature vector of each channel is obtained.

[0070] The S400 integrates the frequency domain features of time-series data from multiple channels. Using the integrated multimodal frequency domain features as input, it calls the trained suspension control parameter prediction model, extracts the time-series features from the multimodal frequency domain features, constructs target time-series features for predicting suspension control parameters from the extracted time-series features, and predicts the vehicle suspension control parameters based on the target time-series features.

[0071] Among them, the suspension control parameter prediction model is trained on a long short-term memory network containing a self-attention mechanism based on the frequency domain features of historical multimodal time series data under various road conditions.

[0072] Multimodal frequency domain features refer to the frequency domain features obtained by integrating features from multiple channels (such as vehicle dynamics, control input, and environmental perception) and processing these features into frequency domain features that meet the input requirements of the suspension control parameter prediction model. Target time-series features refer to the key features selected by the model for predicting optimal suspension control parameters, which may include energy, dominant frequency, and time-series rate of change in certain specific frequency bands.

[0073] The suspension system includes passive suspension systems, semi-active suspension systems, and active suspension systems. In this embodiment, the predicted suspension control parameters can also be referred to as active suspension control parameters, which are physical quantities used to control the active suspension system, including but not limited to the damper damping coefficient, active air spring stiffness, and active suspension actuator output force. It is understood that in other embodiments, the suspension control parameters of the semi-active suspension system can also be predicted, thereby controlling the semi-active suspension system.

[0074] In practical applications, vehicle suspension control can be modeled as a prediction problem of multimodal temporal data. A prediction model can be designed by combining a bidirectional LSTM (Long Short-Term Memory) network with a Transformer structure, and incorporating an attention mechanism. This allows for the training of a cross-modal deep prediction model that can be used to predict suspension control parameters using multimodal temporal data. Specifically, model training can proceed as follows: First, addressing the temporal dependencies and multimodal fusion requirements in vehicle suspension control parameter prediction, a backbone network for the suspension control parameter prediction model is constructed based on an LSTM-Transformer hybrid structure. This backbone network includes a temporal feature extraction module and a cross-modal attention fusion module. The temporal feature extraction module uses a bidirectional LSTM layer to capture the long-term dependencies of vehicle dynamics parameters. The cross-modal attention fusion module dynamically allocates weights for temporal features from different modalities through a multi-head self-attention mechanism.

[0075] Subsequently, time-series data on vehicle dynamics, vehicle handling, environmental perception, and optimal suspension control parameters were collected under varying road conditions and complexity over historical periods. Then, the collected time-series data on vehicle dynamics, vehicle handling, and environmental perception were labeled with the optimal suspension control parameters to obtain the training dataset. Next, based on the training dataset, a course-based learning strategy was adopted, gradually transitioning the training process from simple urban road scenarios to complex road conditions (such as rain, snow, and icy roads) to ensure the model mastered different driving modes in stages. An adaptive algorithm balancing convergence speed and stability was selected as the optimizer. A composite loss function was used, incorporating a mean squared error (MAE) term and control parameter smoothness constraints to avoid abrupt changes in output commands. The loss function, combining MAE and control smoothness constraints, prevented abrupt changes in the numerical values ​​of suspension control parameters.

[0076]

[0077] in, This represents the actual optimal suspension control parameters. This represents the rate of change of the actual optimal suspension control parameters; These are the suspension control parameters predicted by the model. To predict the rate of change of suspension control parameters, For smoothing coefficients, This represents the total loss value.

[0078] Based on the aforementioned loss function, the difference between the actual optimal suspension control parameters and the predicted suspension control parameters is fitted to obtain the trained suspension control parameter prediction model. Subsequently, the trained suspension control parameter prediction model is deployed in the vehicle control system.

[0079] In practice, the controller integrates frequency domain features from multiple channels to obtain multimodal frequency domain features. ,in, This represents the time-series data of vehicle dynamics (suspension displacement, lateral acceleration). This indicates the timing data of vehicle control (accelerator, brake, steering wheel angle). This represents the environmental perception time-series data (road adhesion coefficient, weather condition coding). Next, the multimodal frequency domain features are input into the trained suspension control parameter prediction model. The model uses a time-series feature extraction module to extract the time-series features from the frequency domain features of each mode, resulting in multiple time-series features. Subsequently, the target time-series features for predicting suspension control parameters are selected from these extracted features. Finally, based on the target time-series features, a multi-task learning architecture is used to predict the suspension control parameters. Specific control quantities... The output layer obtains the predicted optimal suspension control parameters through a fully connected layer and a Tanh activation function.

[0080]

[0081] in, , This represents the weights and biases of the suspension control parameter prediction layer.

[0082] S500 is a suspension system that controls a vehicle based on vehicle suspension control parameters.

[0083] In this embodiment, taking the controlled suspension system as an active suspension system as an example, after the controller predicts the optimal suspension control parameters, it can use control algorithms such as PID algorithm and fuzzy control method to convert the suspension control parameters into control signals for specific actuators (such as PWM signal to control solenoid valve and voltage signal to drive motor). Then, the control signals are sent to the active suspension system to adjust the action of the active suspension system. During this process, the vehicle's status, such as actual damping value and vehicle posture, can be monitored and fed back in real time for closed-loop adjustment to minimize the possibility of vehicle rollover.

[0084] There are no restrictions on how the number of adaptive modes is determined. For example... Figure 2As shown, in some embodiments, for the timing data under each channel, the adaptive modality number of the timing data under that channel is determined, including:

[0085] S320 determines the spectral entropy of the power spectrum of the time-series data for each channel.

[0086] S340, based on spectral entropy, determines the adaptive mode number of the time-series data of a channel.

[0087] The power spectrum describes the energy distribution of a signal at different frequencies. Spectral entropy is an information theory-based metric used to measure the "disorder" or uncertainty of a signal's frequency domain distribution. For example, if a signal's energy is concentrated at a few frequencies, the spectral entropy is low, representing fewer frequency domain components, and the signal can be represented by fewer modes; if the energy is distributed across multiple frequencies, the spectral entropy is high, representing more frequency domain components, and therefore, more modes are needed to accurately represent the signal.

[0088] In practice, for each channel's time-series data (i.e., signal data), FFT or wavelet transform can be performed on the time-series data of that channel to obtain a complex spectrum. Subsequently, based on the complex spectrum, the power spectrum of the time-series data can be calculated. Next, the spectral entropy is calculated based on the power spectrum using the following method. :

[0089]

[0090] in, Indicates the first The power spectrum of each modal data, and .

[0091] Furthermore, after obtaining the spectral entropy, the number of adaptive modes can be determined through a segmented mapping method. For example, different spectral entropy ranges can be pre-defined to correspond to different adaptive mode numbers K. Then, the adaptive mode number K can be determined by finding the spectral entropy interval in which the calculated spectral entropy falls.

[0092] In other embodiments, a regression model for predicting the optimal number of modes can be pre-trained. The optimal number of modes is obtained by inputting spectral entropy and time-series data into the regression model. The regression model can be a random forest or a neural network, etc., and is not limited to any particular model.

[0093] In this embodiment, by calculating the spectral entropy of each channel signal, the optimal number of modes required for mode decomposition can be accurately determined, thereby extracting frequency domain features more effectively and providing high-quality input features for subsequent suspension control parameter prediction.

[0094] like Figure 3As shown, in some exemplary embodiments, S340 includes:

[0095] S342 determines the signal change rate of the timing data of the channel.

[0096] S344 determines the adaptive penalty factor based on the signal change rate and the preset initial differential factor.

[0097] S346 determines the number of adaptive modes of the time-series data of a channel based on an adaptive penalty factor and spectral entropy.

[0098] The rate of change of the signal is used to describe the degree of fluctuation of the signal in the time domain. The initial difference factor is a preset constant used to control the strength of the influence of the rate of change of the signal on the penalty factor. The adaptive penalty factor is a coefficient used to control the decomposition accuracy and the number of modes; in this embodiment, it can be based on... Signal change rate adaptive penalty factor .

[0099] In practice, the time-domain volatility of time-series data can be assessed using the finite difference method to determine whether it is a non-stationary signal. Then, an adaptive penalty factor can be determined by the signal change rate and a preset initialization difference factor. :

[0100]

[0101] in, Indicates calculation The first difference, This is used to initialize the difference factor. That is, the greater the signal fluctuation, the larger the adaptive penalty factor. The larger the value, the more it helps limit the bandwidth of each mode and prevent aliasing.

[0102] Next, based on the penalty factor and spectral entropy Determine the number of adaptive modes K:

[0103]

[0104] in, and All are penalty factors, and It is a constant.

[0105] In this embodiment, instead of relying on fixed parameters, the number of adaptive modes is determined by the signal change rate and spectral entropy. The adaptive factor is dynamically adjusted according to the signal itself to determine a more reasonable number of modes. This can reduce the problems of mode aliasing, over-decomposition or under-decomposition, and has better versatility for data from different road conditions, driving behaviors and sensor types.

[0106] like Figure 4As shown, in some exemplary embodiments, modal decomposition of time-series data is performed based on adaptive modal number to extract frequency domain features of the time-series data, including:

[0107] S420, based on adaptive mode number, decomposes time series data into multiple intrinsic mode function components, and the intrinsic mode function components are matched with the adaptive mode number.

[0108] S440, determine the variance of each intrinsic mode function component, and based on the variance, determine the weight of each intrinsic mode function component.

[0109] S460, based on the weights of each intrinsic mode function component, performs weighted fusion of each intrinsic mode function component to obtain the frequency domain characteristics of the channel's time-series data.

[0110] The intrinsic mode function component, or IMF component, refers to each mode obtained after decomposing time series data using the Variational Mode Decomposition (VMD) algorithm.

[0111] In this embodiment, an improved VMD algorithm based on spectral entropy to determine the adaptive mode number is used to perform variational mode decomposition on the time-series data under each channel, extracting frequency domain features. Specifically, after determining the adaptive mode K of the time-series data under each channel, a variational mode decomposition algorithm can be used to perform variational mode decomposition on the time-series data under each channel, extracting frequency domain features from the data of each channel. After decomposition, the IMF component groups are obtained:

[0112]

[0113] in, For channel data Decomposition One IMF component.

[0114] Next, the variance of each IMF component can be determined to measure its volatility or magnitude. A larger variance indicates a more "active" or "important" IMF component. Then, the weights of each IMF are determined based on the variance. :

[0115]

[0116] in, Indicates the first One IMF component, Indicates the first The weighted values ​​of each IMF component, and , This represents the variance of the IMF components.

[0117] Finally, based on the weights of each IMF... The frequency domain characteristics of the time series data under this channel are obtained by weighted fusion of all IMF components under this channel. Specifically:

[0118] The final frequency domain features obtained by fusion are ,and Using the above method, variational mode decomposition is performed on the time-series data of each channel to obtain the frequency domain characteristics of the time-series data of each channel.

[0119] It is understood that in other embodiments, EMD (Empirical Mode Decomposition) or other mode decomposition methods can also be used to perform mode decomposition on the time-series data under each channel and extract frequency domain features.

[0120] In this embodiment, by performing variational mode decomposition on the time-series data under the channel, the mode aliasing problem can be reduced, and different frequency components can be separated more accurately.

[0121] In some exemplary embodiments, the suspension control parameter prediction model includes a bidirectional LSTM layer, a fully connected layer, and a convolutional neural network layer, extracting temporal features from the frequency domain features of each modality, including:

[0122] Temporal features of vehicle body dynamics time series data are extracted using a bidirectional LSTM layer.

[0123] By mapping vehicle control timing data to a high-dimensional space through a fully connected layer, the timing characteristics of the vehicle control timing data are obtained.

[0124] Temporal features of environmentally perceptual time-series data are proposed using convolutional neural networks.

[0125] In this embodiment, to address the heterogeneity issue of data from different sensors, a modal embedding layer can be used to perform feature mapping on the input multimodal time-series data and extract time-series features. Specifically, the time-series feature module of the suspension control parameter prediction model includes a bidirectional LSTM layer (forward LSTM and backward LSTM), a fully connected layer, and a convolutional neural network layer. The controller can process the vehicle dynamics time-series data... Directly inputting into a bidirectional LSTM layer, the processing procedure of the bidirectional LSTM layer is as follows:

[0126]

[0127] in, and They represent in Similarly, the hidden states of the forward LSTM and backward LSTM at each time point... Indicates in The hidden state of the forward LSTM at each time point. In order to be in The hidden state of the reverse LSTM at each time point. It represents the multimodal frequency domain features, which are the splicing and integration results of the frequency domain features after multi-channel mode decomposition.

[0128] Subsequently, the time-series characteristics of vehicle dynamics are obtained by using the outputs of the forward LSTM and the backward LSTM. :

[0129]

[0130] At the same time, vehicle control data is transmitted through the fully connected layer (FC). Mapping to a high-dimensional space, the features of each time step are mapped to obtain the temporal features of the vehicle control time-series data. Furthermore, the environmental perception time-series data... The data is input into a convolutional neural network, specifically using a one-dimensional convolutional layer (Conv1D) and a max-pooling layer (MaxPool1D) to extract the embedded representation of the environmentally perceptual temporal data. The processing procedure is as follows:

[0131]

[0132] in, Embedded representations of environmental perception features.

[0133] Next, the extracted temporal features can be uniformly fed into the cross-modal attention fusion module in the model. The interaction weights of the temporal features of different modalities are dynamically allocated through the multi-head self-attention mechanism. Finally, based on the interaction weights, the temporal features are weighted and fused to obtain the target temporal features used to predict suspension control parameters.

[0134] Interaction weights are used to characterize the importance of temporal features at each time step in the current task (such as suspension control), and are automatically learned by the attention mechanism. The target temporal features are the most representative temporal features extracted after weighted fusion, and can be directly used for the final suspension control parameter prediction.

[0135] Specifically, the cross-modal attention fusion module employs a multi-head self-attention (MHSA) mechanism to calculate the interaction weights between temporal features and predictions from different modalities:

[0136]

[0137] in, These represent the query matrix, key matrix, and value matrix, respectively. This is the transpose of the key matrix. For feature dimensions.

[0138] Next, based on the interaction weights, important temporal features are retained through a gating mechanism, and the retained temporal features are fused to obtain the final target temporal features. Output via residual connections and layer normalization (LayerNorm):

[0139]

[0140] This includes time-series features extracted from vehicle dynamics time-series data, vehicle handling time-series data, and environmental perception time-series data. It includes eight parallel attention heads. The core purpose of using eight parallel attention heads is to enhance the model's ability to dynamically interact with complex multimodal time-series data by modeling multiple independent feature subspaces. Each attention head can learn different attention patterns, and the final result is a more comprehensive target time-series feature formed by concatenation and fusion.

[0141] Furthermore, a multi-task learning architecture is employed to predict suspension control parameters, and the output layer obtains the optimal suspension control parameters through a fully connected layer and a Tanh activation function.

[0142]

[0143] in, , This represents the weights and biases of the prediction layer for active suspension control parameters.

[0144] In this embodiment, by designing a modal embedding layer to perform feature mapping on multimodal time series data, different sensor data can be mapped to a vector space of the same dimension after passing through their respective embedding layers, which facilitates subsequent fusion and processing. Furthermore, heterogeneous data has higher robustness after embedding and stronger adaptability to different road conditions and driving behaviors. By performing weighted fusion in a unified space, the fused target time series features have stronger expressive and generalization capabilities.

[0145] To enhance the generalization ability of the suspension control parameter prediction model, a multimodal time-series training model under multiple road conditions can be employed. For example... Figure 5 As shown, in some exemplary embodiments, before acquiring vehicle dynamics timing data, vehicle handling timing data, and environmental perception timing data during vehicle operation, the method further includes:

[0146] S102, acquire vehicle body dynamics time-series data, vehicle handling time-series data and environmental perception time-series data under different road conditions to obtain multimodal time-series data.

[0147] S104 performs feature alignment on multimodal time series data through a gradient inversion layer.

[0148] S106. Based on the multimodal time-series data after feature alignment, the pre-constructed initial suspension control parameter prediction model is trained using an adversarial training strategy to obtain the trained suspension control parameter prediction model.

[0149] In this embodiment, road conditions of varying complexity include, but are not limited to, urban roads, highways, mountain roads, rainy roads, and icy roads. Multimodal time-series data can be obtained by pre-collecting vehicle body dynamics time-series data (suspension displacement, lateral acceleration), vehicle handling time-series data (accelerator, brake, steering wheel angle), and environmental perception time-series data (road surface adhesion coefficient, weather condition coding) under different road conditions. Subsequently, these data are aligned according to time sequence to form a multimodal time-series dataset.

[0150] Next, an initial neural network model with a gradient inversion layer is constructed. This model consists of a feature extractor, a road condition classifier, and a suspension control parameter predictor. During training, multimodal time-series data is input into the feature extractor to obtain feature vectors. These feature vectors are then passed to the road condition classifier and the suspension control parameter predictor, respectively, through the gradient inversion layer. The gradient inversion layer transmits features normally during forward propagation, but during backward propagation, it multiplies the gradient calculated by the road condition classifier by a negative coefficient (i.e., inverts the gradient direction) and feeds it back to the feature extractor. In this way, the feature extractor updates its parameters in a direction that confuses the road condition classifier, thereby learning a feature representation robust to road condition changes; that is, the feature distributions under different road conditions are aligned.

[0151] Finally, adversarial training is performed based on the feature-aligned multimodal time-series data. In each iteration, the model simultaneously calculates two losses: suspension control parameter prediction loss (e.g., mean squared error) and road condition classification loss (e.g., cross-entropy). Through gradient inversion, the feature extractor is optimized to simultaneously minimize the suspension control prediction loss and maximize the road condition classification loss (i.e., making the road condition classifier unable to distinguish between different road conditions), while the road condition classifier is optimized to minimize the road condition classification loss. Through this adversarial training strategy, the model ultimately learns a generalization ability that can adapt to different road conditions, thus obtaining the trained suspension control parameter prediction model.

[0152] In this embodiment, road condition-related features are suppressed by gradient inversion layer, and adversarial training forces the feature extractor to ignore road condition-related features and focus on essential features directly related to suspension control, thereby improving the model's generalization ability under unknown road conditions.

[0153] To meet the real-time requirements of onboard electronic control units, the model can be compressed before deployment. For example... Figure 6 As shown, in some exemplary embodiments, after S106, the method further includes:

[0154] S108 removes redundant convolutional kernels from the trained suspension control parameter prediction model through channel pruning to obtain a compressed suspension control parameter prediction model, which is then deployed on the vehicle.

[0155] Channel pruning is a model compression technique that reduces the number of model parameters and computational cost by removing unimportant channels (i.e., output channels of convolutional kernels) from convolutional neural networks. In suspension control parameter prediction models, if the output channels of certain convolutional kernels contribute very little to the final prediction result, they can be removed without significantly affecting model performance.

[0156] In practice, this can be done by evaluating the importance of the output channels of each convolutional layer in the suspension control parameter prediction model. Then, based on the target compression ratio and model performance requirements, the proportion of channels to be pruned in each layer is determined (e.g., removing 10% of the channels). Next, for each convolutional layer, unimportant output channels are removed, and the corresponding input channels in the next layer are also removed. The pruned model parameters W′ satisfy:

[0157]

[0158] in, It is a binary mask. This is the pruning threshold. These are the complete model parameters before pruning; This represents the activation value of the j-th neuron in the i-th layer of the network before pruning. When it is greater than the pruning threshold... When, the corresponding mask The value is 1.

[0159] Since model performance may degrade, fine-tuning on the training data can be performed to restore performance. Furthermore, a quantized deployment approach can be used to convert the trained suspension control parameter prediction model into a format suitable for onboard hardware deployment and integrate it into the vehicle control system.

[0160] In other embodiments, to ensure the model meets the real-time requirements of vehicle control, a closed-loop verification system can be built on a high-performance test platform. By injecting sensor data streams from typical operating conditions such as urban roads, highways, and rainy / snowy weather, the end-to-end inference latency from multimodal time-series data input to vehicle suspension control parameter output is tested, while the frame rate stability of the model under extreme road condition disturbances is evaluated. The verification process simultaneously monitors the resource utilization of the onboard electronic control unit to ensure that the lightweight model maintains millisecond-level active suspension control response capability under pruning and INT8 quantization conditions, providing a fast and reliable guarantee for actual vehicle deployment.

[0161] To provide a clearer explanation of the vehicle suspension control method provided in this application, a specific embodiment is described below, which includes the following:

[0162] S1 acquires vehicle dynamics timing data, vehicle handling timing data, and environmental perception timing data during vehicle operation.

[0163] Specifically, vehicle dynamics time-series data includes suspension displacement and lateral acceleration; vehicle handling time-series data includes throttle, brake, and steering wheel angle; and environmental perception time-series data includes road surface adhesion coefficient and weather condition coding. For example, the aforementioned vehicle dynamics time-series data, vehicle handling time-series data, and environmental perception time-series data will be referred to as... , and .

[0164] S2 combines vehicle dynamics timing data, vehicle handling timing data, and environmental perception timing data into multi-channel timing data.

[0165] By combination , and This yields multi-channel input data, i.e., multi-channel timing data. .

[0166] S3. For the time-series data under each channel, determine the spectral entropy of the power spectrum of the time-series data of the channel. Based on the spectral entropy, determine the adaptive mode number of the time-series data under the channel, and perform mode decomposition on the time-series data based on the adaptive mode number to extract the frequency domain features of the time-series data.

[0167] Specifically, the signal change rate of the time-series data for each channel is first determined. Based on the signal change rate and a preset initial differential factor, an adaptive penalty factor is determined. Based on the adaptive penalty factor and spectral entropy, the adaptive mode number K of the time-series data for each channel is determined. Based on the adaptive mode number K, the time-series data for each channel is decomposed using VMD to obtain K IMF components. Based on the variance of each IMF component, the weights of the IMF components are determined. The multiple IMF components are then weighted and fused to obtain the frequency domain characteristics of the time-series data for each channel.

[0168] S4 integrates the frequency domain features of time-series data from multiple channels. Using the integrated multimodal frequency domain features as input, the trained suspension control parameter prediction model is called to extract the time-series features from the multimodal frequency domain features. The extracted time-series features are weighted and fused to obtain the target time-series features for predicting suspension control parameters. Based on the target time-series features, the vehicle suspension control parameters are predicted. The suspension control parameter prediction model is trained on a long short-term memory network containing a self-attention mechanism based on the frequency domain features of historical multimodal time-series data under various road conditions.

[0169] Next, when the controller integrates the frequency domain features from multiple channels, it obtains the multimodal frequency domain features. ,in, This represents the time-series data of vehicle dynamics (suspension displacement, lateral acceleration). This indicates the timing data of vehicle control (accelerator, brake, steering wheel angle). This represents time-series data of environmental perception (road adhesion coefficient, weather condition coding). Next, the multimodal frequency domain features are... The data is input into a trained suspension control parameter prediction model; specifically, the controller uses vehicle dynamics time-series data. By directly inputting the data into a bidirectional LSTM layer, the temporal characteristics of the vehicle's dynamics are obtained. The vehicle handling data is then processed through a fully connected (FC) layer. Mapping to a high-dimensional space, the features of each time step are mapped to obtain the temporal features of the vehicle control time-series data. Furthermore, the environmental perception time-series data... The data is input into a convolutional neural network, specifically using one-dimensional convolutional layers (Conv1D) and max-pooling layers (MaxPool1D) to extract embedded representations of the environmental perception temporal data. Next, the extracted temporal features are uniformly fed into the cross-modal attention fusion module of the model. A multi-head self-attention mechanism dynamically assigns interaction weights to the temporal features from different modalities. Finally, based on the interaction weights, the temporal features are weighted and fused to obtain the target temporal features used to predict suspension control parameters. The suspension control parameters of the active suspension system are then predicted based on these target temporal features.

[0170] S5 is a vehicle suspension system that controls the vehicle's suspension based on vehicle suspension control parameters.

[0171] After predicting the optimal suspension control parameters, the controller can use control algorithms such as PID algorithm and fuzzy control method to convert the suspension control parameters into control signals for specific actuators (such as PWM signal to control solenoid valves and voltage signal to drive motors). The control signals are then sent to the active suspension system to adjust the action of the active suspension system. During this process, the vehicle's status, such as actual damping value and vehicle posture, can be monitored and fed back in real time for closed-loop adjustment to minimize the possibility of vehicle rollover.

[0172] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0173] Based on the same inventive concept, this application also provides a vehicle suspension control device for implementing the vehicle suspension control method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more vehicle suspension control device embodiments provided below can be found in the limitations of the vehicle suspension control method described above, and will not be repeated here.

[0174] In one exemplary embodiment, such as Figure 7 As shown, a vehicle suspension control device 700 is provided, including: a data acquisition module 710, a data combination module 720, a feature extraction module 730, a suspension control parameter prediction module 740, and a suspension control module 750, wherein:

[0175] The data acquisition module 710 is used to acquire vehicle dynamics time-series data, vehicle handling time-series data and environmental perception time-series data during vehicle operation.

[0176] The data combination module 720 is used to combine the vehicle dynamics timing data, the vehicle handling timing data, and the environmental perception timing data into multi-channel timing data.

[0177] The feature extraction module 730 is used to determine the adaptive mode number of the time series data under each channel, and perform mode decomposition on the time series data based on the adaptive mode number to extract the frequency domain features of the time series data.

[0178] The suspension control parameter prediction module 740 is used to integrate the frequency domain features of time-series data from multiple channels. Taking the integrated multimodal frequency domain features as input, it calls the trained suspension control parameter prediction model, extracts the time-series features from the multimodal frequency domain features, and performs weighted fusion on the extracted time-series features to obtain the target time-series features for predicting suspension control parameters. Based on the target time-series features, the vehicle suspension control parameters are predicted. The suspension control parameter prediction model is trained on a long short-term memory network containing a self-attention mechanism based on the frequency domain features of historical multimodal time-series data under various road conditions.

[0179] The suspension control module 750 is used to control the vehicle's suspension system based on the vehicle suspension control parameters.

[0180] In some embodiments, the feature extraction module 730 is further configured to determine the spectral entropy of the power spectrum of the time-series data of each channel for each channel, and determine the adaptive mode number of the time-series data of the channel based on the spectral entropy.

[0181] In some embodiments, the feature extraction module 730 is further configured to determine the signal change rate of the time-series data of the channel, determine an adaptive penalty factor based on the signal change rate and a preset initialization difference factor, and determine the adaptive mode number of the time-series data of the channel based on the adaptive penalty factor and the spectral entropy.

[0182] In some embodiments, the feature extraction module 730 is further configured to decompose the time-series data into multiple intrinsic mode function components based on the adaptive mode number, match the intrinsic mode function components with the adaptive mode number, determine the variance of each intrinsic mode function component, determine the weight of each intrinsic mode function component based on the variance, and perform weighted fusion of each intrinsic mode function component based on the weight of each intrinsic mode function component to obtain the frequency domain features of the channel's time-series data.

[0183] In some embodiments, the suspension control parameter prediction model includes a bidirectional LSTM layer, a fully connected layer, and a convolutional neural network layer to extract temporal features from the frequency domain features of each mode. The suspension control parameter prediction module 740 is also used to extract the temporal features of the vehicle dynamics temporal data through the bidirectional LSTM layer, map the vehicle handling temporal data to a high-dimensional space through the fully connected layer to obtain the temporal features of the vehicle handling temporal data, and extract the temporal features of the environmental perception temporal data through the convolutional neural network.

[0184] In some embodiments, the suspension control parameter prediction module 750 is further configured to determine the interaction weights between each temporal feature and the suspension control task through the multi-head self-attention mechanism in the suspension control parameter prediction model, and to perform weighted fusion of the temporal features based on the interaction weights to obtain the target temporal features for predicting the suspension control parameters.

[0185] like Figure 8 As shown, in some embodiments, the device further includes a model training module 702, which is used to acquire vehicle body dynamics time-series data, vehicle handling time-series data and environmental perception time-series data under different road conditions to obtain multimodal time-series data. The multimodal time-series data is then feature-aligned through a gradient inversion layer. Based on the feature-aligned multimodal time-series data, the pre-constructed initial suspension control parameter prediction model is trained through an adversarial training strategy to obtain the trained suspension control parameter prediction model.

[0186] In some embodiments, the apparatus further includes a model deployment module 704, which removes redundant convolutional kernels from the trained suspension control parameter prediction model by channel pruning to obtain a compressed suspension control parameter prediction model, and deploys the compressed suspension control parameter prediction model on the vehicle.

[0187] The modules in the aforementioned vehicle suspension control device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0188] In one exemplary embodiment, this application also provides a vehicle, including a vehicle body, a memory, and a controller. The memory stores a computer program, and the controller executes the computer program to implement the steps in any of the above-described vehicle suspension control method embodiments.

[0189] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores vehicle dynamics timing data, vehicle handling timing data, environmental perception timing data, and suspension control parameters. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a vehicle suspension control method.

[0190] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0191] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in any of the above embodiments of the vehicle suspension control method.

[0192] In one exemplary embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in any of the vehicle suspension control method embodiments described above.

[0193] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in any of the vehicle suspension control method embodiments described above.

[0194] It should be noted that the user information (including but not limited to user device information, vehicle information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0195] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0196] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0197] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A vehicle suspension control method, characterized in that, The method includes: Acquire vehicle dynamics timing data, vehicle handling timing data, and environmental perception timing data during vehicle operation; The vehicle dynamics timing data, the vehicle handling timing data, and the environmental perception timing data are combined into multi-channel timing data. For the time-series data under each channel, the adaptive mode number of the time-series data under the channel is determined, and the time-series data is decomposed based on the adaptive mode number to extract the frequency domain features of the time-series data. The frequency domain features of time-series data from multiple channels are integrated. Using the integrated multimodal frequency domain features as input, a trained suspension control parameter prediction model is called to extract the time-series features from the multimodal frequency domain features. The extracted time-series features are weighted and fused to obtain the target time-series features for predicting suspension control parameters. Based on the target time-series features, the vehicle suspension control parameters are predicted. The suspension control parameter prediction model is trained on a long short-term memory network containing a self-attention mechanism based on the frequency domain features of historical multimodal time-series data under various road conditions. The vehicle's suspension system is controlled based on the vehicle suspension control parameters.

2. The method according to claim 1, characterized in that, The determination of the adaptive modal number of the time-series data for each channel includes: For each channel's time-series data, determine the spectral entropy of the power spectrum of the time-series data for that channel; Based on the spectral entropy, the number of adaptive modes of the time-series data of the channel is determined.

3. The method according to claim 2, characterized in that, The step of determining the adaptive mode number of the time-series data of the channel based on the spectral entropy includes: Determine the signal change rate of the time-series data of the channel; Based on the signal change rate and the preset initial differential factor, an adaptive penalty factor is determined; Based on the adaptive penalty factor and the spectral entropy, the adaptive mode number of the time-series data of the channel is determined.

4. The method according to claim 1, characterized in that, The step of performing mode decomposition on the time-series data based on the adaptive mode number to extract the frequency domain features of the time-series data includes: Based on the adaptive mode number, the time series data is decomposed into multiple intrinsic mode function components, and the intrinsic mode function components are matched with the adaptive mode number; Determine the variance of each intrinsic mode function component, and based on the variance, determine the weight of each intrinsic mode function component; Based on the weights of each intrinsic mode function component, the intrinsic mode function components are weighted and fused to obtain the frequency domain characteristics of the time-series data of the channel.

5. The method according to any one of claims 1 to 4, characterized in that, The suspension control parameter prediction model includes a bidirectional LSTM layer, a fully connected layer, and a convolutional neural network layer. The extraction of temporal features from the frequency domain features of each modality includes: The temporal features of the vehicle dynamics time series data are extracted through the bidirectional LSTM layer; The vehicle control timing data is mapped to a high-dimensional space through the fully connected layer to obtain the timing features of the vehicle control timing data. The convolutional neural network is used to extract temporal features from environmentally perceptual time-series data.

6. The method according to any one of claims 1 to 4, characterized in that, The weighted fusion of the extracted temporal features to obtain target temporal features for predicting suspension control parameters includes: The interaction weights between each temporal feature and the suspension control task are determined by the multi-head self-attention mechanism in the suspension control parameter prediction model. Based on the interaction weights, the time-series features are weighted and fused to obtain target time-series features for predicting suspension control parameters.

7. The method according to any one of claims 1 to 4, characterized in that, Before acquiring the vehicle dynamics timing data, vehicle handling timing data, and environmental perception timing data during vehicle operation, the method further includes: Acquire vehicle body dynamics time-series data, vehicle handling time-series data, and environmental perception time-series data under different road conditions to obtain multimodal time-series data; Feature alignment of the multimodal time-series data is performed using a gradient inversion layer; Based on the feature-aligned multimodal time-series data, the pre-constructed initial suspension control parameter prediction model is trained using an adversarial training strategy to obtain the trained suspension control parameter prediction model.

8. The method according to claim 7, characterized in that, After obtaining the trained suspension control parameter prediction model, the method further includes: By using channel pruning, redundant convolutional kernels in the trained suspension control parameter prediction model are removed to obtain a compressed suspension control parameter prediction model. The compressed suspension control parameter prediction model is deployed on the vehicle.

9. A vehicle suspension control device, characterized in that, The device includes: The data acquisition module is used to acquire vehicle dynamics time-series data, vehicle handling time-series data, and environmental perception time-series data during vehicle operation. The data combination module is used to combine the vehicle dynamics time-series data, the vehicle handling time-series data, and the environmental perception time-series data into multi-channel time-series data. The feature extraction module is used to determine the adaptive mode number of the time series data under each channel, and perform mode decomposition on the time series data based on the adaptive mode number to extract the frequency domain features of the time series data. The suspension control parameter prediction module integrates the frequency domain features of time-series data from multiple channels. Taking the integrated multimodal frequency domain features as input, it calls a trained suspension control parameter prediction model to extract the time-series features from the multimodal frequency domain features. The extracted time-series features are then weighted and fused to obtain target time-series features for predicting suspension control parameters. Based on the target time-series features, the vehicle suspension control parameters are predicted. The suspension control parameter prediction model is trained on a long short-term memory network containing a self-attention mechanism based on the frequency domain features of historical multimodal time-series data under various road conditions. A suspension control module is used to control the vehicle's suspension system based on the vehicle's suspension control parameters.

10. A vehicle, comprising a vehicle body, a memory, and a controller, characterized in that, The memory stores a computer program, characterized in that the controller executes the computer program to implement the steps of the method according to any one of claims 1 to 8.

11. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

Images