Intelligent Control System and Method for Air Suspension Shock Absorber of New Energy Vehicles

Abstract

The present application provides an intelligent control system and method for an air suspension shock absorber of a new energy vehicle, which relates to the field of intelligent control of new energy vehicles. It uses data processing technology based on deep learning to integrate data and encode time-series features for a set of vehicle state data. Then, it aggregates the time-series features of each vehicle state data and correlates the parameters. Based on the global-local semantic mapping enhanced representation of the time-series correlation encoding features between the aggregated and correlated vehicle state parameters, it intelligently predicts the pressure value of the air spring and adaptively adjusts the pressure value of the air spring by comparing it with the real-time parameters. In this way, real-time dynamic adjustment of the air spring pressure can be achieved, which helps to improve the driving experience and driving safety.

Description

Technical Field

[0001] This application relates to the field of intelligent control of new energy vehicles, and more specifically, to an intelligent control system and method for an air suspension shock absorber of a new energy vehicle. Background Art

[0002] In the current situation where the automotive industry is developing rapidly towards intelligence and electrification, new energy vehicles have become the focus of the market. As one of the core components, the air suspension shock absorber can adjust the vehicle height and suspension stiffness by adjusting the pressure of the air spring to adapt to various road conditions and load conditions.

[0003] However, traditional air suspension systems mainly rely on fixed parameter settings or simple feedback mechanisms. On the one hand, fixed parameter settings cannot adapt to changing road conditions and load conditions, resulting in poor suspension performance; on the other hand, simple feedback mechanisms respond slowly and are difficult to adjust the pressure of the air spring in a timely manner when facing complex driving behaviors such as sudden acceleration and braking, as well as sudden road conditions such as potholes and speed bumps. This limitation severely restricts the dynamic adjustment ability and overall performance improvement of the system, affecting the optimization of the driving experience and driving safety.

[0004] Therefore, an intelligent control solution for the air suspension shock absorber of a new energy vehicle is desired. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide an intelligent control system and method for an air suspension shock absorber of a new energy vehicle.

[0006] According to one aspect of this application, an intelligent control system for an air suspension shock absorber of a new energy vehicle is provided, which includes:

[0007] A vehicle state data acquisition module, configured to obtain vehicle state data collected by a sensor assembly deployed on a new energy vehicle to obtain a set of vehicle state data. The sensor assembly includes an acceleration sensor, a vehicle speed sensor, a height sensor, and a vibration frequency sensor, and the vehicle state data includes a vehicle acceleration value, a vehicle speed value, a vehicle height value, and a vehicle vibration frequency value;

[0008] A vehicle state time-series feature extraction module, configured to perform data regularization and time-series feature extraction on the set of vehicle state data to obtain a vehicle acceleration time-series correlation feature implicit encoding vector, a vehicle speed time-series correlation feature implicit encoding vector, a vehicle height time-series change feature implicit encoding vector, and a vehicle vibration frequency time-series correlation feature implicit encoding vector;

[0009] A vehicle state parameter association module, which is used to perform multi-dimensional time series association on the vehicle acceleration time series association feature implicit coding vector, the vehicle speed time series association feature implicit coding vector, the vehicle height time series change feature implicit coding vector, and the vehicle vibration frequency time series association feature implicit coding vector to obtain a time series association coding feature map between vehicle state parameters;

[0010] A vehicle state parameter time series association enhancement module, which is used to perform global-local vehicle state semantic mapping enhancement on the time series association coding feature map between vehicle state parameters to obtain an enhanced time series association coding feature map between vehicle state parameters;

[0011] A pressure value adjustment module, which is used to adjust the pressure value of the air spring based on the enhanced time series association coding feature map between vehicle state parameters.

[0012] According to another aspect of the present application, there is provided an intelligent control method for an air suspension shock absorber of a new energy vehicle, which includes:

[0013] Obtain the vehicle state data collected by the sensor assembly deployed on the new energy vehicle to obtain a set of vehicle state data. The sensor assembly includes an acceleration sensor, a vehicle speed sensor, a height sensor, and a vibration frequency sensor. The vehicle state data includes a vehicle acceleration value, a vehicle speed value, a vehicle height value, and a vehicle vibration frequency value;

[0014] Perform data regularization and time series feature extraction on the set of vehicle state data to obtain a vehicle acceleration time series association feature implicit coding vector, a vehicle speed time series association feature implicit coding vector, a vehicle height time series change feature implicit coding vector, and a vehicle vibration frequency time series association feature implicit coding vector;

[0015] Perform multi-dimensional time series association on the vehicle acceleration time series association feature implicit coding vector, the vehicle speed time series association feature implicit coding vector, the vehicle height time series change feature implicit coding vector, and the vehicle vibration frequency time series association feature implicit coding vector to obtain a time series association coding feature map between vehicle state parameters;

[0016] Perform global-local vehicle state semantic mapping enhancement on the time series association coding feature map between vehicle state parameters to obtain an enhanced time series association coding feature map between vehicle state parameters;

[0017] Based on the enhanced time series association coding feature map between vehicle state parameters, adjust the pressure value of the air spring.

[0018] Compared with the prior art, the intelligent control system and method for the air suspension shock absorber of a new energy vehicle provided by this application adopt a data processing technology based on deep learning to integrate the vehicle state data set and encode the time series features. Then, aggregate the time series features of each vehicle state data and correlate the parameters. Based on the global-local semantic mapping enhanced representation of the time series correlation encoded features between the aggregated and correlated vehicle state parameters, the pressure value of the air spring is intelligently predicted, and the comparison between it and the real-time parameters is used to adaptively adjust the pressure value of the air spring. In this way, real-time dynamic adjustment of the air spring pressure can be achieved, which helps to improve the driving experience and driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. In the drawings:

[0020] Figure 1 It is a system block diagram of the intelligent control system for the air suspension shock absorber of a new energy vehicle according to an embodiment of this application.

[0021] Figure 2 It is a block diagram of the vehicle state time series feature extraction module in the intelligent control system for the air suspension shock absorber of a new energy vehicle according to an embodiment of this application.

[0022] Figure 3 It is a block diagram of the vehicle state parameter correlation module in the intelligent control system for the air suspension shock absorber of a new energy vehicle according to an embodiment of this application.

[0023] Figure 4 It is a block diagram of the vehicle state parameter time series correlation enhancement module in the intelligent control system for the air suspension shock absorber of a new energy vehicle according to an embodiment of this application.

[0024] Figure 5 It is a flowchart of the intelligent control method for the air suspension shock absorber of a new energy vehicle according to an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] Next, exemplary embodiments according to this application will be described in detail with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments of this application. It should be understood that this application is not limited by the exemplary embodiments described here.

[0026] In the wave of the automotive industry's accelerating development towards intelligence and electrification, new energy vehicles have become the focus of market attention. As one of the key components, the air suspension shock absorber can not only adjust the vehicle height but also change the suspension stiffness by regulating the pressure of the air spring, thus adapting to diverse road conditions and load requirements.

[0027] However, traditional air suspension systems usually rely on fixed parameter settings or relatively simple feedback adjustment mechanisms. On the one hand, it is difficult for fixed parameter settings to flexibly cope with complex and variable road conditions and load conditions, resulting in the suspension performance being difficult to reach the ideal state; on the other hand, simple feedback adjustment has the problem of slow response speed, especially during violent driving behaviors such as rapid acceleration and sudden braking, as well as when dealing with sudden situations such as potholed roads and speed bumps, it is difficult to adjust the pressure of the air spring in a timely manner. These deficiencies significantly limit the system's dynamic response ability and overall performance improvement, not only hindering the optimization of ride comfort but also having a certain impact on driving safety.

[0028] To address the above technical problems, the technical concept of this application is to obtain vehicle state data collected by a sensor component deployed in a new energy vehicle to obtain a set of vehicle state data (vehicle acceleration value, vehicle speed value, vehicle height value, and vehicle vibration frequency value), and use deep learning-based data analysis and feature analysis technologies to perform data integration and time-series feature encoding on the set of vehicle state data. Then, aggregate and correlate the parameters of each vehicle state data time-series feature, and based on the global-local semantic mapping enhanced representation of the time-series correlation encoding features between the aggregated and correlated vehicle state parameters, intelligently predict the pressure value of the air spring, and compare it with the real-time parameters to adaptively adjust the pressure value of the air spring. Through the real-time data acquisition and processing capabilities of this application, the system can quickly respond to complex road condition changes and driving behaviors, realize real-time dynamic adjustment of the air spring pressure, thus significantly improving ride comfort, reducing the sense of bumpiness, providing a smoother driving experience, and enhancing driving safety.

[0029] Figure 1 The system block diagram of the intelligent control system for the air suspension shock absorber of a new energy vehicle according to an embodiment of the present application. As Figure 1As shown, in the intelligent control system 100 of the air suspension shock absorber of a new energy vehicle, it includes: a vehicle state data acquisition module 110, which is used to obtain the vehicle state data collected by the sensor assembly deployed on the new energy vehicle to obtain a set of vehicle state data. The sensor assembly includes an acceleration sensor, a vehicle speed sensor, a height sensor, and a vibration frequency sensor. The vehicle state data includes a vehicle acceleration value, a vehicle speed value, a vehicle height value, and a vehicle vibration frequency value; a vehicle state time series feature extraction module 120, which is used to perform data regularization and time series feature extraction on the set of vehicle state data to obtain a vehicle acceleration time series correlation feature implicit coding vector, a vehicle speed time series correlation feature implicit coding vector, a vehicle height time series change feature implicit coding vector, and a vehicle vibration frequency time series correlation feature implicit coding vector; a vehicle state parameter correlation module 130, which is used to perform multi-dimensional time series correlation on the vehicle acceleration time series correlation feature implicit coding vector, the vehicle speed time series correlation feature implicit coding vector, the vehicle height time series change feature implicit coding vector, and the vehicle vibration frequency time series correlation feature implicit coding vector to obtain a time series correlation coding feature map between vehicle state parameters; a vehicle state parameter time series correlation enhancement module 140, which is used to perform global-local vehicle state semantic mapping enhancement on the time series correlation coding feature map between vehicle state parameters to obtain an enhanced time series correlation coding feature map between vehicle state parameters; a pressure value adjustment module 150, which is used to adjust the pressure value of the air spring based on the enhanced time series correlation coding feature map between vehicle state parameters.

[0030] In the embodiment of the present application, the vehicle state data acquisition module 110 is configured to obtain the vehicle state data collected by the sensor assembly deployed on the new energy vehicle to obtain a set of vehicle state data. The sensor assembly includes an acceleration sensor, a vehicle speed sensor, an altitude sensor, and a vibration frequency sensor. The vehicle state data includes a vehicle acceleration value, a vehicle speed value, a vehicle altitude value, and a vehicle vibration frequency value. It should be understood that the acceleration value can reflect the dynamic behaviors of the vehicle such as acceleration, deceleration, sudden acceleration, and sudden braking. During sudden acceleration, the center of gravity of the vehicle moves backward, and the air suspension system needs to adjust the air spring pressure to maintain the balance and stability of the vehicle body and prevent the front of the vehicle from rising excessively. During sudden braking, the center of gravity of the vehicle moves forward, and appropriate air spring pressure can prevent the front of the vehicle from sinking excessively and ensure the handling performance and safety of the vehicle. Different vehicle speeds also affect the requirements for the air suspension system. When driving at high speed, the vehicle needs a stiffer suspension to improve stability and reduce the swaying and rolling of the vehicle body. When driving at low speed, a relatively softer suspension can provide better comfort and absorb road bumps. The vehicle altitude value reflects the ground clearance of the vehicle and the posture of the vehicle body. When the vehicle passes through different road conditions, such as raised speed bumps, potholed roads, etc., the vehicle altitude will change. By obtaining the vehicle altitude value, the intelligent control system can determine the current road condition of the vehicle. For example, when the vehicle altitude suddenly drops, it may be due to encountering a pothole, and at this time, the air spring pressure needs to be increased to buffer the impact and protect the vehicle and passengers. When the vehicle altitude rises, it may be because it has passed over a raised object, and the system can appropriately adjust the pressure to restore the normal vehicle body posture. In addition, the vehicle altitude value is also related to the load of the vehicle. When the vehicle is fully loaded, the vehicle altitude will decrease, and the air spring pressure needs to be adjusted accordingly to maintain a proper suspension state. The vibration frequency value can reflect the flatness of the road surface and the working state of the vehicle suspension system. Different road conditions will generate vibrations of different frequencies. For example, a rough road surface will cause the vehicle to generate high-frequency vibrations, while the vibration frequency of a flat road surface is relatively low. By obtaining the vehicle vibration frequency value, the intelligent control system can understand the road surface conditions. For a road surface with high-frequency vibrations, increasing the air spring pressure can increase the stiffness of the suspension and reduce vehicle vibrations. For a road surface with low-frequency vibrations, the pressure can be appropriately reduced to improve comfort. Generally speaking, the obtained vehicle state data is the basic basis for adjusting the air spring pressure value. By analyzing the obtained vehicle state data, the actual operating state of the vehicle can be understood, and based on this, the corresponding adjustment of the air spring pressure value can be made to adapt to changing road conditions and driving conditions, ensuring the best driving experience and driving safety.

[0031] The following is a detailed elaboration of a specific implementation process of "obtaining the vehicle state data collected by the sensor assembly deployed on the new energy vehicle to obtain a set of vehicle state data":

[0032] First, it is the selection and installation of sensors. To accurately measure the acceleration changes of a vehicle during driving, a three-axis acceleration sensor such as ADXL345 with high precision and fast response speed is often selected. It can precisely sense the acceleration of the vehicle in the X, Y, and Z directions. The range selection needs to be based on the actual usage scenario of the vehicle, generally determined among ±2g, ±4g, ±8g, or ±16g to adapt to the acceleration changes under different driving conditions such as hard acceleration and hard braking. During installation, it is firmly placed near the center of gravity of the vehicle chassis and fixed by bolts or special glue. It is necessary to ensure the correct installation direction so that its coordinate axes are strictly consistent with the front-back, left-right, and up-down directions of the vehicle to avoid measurement errors caused by improper installation. The selection of the vehicle speed sensor cannot be ignored either. The magnetoelectric vehicle speed sensor has a lower cost and is suitable for ordinary vehicle models; while the Hall vehicle speed sensor has higher accuracy and stronger anti-interference ability and is widely used in new energy vehicles with strict requirements for vehicle speed measurement. When installing the vehicle speed sensor, common installation positions are the output shaft of the vehicle transmission or the wheel hub. If installed on the output shaft of the transmission, it is necessary to ensure that the gear of the sensor meshes correctly with the gear ring on the output shaft; if installed on the wheel hub, it is necessary to ensure that the gap between the sensor and the hub is within the specified range of 0.5 - 2mm to ensure the accuracy of the measurement. For the selection of the height sensor, a potentiometer type or an optoelectronic type can be considered. The potentiometer type height sensor has a simple structure and low cost, while the optoelectronic height sensor has high precision and good reliability. For example, the optoelectronic height sensor of Omron can provide accurate height measurement data. The height sensor is generally installed in the vehicle suspension system, such as near the shock absorber or control arm. During the installation process, special attention should be paid to its position and angle. Only in this way can the stroke change of the suspension be accurately measured and the height information of the vehicle be accurately reflected. The piezoelectric vibration sensor is mostly selected for the vibration frequency sensor, which has the advantages of high sensitivity and fast response speed and can accurately capture the vibration frequency of the vehicle. During installation, it is tightly installed in key parts of the vehicle, such as the engine, chassis, or body, to ensure full contact between the sensor and the measured part to effectively obtain vibration signals.

[0033] After the sensors are installed, the data acquisition phase begins. The signals output by the acceleration sensors are of two types: analog and digital. Analog signals need to be converted into digital signals with the help of an analog-to-digital converter (ADC) before subsequent processing can be carried out; digital signals can be directly transmitted to the data acquisition module. The vehicle speed sensor outputs a pulse signal, and the vehicle speed can be calculated by measuring the frequency or period of the pulse. The signal output by the height sensor may be a voltage signal or a digital signal, and corresponding processing is required according to the sensor type. The vibration frequency sensor outputs an electrical signal related to the vibration frequency, which also needs to be appropriately processed and converted. The data acquisition work is completed by a microcontroller (such as Arduino, STM32, etc.), which is responsible for receiving the signals from each sensor and performing preliminary processing and storage. During this process, the sampling frequency needs to be reasonably set. For acceleration and vibration frequency data, the sampling frequency is generally set between 100 Hz and 1000 Hz, which can effectively capture their rapid changes; for vehicle speed and height data, the sampling frequency is relatively low, and a setting of 10 Hz to 50 Hz is sufficient.

[0034] Then, the collected data needs to be transmitted and stored. Data transmission can be carried out in wired or wireless ways. Wired transmission often uses the CAN bus. Due to its high reliability and strong anti-interference ability, the CAN bus has become a common choice for in-vehicle data transmission; wireless transmission can utilize technologies such as Bluetooth, Wi-Fi, 4G / 5G, etc. to achieve remote real-time data transmission. To ensure data security and integrity, data needs to be encoded and encrypted during transmission. After the data is transmitted to the vehicle's central processor or a remote server, it is stored in a database. The choice of database should be based on actual needs and data characteristics. Relational databases (such as MySQL, Oracle, etc.) and non-relational databases (such as MongoDB, Redis, etc.) each have their own advantages. Reasonable organization and management of the stored data contribute to subsequent data processing and analysis.

[0035] Finally, data preprocessing and quality control are carried out. During data preprocessing, filtering methods are often used to remove noise and interference signals. Mean filtering, median filtering, Kalman filtering, etc. are all commonly used methods. For example, using Kalman filtering for acceleration data can significantly reduce measurement errors and improve data accuracy. At the same time, data normalization is performed to unify the data collected by different sensors into the same scale range, which is convenient for subsequent analysis and model training. In terms of quality control, strict quality checks need to be carried out on the collected data to judge the validity of the data. For example, check whether the vehicle speed data is within a reasonable range and whether there are abnormal fluctuations in the acceleration data. For invalid or abnormal data, it needs to be marked, excluded, and corresponding measures should be taken to repair or supplement it to ensure reliable data quality, providing a solid guarantee for the intelligent control results obtained from subsequent data analysis.

[0036] In the embodiment of the present application, the vehicle state time-series feature extraction module 120 is used to perform data regularization and time-series feature extraction on the set of vehicle state data to obtain a vehicle acceleration time-series correlation feature implicit encoding vector, a vehicle speed time-series correlation feature implicit encoding vector, a vehicle height time-series change feature implicit encoding vector, and a vehicle vibration frequency time-series correlation feature implicit encoding vector. Specifically, Figure 2 is a block diagram of the vehicle state time-series feature extraction module in the intelligent control system of the air suspension shock absorber of a new energy vehicle according to the embodiment of the present application. As Figure 2 shown, the vehicle state time-series feature extraction module 120 includes: a vehicle state data splitting and regularization unit 121, which is used to split the set of vehicle state data according to the parameter sample dimension, and regularize the split data according to the time dimension to obtain a time series of vehicle acceleration values, a time series of vehicle speed values, a time series of vehicle height values, and a time series of vehicle vibration frequency values; a vehicle state time-series feature generation unit 122, which is used to perform time-series feature extraction based on one-dimensional convolution on the time series of vehicle acceleration values, the time series of vehicle speed values, the time series of vehicle height values, and the time series of vehicle vibration frequency values respectively to obtain the vehicle acceleration time-series correlation feature implicit encoding vector, the vehicle speed time-series correlation feature implicit encoding vector, the vehicle height time-series change feature implicit encoding vector, and the vehicle vibration frequency time-series correlation feature implicit encoding vector.

[0037] In the embodiment of the present application, the vehicle state data splitting and regularization unit 121 is used to split the set of vehicle state data according to the parameter sample dimension, and regularize the split data according to the time dimension to obtain a time series of vehicle acceleration values, a time series of vehicle speed values, a time series of vehicle height values, and a time series of vehicle vibration frequency values. Correspondingly, considering different vehicle state parameters, such as acceleration, speed, height, and vibration frequency, each of them has unique physical meanings and change laws. In order to be able to separate these different types of parameters independently for observation and analysis to deeply understand the characteristics of each parameter. In the technical solution of the present application, the set of vehicle state data is split according to the parameter sample dimension, and the split data is regularized according to the time dimension to obtain a time series of vehicle acceleration values, a time series of vehicle speed values, a time series of vehicle height values, and a time series of vehicle vibration frequency values. In this way, the characteristics of each parameter in the time dimension can be mined. For example, from the time series of vehicle speed values, it may be found that the driving speed pattern of the vehicle at different times of the day, or the speed change trend during a long-distance driving; from the time series of vehicle vibration frequency, the periodic change of the vibration frequency under specific road conditions can be identified, and these characteristics are very crucial for understanding the driving behavior and working conditions of the vehicle.

[0038] In an embodiment of the present application, the vehicle state time-series feature generation unit 122 is configured to perform time-series feature extraction based on one-dimensional convolution on the time series of the vehicle acceleration value, the time series of the vehicle speed value, the time series of the vehicle height value, and the time series of the vehicle vibration frequency value respectively to obtain the vehicle acceleration time-series correlation feature implicit encoding vector, the vehicle speed time-series correlation feature implicit encoding vector, the vehicle height time-series change feature implicit encoding vector, and the vehicle vibration frequency time-series correlation feature implicit encoding vector. It should be understood that, in order to further capture and refine the time-series feature information of each vehicle state parameter in the time dimension, in the technical solution of the present application, time-series feature extraction based on one-dimensional convolution is performed on the time series of the vehicle acceleration value, the time series of the vehicle speed value, the time series of the vehicle height value, and the time series of the vehicle vibration frequency value respectively to slide the convolution kernel on the time series, automatically extract the local features in the data, and capture the local dependence relationship of the data in the time dimension, such as the change pattern of the acceleration of the vehicle in a short period of time, the instantaneous fluctuation of the speed, etc., to obtain the vehicle acceleration time-series correlation feature implicit encoding vector, the vehicle speed time-series correlation feature implicit encoding vector, the vehicle height time-series change feature implicit encoding vector, and the vehicle vibration frequency time-series correlation feature implicit encoding vector.

[0039] In an embodiment of the present application, the vehicle state parameter association module 130 is configured to perform multi-dimensional time-series association on the vehicle acceleration time-series correlation feature implicit encoding vector, the vehicle speed time-series correlation feature implicit encoding vector, the vehicle height time-series change feature implicit encoding vector, and the vehicle vibration frequency time-series correlation feature implicit encoding vector to obtain a time-series association coding feature map between vehicle state parameters. Specifically, Figure 3 It is a block diagram of a vehicle state parameter association module in an intelligent control system of a new energy vehicle air suspension shock absorber according to an embodiment of the present application. As Figure 3 shown, the vehicle state parameter association module 130 includes: a vehicle state feature multi-dimensional aggregation coding unit 131, configured to perform vehicle state feature multi-dimensional aggregation coding on the vehicle acceleration time-series correlation feature implicit encoding vector, the vehicle speed time-series correlation feature implicit encoding vector, the vehicle height time-series change feature implicit encoding vector, and the vehicle vibration frequency time-series correlation feature implicit encoding vector to obtain a vehicle state multi-dimensional time-series aggregation matrix; a time-series association coding feature generation unit 132 for vehicle state parameters, configured to input the vehicle state multi-dimensional time-series aggregation matrix into an association feature extractor between vehicle state parameters based on a two-dimensional convolutional neural network to obtain the time-series association coding feature map between vehicle state parameters.

[0040] In the embodiment of the present application, the vehicle state feature multi-dimensional aggregation encoding unit 131 is configured to perform vehicle state feature multi-dimensional aggregation encoding on the vehicle acceleration time-series correlation feature implicit encoding vector, the vehicle speed time-series correlation feature implicit encoding vector, the vehicle height time-series change feature implicit encoding vector, and the vehicle vibration frequency time-series correlation feature implicit encoding vector to obtain a vehicle state multi-dimensional time-series aggregation matrix. Accordingly, considering the complex time-series correlation relationships among various vehicle parameters, for example, the change in vehicle speed may affect the vehicle vibration frequency and is also related to the change in vehicle height (such as when passing through an undulating road surface). Therefore, in order to discover the synergistic effects and mutual influence patterns among different parameter features to more accurately reflect the true operating condition of the vehicle, the present application performs vehicle state feature multi-dimensional aggregation encoding on the vehicle acceleration time-series correlation feature implicit encoding vector, the vehicle speed time-series correlation feature implicit encoding vector, the vehicle height time-series change feature implicit encoding vector, and the vehicle vibration frequency time-series correlation feature implicit encoding vector so as to integrate the feature information from different dimensions together, form a more comprehensive and integrated description of the vehicle state, and obtain a vehicle state multi-dimensional time-series aggregation matrix. Specifically, in a specific embodiment of the present application, the vehicle state multi-dimensional time-series aggregation matrix is obtained by arranging the vehicle acceleration time-series correlation feature implicit encoding vector, the vehicle speed time-series correlation feature implicit encoding vector, the vehicle height time-series change feature implicit encoding vector, and the vehicle vibration frequency time-series correlation feature implicit encoding vector in a matrix.

[0041] In the embodiment of the present application, the temporal correlation encoding feature generation unit 132 of vehicle state parameters is configured to input the multi-dimensional temporal aggregation matrix of vehicle states into the correlation feature extractor of vehicle state parameters based on a two-dimensional convolutional neural network to obtain the temporal correlation encoding feature map of vehicle state parameters. It should be understood that in order to explore the mutual correlation and influence relationship of each vehicle parameter in the time dimension, the present application inputs the multi-dimensional temporal aggregation matrix of vehicle states into the correlation feature extractor of vehicle state parameters based on a two-dimensional convolutional neural network to obtain the temporal correlation encoding feature map of vehicle state parameters. Those of ordinary skill in the art should know that a two-dimensional convolutional neural network (2D-CNN) is good at processing data with a two-dimensional spatial structure and can process the row and column information in the matrix simultaneously. By using the multi-dimensional temporal aggregation matrix as the input of the 2D-CNN, the advantages of the 2D-CNN in processing this type of structured data can be fully utilized to explore the potential relationship between rows and columns in the matrix, that is, the complex correlation between different parameters and between different time points of the same parameter. In this way, feature extraction can be performed in the two-dimensional space of the parameter dimension and the time dimension, so as to more comprehensively and deeply explore the correlation features of vehicle state parameters. For example, it can simultaneously capture the correlation pattern between the change in vehicle acceleration within a certain period and the change in vehicle height within the same period, providing an intuitive basis for further understanding the vehicle running state.

[0042] In the embodiment of the present application, the temporal correlation enhancement module 140 of vehicle state parameters is configured to perform global-local vehicle state semantic mapping enhancement on the temporal correlation encoding feature map of vehicle state parameters to obtain an enhanced temporal correlation encoding feature map of vehicle state parameters. Specifically, Figure 4 It is a block diagram of the temporal correlation enhancement module of vehicle state parameters in the intelligent control system of the air suspension shock absorber of a new energy vehicle according to an embodiment of the present application. As Figure 4As shown in the figure, the vehicle state parameter temporal correlation enhancement module 140 includes: a vehicle state feature global key semantic information extraction unit 141, configured to input the temporal correlation encoded feature map between the vehicle state parameters into a global key semantic information extraction network to obtain a global temporal key semantic information encoded vector between the vehicle state parameters; a to-be-processed feature window determination unit 142, configured to extract a channel feature vector between to-be-processed vehicle state parameters from the temporal correlation encoded feature map between the vehicle state parameters, and determine a vehicle state local neighborhood semantic association window of the channel feature vector between the to-be-processed vehicle state parameters; a semantic mapping guidance factor calculation unit 143, configured to calculate a vehicle state local key semantic information encoded vector corresponding to the channel feature vector between the to-be-processed vehicle state parameters, and perform semantic contrast analysis on the vehicle state local key semantic information encoded vector and the global temporal key semantic information encoded vector between the vehicle state parameters to determine a vehicle state semantic mapping guidance factor of the channel feature vector between the to-be-processed vehicle state parameters; and a to-be-processed vehicle state parameter enhanced channel feature generation unit 144, configured to modulate the channel feature vector between the to-be-processed vehicle state parameters based on the vehicle state semantic mapping guidance factor to obtain an enhanced channel feature vector between the to-be-processed vehicle state parameters, where the enhanced channel feature vector between the to-be-processed vehicle state parameters is the channel feature vector at the pixel position (i, j) of the enhanced temporal correlation encoded feature map between the vehicle state parameters.

[0043] It should be understood that the vehicle operating state is a complex system, which includes both overall trends and long-term patterns, as well as local instantaneous changes and detailed features. When traditional convolutional neural networks process the temporal correlation encoded feature map between vehicle state parameters, due to the limited receptive field, it is difficult to effectively model long-distance dependence relationships and accurately extract global semantic background information. Moreover, the analysis of vehicle state parameters requires a high level of fine-grained information expression of local features. However, when convolutional operations capture local semantics of the vehicle state, information dilution is likely to occur, resulting in insufficient expression of some subtle changes (such as minor changes in vehicle vibration frequency, edge features of height changes, etc.). Therefore, the present application introduces a global-local vehicle state semantic mapping enhancement mechanism to enhance the temporal correlation encoded feature map between the vehicle state parameters to obtain an enhanced temporal correlation encoded feature map between the vehicle state parameters. In particular, this mechanism integrates the collaborative modeling means of global semantics and local semantics, fuses the global semantic background information with local feature details, constructs a mapping guidance factor, and uses this factor to perform modulation and strengthening operations on each channel in the feature space, thereby generating features with a higher task correlation degree and stronger expression ability to improve the understanding of the overall vehicle operating state.

[0044] Specifically, first, the temporal correlation encoding feature map between the vehicle state parameters needs to be input into the global key semantic information extraction network to obtain the global temporal key semantic information encoding vector between the vehicle state parameters. The above process can be expressed by the formula: ; where is the temporal correlation encoding feature map between the vehicle state parameters, is the set of real numbers, and are respectively the height and width of each feature matrix along the channel dimension, is the number of channels of is the feature decomposition operation, and are respectively the 1st, 2nd, th, and th temporal correlation encoding feature vectors in the set of temporal correlation encoding feature vectors between the vehicle state parameters after decomposition, and are respectively the corresponding weight matrix and bias vector, is matrix multiplication, is the scoring vector, is the th temporal correlation significant factor in the set of temporal correlation significant factors between the vehicle state parameters, is the normalization function, is the number of vectors in the set of temporal correlation encoding feature vectors between the vehicle state parameters, is the global temporal key semantic information encoding vector between the vehicle state parameters.

[0045] It should be understood that during the vehicle operation state, there are long-distance dependency relationships among vehicle state parameters (such as acceleration, speed, height, and vibration frequency, etc.) at different times. For example, during the vehicle acceleration process, from the start of acceleration to reaching a stable speed, the changes in the acceleration value and the speed value during this process are continuous and mutually related, and this kind of correlation may span a relatively long time series. The global key semantic information extraction network can capture these long-distance dependency relationships from the time-series correlation encoded feature map among vehicle state parameters. It constructs a global semantic template to uniformly encode the entire feature space, enabling the network to pay attention to the potential connections among different times and different vehicle parameters, thereby more accurately understanding the changing trend of the vehicle operation state. Moreover, through its special construction and operations, the global key semantic information extraction network can not only expand the receptive field but also model the potential dependency relationships across pixels and across feature channels. Specifically, in the time-series correlation encoded feature map among vehicle state parameters, there may be complex dependency relationships among different pixels (representing the values of different times or different parameters) and different feature channels (representing different vehicle state parameters). This network can mine these potential dependency relationships. For example, it can discover the hidden connection between the change in the vehicle vibration frequency and the vehicle speed and height, thereby more deeply understanding the essence of the vehicle operation state and providing more powerful support for optimizing the control of the suspension system.

[0046] Specifically, in the embodiment of the present application, the to-be-processed feature window determination unit is configured to: extract the channel feature vector among vehicle state parameters at the pixel position (i, j) from the time-series correlation encoded feature map among vehicle state parameters as the to-be-processed channel feature vector among vehicle state parameters, and this process can be expressed by the formula: ; where is the channel feature vector among vehicle state parameters at the pixel position (i, j) in and

[0047] Based on the feature distribution characteristics of the to-be-processed channel feature vector among vehicle state parameters, determine the local neighborhood semantic association window of the vehicle state for the to-be-processed channel feature vector among vehicle state parameters, where the to-be-processed channel feature vector among vehicle state parameters is located at the center position of the local neighborhood semantic association window of the vehicle state, and this process can be expressed by the formula: ; where is the to-be-processed channel feature vector among vehicle state parameters, is used to calculate the square of the Euclidean norm of is the logarithmic function value with base 2, is rounding up, is the central position value of the channel feature vector between the vehicle state parameters to be processed in the vehicle state local neighborhood semantic association window, is the size of the vehicle state local neighborhood semantic association window.

[0048] It should be understood that the temporal correlation coding feature map between vehicle state parameters contains various information about the vehicle running state and is presented in the form of pixels. By extracting the channel feature vector between vehicle state parameters at the pixel position (i, j) as the channel feature vector between the vehicle state parameters to be processed, it is possible to focus on the local feature details in the vehicle running state, which is conducive to subsequent more refined analysis. It should be noted that in this step, the channel feature vector between the vehicle state parameters to be processed is the channel feature vector at any pixel position in the temporal correlation coding feature map between the vehicle state parameters.

[0049] Correspondingly, considering that in the entire vehicle state analysis process, the previous steps have processed the temporal correlation coding feature map between vehicle state parameters from a global perspective and obtained some global semantic information. However, to understand the vehicle state more precisely, it is necessary to transition from the global to the local. The channel feature vector between the vehicle state parameters to be processed represents the local information of the vehicle state at a specific position. Determining the vehicle state local neighborhood semantic association window by analyzing its feature distribution characteristics is the key operation to achieve this transition from the global to the local. That is, accurately determining the vehicle state local neighborhood semantic association window can enable the model to capture the local semantic association between vehicle state parameters more accurately. Specifically, the vehicle state local neighborhood semantic association window focuses on high-semantic correlation points, enabling the model to understand the vehicle state at a specific position and moment more deeply and avoiding being interfered by irrelevant information. It is worth mentioning that the determination of the vehicle state local neighborhood semantic association window is closely related to the ideas of regional dependence modeling and information reduction. In regions where the feature distribution is relatively concentrated and the semantic correlation is strong, the window range can be appropriately reduced to focus on key information; in regions where the feature distribution is relatively dispersed and more context information needs to be considered, the window range can be expanded. For example, when the vehicle is in a steady driving state, the change of relevant parameters is relatively small, and the receptive field range can be relatively small; while when the vehicle encounters complex road conditions or performs intense driving operations, the parameter change is large, and the receptive field range can be expanded accordingly to better capture the change of the vehicle state.

[0050] Specifically, in the embodiments of the present application, the semantic mapping guiding factor calculation unit includes: a vehicle state local key semantic information extraction subunit, configured to input all the channel feature vectors between vehicle state parameters in the vehicle state local neighborhood semantic association window into a local neighborhood association semantic information extraction network to obtain the vehicle state local key semantic information coding vector corresponding to the channel feature vector between the to-be-processed vehicle state parameters; a vehicle state semantic mapping guiding factor calculation subunit, configured to determine the vehicle state semantic mapping guiding factor of the channel feature vector between the to-be-processed vehicle state parameters based on the semantic contrast analysis between the global time-series key semantic information coding vector between vehicle state parameters and the vehicle state local key semantic information coding vector corresponding to the channel feature vector between the to-be-processed vehicle state parameters.

[0051] More specifically, in the embodiments of the present application, the vehicle state local key semantic information extraction subunit is configured to input all the channel feature vectors between vehicle state parameters in the vehicle state local neighborhood semantic association window into a local neighborhood association semantic information extraction network to obtain the vehicle state local key semantic information coding vector corresponding to the channel feature vector between the to-be-processed vehicle state parameters. The above process can be expressed as: ; where is the central position value of the channel feature vector between the to-be-processed vehicle state parameters in the vehicle state local neighborhood semantic association window, is the channel feature vector between vehicle state parameters at the pixel position (k, l) in the vehicle state local neighborhood semantic association window, is the corresponding vehicle state local key semantic information coding vector.

[0052] It should be understood that the channel feature vectors within the vehicle state local neighborhood semantic association window cover a large amount of fine-grained semantic data, such as the minute changes in various state parameters of the vehicle within a short period of time. The local neighborhood association semantic information extraction network can accurately identify and extract the key semantic information highly relevant to the to-be-processed feature points from them. For example, when the vehicle passes over a speed bump, the network can accurately extract information related to the sharp change in vibration frequency, the instantaneous change in vehicle body height, and the short-term fluctuation of acceleration from numerous feature vectors, which is crucial for judging the specific state of the vehicle at this time. Moreover, in actual data processing, not all information within the association window is substantially helpful for understanding the vehicle state of the to-be-processed feature points, and there are many redundant semantic information. Through the built-in correlation measurement mechanism of the local neighborhood association semantic information extraction network, these information can be screened to exclude those with little relevance to the current to-be-processed feature points, thereby avoiding interference from these redundant information to subsequent analysis.

[0053] More specifically, in the embodiments of the present application, the vehicle state semantic mapping guidance factor calculation subunit is configured to: based on the channel feature vector between the to-be-processed vehicle state parameters, perform eigen-diagonal double-mediated semantic interaction optimization on the vehicle state local key semantic information coding vector corresponding to the channel feature vector between the to-be-processed vehicle state parameters to obtain an optimized vehicle state local key semantic information coding vector of the channel feature vector between the to-be-processed vehicle state parameters. This process can be expressed by the formula: ; where represents matrix multiplication, is a vector and is the correlation matrix of is a matrix is the diagonal vector composed of the diagonal elements of the matrix is a matrix is the eigenvector composed of the eigenvalues of the matrix is corresponding optimized vehicle state local key semantic information coding vector, is element-wise addition by position;

[0054] Calculate the Mahalanobis distance between the global temporal key semantic information coding vector between the vehicle state parameters and the optimized vehicle state local key semantic information coding vector of the channel feature vector between the to-be-processed vehicle state parameters as the vehicle state semantic mapping guidance factor of the channel feature vector between the to-be-processed vehicle state parameters. This process can be expressed by the formula: ; where is corresponding optimized vehicle state local key semantic information coding vector, is the global temporal key semantic information coding vector between vehicle state parameters, is and is the inverse matrix of the covariance matrix between is the transpose operation, is and is the vehicle state semantic mapping guidance factor between

[0055] Here, for the vehicle state local key semantic information encoding vector obtained by local semantic information extraction, based on the concept of "like attracts like", the interaction between entities with similar semantic neighborhood types in the local semantic interaction system, that is, the interaction between the vehicle state local key semantic information encoding vector and the channel feature vector between the vehicle state parameters to be processed, can be further characterized. That is, the vehicle state local key semantic information encoding vectors with associated similar semantic neighborhood windows are interacted based on the semantic similarity drive of the channel feature vector between the vehicle state parameters to be processed, so that the vehicle state local key semantic information encoding vectors and the channel feature vector between the vehicle state parameters to be processed with similar attributes are more inclined to form a feature entity association. Specifically, the diagonal vector and eigenvector of the association representation between the vehicle state local key semantic information encoding vector and the channel feature vector between the vehicle state parameters to be processed are used as the mediation-mediated effect representation and the target-mediated effect representation respectively, to represent the indirect influence of the associated semantic similarity on the target variable through the mediation variable by means of mediation effect modeling, so as to perform local semantic logic mediation inference through the "like attracts like" semantic network structure, and further improve the subsequent semantic mapping alignment.

[0056] It should be understood that by calculating the Mahalanobis distance between the global temporal key semantic information encoding vector between vehicle state parameters and the optimized vehicle state local key semantic information encoding vector between the vehicle state parameters to be processed, the semantic mapping relationship of the channel feature vector to be processed in the global context can be clarified. Specifically, the global temporal key semantic information encoding vector between vehicle state parameters represents the overall trend and long-term pattern of the vehicle state, which is a description of the vehicle state from a macroscopic perspective; while the optimized vehicle state local key semantic information encoding vector focuses on the local detailed information of the feature points to be processed. The calculation of the Mahalanobis distance can measure the difference and correlation degree between the two, so as to determine the position and meaning of the local feature in the global semantic background. This step is similar to the feature alignment idea, where the global semantics provides a reference framework for the local semantics, enabling the local features to be compared and analyzed in the global context. Generally speaking, the semantic contrast operation (implemented by calculating the Mahalanobis distance) not only strengthens the global consistency of the vehicle state feature expression, but also retains the local personalized details. It generates a specific vehicle state semantic mapping guiding factor in the form of semantic mapping, which can provide a core basis for the subsequent final vehicle state feature enhancement operation.

[0057] Finally, based on the vehicle state semantic mapping guidance factor, modulate the channel feature vector between the to-be-processed vehicle state parameters to obtain an enhanced channel feature vector between the to-be-processed vehicle state parameters, where the enhanced channel feature vector between the to-be-processed vehicle state parameters is the channel feature vector at the pixel position (i, j) of the enhanced temporal correlation encoding feature map between the enhanced vehicle state parameters. The above process can be expressed by the formula: ; where is the channel feature vector between the to-be-processed vehicle state parameters, is and the vehicle state semantic mapping guidance factor between them, is the channel feature vector at the pixel position (i, j) in the enhanced temporal correlation encoding feature map between the enhanced vehicle state parameters, that is, the is modulated to obtain the enhanced channel feature vector between the to-be-processed vehicle state parameters.

[0058] It should be understood that the importance of different channel features in the original temporal correlation encoding feature map between vehicle state parameters for describing the vehicle running state is not exactly the same, and under different driving scenarios and task requirements, the priorities of each channel feature will change. The vehicle state semantic mapping guidance factor reflects the degree of association and difference between the channel feature vector between the to-be-processed vehicle state parameters and the global semantics. When modulating the channel feature vector between the to-be-processed vehicle state parameters by using the multiplicative adjustment method based on this factor, the weight of this channel feature can be adjusted according to its degree of fit with the global semantics. This will increase the feature weights of the channels that are more critical for accurately judging the vehicle state and performing suspension adjustment in the original feature map, while the weights of other relatively less important channel features will be correspondingly reduced, so that the model can focus more on important feature information during specific processing. And by adjusting each channel feature vector between the to-be-processed vehicle state parameters in the original temporal correlation encoding feature map between vehicle state parameters based on the vehicle state semantic mapping guidance factor, the semantic information carried by each channel feature can be more accurately expressed and strengthened, so as to further enrich the representation ability of the enhanced temporal correlation encoding feature map composed of it, enabling the enhanced feature map to describe the vehicle running state more comprehensively and meticulously, and thus providing strong support for accurately judging the actual needs of the vehicle and making corresponding control decisions in the future.

[0059] In the embodiment of the present application, the pressure value adjustment module 150 is configured to adjust the pressure value of the air spring based on the enhanced temporal correlation encoded feature map between vehicle state parameters. Specifically, in the embodiment of the present application, the pressure value adjustment module is configured to: input the enhanced temporal correlation encoded feature map between vehicle state parameters into a model predictive controller to obtain suspension optimization parameters, where the suspension optimization parameters include the pressure value of the air spring; send the suspension optimization parameters to the actuator layer, and the actuator layer sends a control instruction to the air spring based on the comparison between the suspension optimization parameters and the real-time suspension parameters to adjust the pressure value of the air spring.

[0060] It should be understood that inputting the enhanced temporal correlation encoded feature map between vehicle state parameters into a model predictive controller to obtain suspension optimization parameters means using the temporal correlation encoded feature map between vehicle state parameters to perform decoding processing on the enhanced temporal correlation encoded feature map between vehicle state parameters, so as to intelligently predict the pressure value of the air spring. In particular, the model predictive controller usually analyzes and calculates the input feature map based on a certain mathematical model and algorithm. The enhanced feature map provides rich semantic information for the model, enabling the model to better learn and understand the relationship between the vehicle state and the suspension parameters. By analyzing these features, the model can predict the suspension optimization parameters most suitable for the current vehicle state according to its internal algorithms and rules. In a specific embodiment of the present application, inputting the enhanced temporal correlation encoded feature map between vehicle state parameters into a model predictive controller to obtain suspension optimization parameters, where the suspension optimization parameters include the pressure value of the air spring, includes: multiplying the decoding weight matrix of the model predictive controller with each feature matrix of the enhanced temporal correlation encoded feature map between vehicle state parameters along the channel dimension to obtain the decoded enhanced temporal correlation encoded feature map between vehicle state parameters, and accumulating and summing the feature values at all positions of the decoded enhanced temporal correlation encoded feature map between vehicle state parameters to obtain the suspension optimization parameters.

[0061] Subsequently, the suspension optimization parameters are sent to the actuator layer, and the actuator layer sends a control instruction to the air spring based on the comparison between the suspension optimization parameters and the real-time suspension parameters to adjust the pressure value of the air spring. It should be understood that the working conditions during vehicle driving are complex and changeable, and the actual real-time suspension parameters will constantly fluctuate. Comparing the suspension optimization parameters with the real-time suspension parameters can monitor the difference between the current state and the ideal state of the system in real time, enabling the actuator layer to accurately judge the deviation between the current air spring pressure and the ideal pressure, and then issue an accurate control instruction to achieve precise adjustment of the air spring pressure value and meet the driving requirements of the vehicle under various complex road conditions.

[0062] The following is a detailed elaboration of a specific implementation process of "sending the suspension optimization parameters to the actuator layer, and the actuator layer sending a control instruction to the air spring based on the comparison between the suspension optimization parameters and the real-time suspension parameters to adjust the pressure value of the air spring":

[0063] First, the suspension optimization parameters obtained through calculation are reliably transmitted to the actuator layer by means of an efficient data transmission network inside the vehicle, such as a Controller Area Network (CAN) bus. The CAN bus has advantages such as high reliability and strong anti-interference ability, which can ensure that data is not lost or incorrect during transmission, enabling the actuator layer to accurately receive the suspension optimization parameters.

[0064] While receiving the suspension optimization parameters, the actuator layer is also continuously acquiring the real-time suspension parameters. A variety of sensors are equipped on the vehicle to monitor these real-time parameters. The pressure sensor is responsible for accurately measuring the current actual pressure of the air spring. It can sense the subtle changes in the internal pressure of the air spring in real time and convert these physical quantities into electrical signals for output. The displacement sensor focuses on monitoring the position information closely related to the suspension state, such as the degree of compression or extension of the suspension. This information helps to comprehensively understand the real-time working condition of the suspension. These sensors collect data at a very high frequency and transmit the collected data to the actuator layer in a timely manner through dedicated data transmission lines.

[0065] The control unit, which is the core component inside the actuator layer, is the key to processing and decision-making. The control unit is usually composed of a powerful microprocessor, and complex and refined control algorithm programs are pre-written inside it. When the control unit receives the suspension optimization parameters and real-time parameters, it immediately starts the comparison operation. Taking the air spring pressure as an example, the control unit will accurately calculate the difference between the current actual pressure of the air spring and the ideal air spring pressure, and at the same time analyze the trend of pressure change to determine whether the pressure is continuously rising, falling, or remaining stable.

[0066] To achieve precise control, the control unit generates control instructions based on preset control logics and algorithm rules. Among them, the PID control algorithm (Proportional-Integral-Derivative control) plays an important role in this process. The PID control algorithm comprehensively considers the current error (the difference between the actual pressure and the ideal pressure), the integral of the error (reflecting the accumulation of errors over a period of time), and the derivative of the error (reflecting the rate of change of the error). By reasonably adjusting and operating these three parameters, the PID algorithm can output a suitable control signal. When the current pressure of the air spring is lower than the ideal air spring pressure value, the PID algorithm will output a control signal to increase the pressure according to the calculation result; conversely, when the pressure is too high, the output control signal will prompt the pressure to decrease.

[0067] After the control instruction is generated, the next step is the execution phase. The actuator layer amplifies the control signal through a drive circuit such as a power amplifier. Since the signal power output by the control unit is often small and cannot directly drive the actuator to work, a power amplifier is needed to amplify the signal to a sufficient intensity. The actuator mainly includes key components such as solenoid valves and air pumps. When the control instruction requires increasing the air spring pressure, the amplified signal by the power amplifier will drive the solenoid valve to open and start the air pump to work simultaneously. The air pump starts to fill the air spring with gas, and as the gas is continuously injected, the pressure of the air spring gradually rises. During this process, the pressure sensor continuously monitors the change of pressure and feeds the new pressure data back to the control unit of the actuator layer in real time. The control unit will continuously adjust the control instruction according to these feedback data to ensure that the air spring pressure can quickly and accurately approach the ideal air spring pressure value. If the pressure rises too fast or exceeds the ideal air spring pressure value, the control unit will timely adjust the control instruction to reduce the working intensity of the air pump or close the solenoid valve to prevent the pressure from being too high.

[0068] Conversely, when it is necessary to reduce the air spring pressure, the control instruction will cause the solenoid valve to open a specific exhaust passage to let the gas in the air spring out. During the exhaust process, the pressure sensor also monitors the pressure change in real time, and the control unit adjusts the opening degree of the solenoid valve in real time according to the feedback data to achieve precise control of the pressure reduction speed and ensure that the pressure drops smoothly to the ideal air spring pressure value.

[0069] During the entire adjustment process, a tight closed-loop control system is formed between the actuator layer and the sensor. The sensor continuously provides real-time data for the actuator layer. The actuator layer analyzes and makes decisions based on these data and ideal parameters, generates control instructions and adjusts the actions of the actuator, so as to achieve dynamic and precise adjustment of the air spring pressure.

[0070] In summary, the intelligent control system 100 of the air suspension shock absorber for new energy vehicles based on the embodiments of the present application is elucidated. It uses data processing technology based on deep learning to integrate and encode the time-series features of the set of vehicle state data. Then, it aggregates the time-series features of each vehicle state data and correlates the parameters. Based on the global-local semantic mapping enhanced representation of the time-series correlation encoding features between the aggregated and correlated vehicle state parameters, it intelligently predicts the pressure value of the air spring and adaptively adjusts the pressure value of the air spring by comparing it with the real-time parameters. In this way, real-time dynamic adjustment of the air spring pressure can be achieved, which helps to improve the driving experience and driving safety.

[0071] Figure 5 It is a flowchart of the intelligent control method for the air suspension shock absorber of new energy vehicles according to the embodiments of the present application. As Figure 5As shown, in the intelligent control method of the air suspension shock absorber of a new energy vehicle, it includes: S110, obtaining the vehicle state data collected by the sensor assembly deployed on the new energy vehicle to obtain a set of vehicle state data. The sensor assembly includes an acceleration sensor, a vehicle speed sensor, a height sensor, and a vibration frequency sensor. The vehicle state data includes a vehicle acceleration value, a vehicle speed value, a vehicle height value, and a vehicle vibration frequency value; S120, performing data regularization and time series feature extraction on the set of vehicle state data to obtain a vehicle acceleration time series correlation feature implicit coding vector, a vehicle speed time series correlation feature implicit coding vector, a vehicle height time series change feature implicit coding vector, and a vehicle vibration frequency time series correlation feature implicit coding vector; S130, performing multi-dimensional time series correlation on the vehicle acceleration time series correlation feature implicit coding vector, the vehicle speed time series correlation feature implicit coding vector, the vehicle height time series change feature implicit coding vector, and the vehicle vibration frequency time series correlation feature implicit coding vector to obtain a time series correlation coding feature map between vehicle state parameters; S140, performing global-local vehicle state semantic mapping enhancement on the time series correlation coding feature map between vehicle state parameters to obtain an enhanced time series correlation coding feature map between vehicle state parameters; S150, based on the enhanced time series correlation coding feature map between vehicle state parameters, adjusting the pressure value of the air spring.

[0072] Here, those skilled in the art can understand that the specific operations of each step in the above intelligent control method of the air suspension shock absorber of a new energy vehicle have been described in detail above with reference to Figures 1 to 4 the description of the intelligent control system of the air suspension shock absorber of the new energy vehicle mentioned above, and therefore, its repeated description will be omitted.

[0073] In summary, based on the intelligent control method of the air suspension shock absorber of the new energy vehicle according to the embodiments of the present application, it is clarified that it uses data processing technology based on deep learning to integrate data and perform time series feature coding on the set of vehicle state data. Then, it aggregates the time series features of each vehicle state data and correlates the parameters. Based on the global-local semantic mapping enhancement representation of the time series correlation coding features between the aggregated and correlated vehicle state parameters, it intelligently predicts the pressure value of the air spring and adaptively adjusts the pressure value of the air spring by comparing it with the real-time parameters. In this way, the real-time dynamic adjustment of the air spring pressure can be achieved, which helps to improve the driving experience and driving safety.

Claims

1. An intelligent control system for an air suspension shock absorber of a new energy vehicle, characterized in that, Comprising: A vehicle state data acquisition module, configured to obtain a set of vehicle state data collected by a sensor assembly deployed on a new energy vehicle to obtain a set of vehicle state data. The sensor assembly includes an acceleration sensor, a vehicle speed sensor, a height sensor, and a vibration frequency sensor. The vehicle state data includes a vehicle acceleration value, a vehicle speed value, a vehicle height value, and a vehicle vibration frequency value; A vehicle state time-series feature extraction module, configured to perform data regularization and time-series feature extraction on the set of vehicle state data to obtain a vehicle acceleration time-series correlation feature implicit coding vector, a vehicle speed time-series correlation feature implicit coding vector, a vehicle height time-series change feature implicit coding vector, and a vehicle vibration frequency time-series correlation feature implicit coding vector; A vehicle state parameter correlation module, configured to perform multi-dimensional time-series correlation on the vehicle acceleration time-series correlation feature implicit coding vector, the vehicle speed time-series correlation feature implicit coding vector, the vehicle height time-series change feature implicit coding vector, and the vehicle vibration frequency time-series correlation feature implicit coding vector to obtain a time-series correlation coding feature map between vehicle state parameters; A vehicle state parameter time-series correlation enhancement module, configured to perform global-local vehicle state semantic mapping enhancement on the time-series correlation coding feature map between vehicle state parameters to obtain an enhanced time-series correlation coding feature map between vehicle state parameters; A pressure value adjustment module, configured to adjust the pressure value of the air spring based on the enhanced time-series correlation coding feature map between vehicle state parameters.

2. The intelligent control system of the air suspension shock absorber for new energy vehicles according to claim 1, characterized in that The vehicle state time-series feature extraction module includes: A vehicle state data splitting and regularization unit, configured to split the set of vehicle state data according to the parameter sample dimension, and regularize the split data according to the time dimension to obtain a time series of vehicle acceleration values, a time series of vehicle speed values, a time series of vehicle height values, and a time series of vehicle vibration frequency values; A vehicle state time-series feature generation unit, configured to perform time-series feature extraction based on one-dimensional convolution on the time series of vehicle acceleration values, the time series of vehicle speed values, the time series of vehicle height values, and the time series of vehicle vibration frequency values respectively to obtain the vehicle acceleration time-series correlation feature implicit coding vector, the vehicle speed time-series correlation feature implicit coding vector, the vehicle height time-series change feature implicit coding vector, and the vehicle vibration frequency time-series correlation feature implicit coding vector.

3. The intelligent control system of the air suspension shock absorber for new energy vehicles according to claim 2, characterized in that, The vehicle state parameter correlation module includes: A vehicle state feature multi-dimensional aggregation coding unit, configured to perform vehicle state feature multi-dimensional aggregation coding on the vehicle acceleration time-series correlation feature implicit coding vector, the vehicle speed time-series correlation feature implicit coding vector, the vehicle height time-series change feature implicit coding vector, and the vehicle vibration frequency time-series correlation feature implicit coding vector to obtain a vehicle state multi-dimensional time-series aggregation matrix; A vehicle state parameter temporal correlation encoding feature generation unit is configured to input the vehicle state multi-dimensional temporal aggregation matrix into a correlation feature extractor between vehicle state parameters based on a two-dimensional convolutional neural network to obtain a temporal correlation encoding feature map between vehicle state parameters.

4. The intelligent control system of the air suspension shock absorber for new energy vehicles according to claim 3, characterized in that, The vehicle state parameter temporal correlation enhancement module includes: A vehicle state feature global key semantic information extraction unit is configured to input the temporal correlation encoding feature map between vehicle state parameters into a global key semantic information extraction network to obtain a global temporal key semantic information encoding vector between vehicle state parameters; A to-be-processed feature window determination unit is configured to extract a channel feature vector between to-be-processed vehicle state parameters from the temporal correlation encoding feature map between vehicle state parameters, and determine a vehicle state local neighborhood semantic correlation window of the channel feature vector between to-be-processed vehicle state parameters; A semantic mapping guidance factor calculation unit is configured to calculate a vehicle state local key semantic information encoding vector corresponding to the channel feature vector between to-be-processed vehicle state parameters, and perform semantic comparison and analysis on the vehicle state local key semantic information encoding vector and the global temporal key semantic information encoding vector between vehicle state parameters to determine a vehicle state semantic mapping guidance factor of the channel feature vector between to-be-processed vehicle state parameters; A to-be-processed enhanced channel feature generation unit between vehicle state parameters is configured to modulate the channel feature vector between to-be-processed vehicle state parameters based on the vehicle state semantic mapping guidance factor to obtain an enhanced channel feature vector between to-be-processed vehicle state parameters, where the enhanced channel feature vector between to-be-processed vehicle state parameters is the channel feature vector at the pixel position (i, j) of the enhanced temporal correlation encoding feature map between vehicle state parameters.

5. The intelligent control system of the air suspension shock absorber for a new energy vehicle according to claim 4, characterized in that, The to-be-processed feature window determination unit is configured to: Extract a channel feature vector between vehicle state parameters at the pixel position (i, j) from the temporal correlation encoding feature map between vehicle state parameters as the channel feature vector between to-be-processed vehicle state parameters; Determine the vehicle state local neighborhood semantic correlation window of the channel feature vector between to-be-processed vehicle state parameters based on the feature distribution characteristics of the channel feature vector between to-be-processed vehicle state parameters, where the channel feature vector between to-be-processed vehicle state parameters is located at the center position of the vehicle state local neighborhood semantic correlation window.

6. The intelligent control system of the air suspension shock absorber for a new energy vehicle according to claim 5, wherein The semantic mapping guidance factor calculation unit includes: A vehicle state local key semantic information extraction subunit is configured to input all channel feature vectors between vehicle state parameters in the vehicle state local neighborhood semantic correlation window into a local neighborhood correlation semantic information extraction network to obtain the vehicle state local key semantic information encoding vector corresponding to the channel feature vector between to-be-processed vehicle state parameters; A vehicle state semantic mapping guidance factor calculation subunit, configured to determine the vehicle state semantic mapping guidance factor of the channel feature vector between the to-be-processed vehicle state parameters based on the semantic comparison analysis between the global temporal key semantic information encoding vector among the vehicle state parameters and the vehicle state local key semantic information encoding vector corresponding to the channel feature vector between the to-be-processed vehicle state parameters.

7. The intelligent control system of the air suspension shock absorber for new energy vehicles according to claim 6, characterized in that, The vehicle state semantic mapping guidance factor calculation subunit is configured to: Based on the channel feature vector between the to-be-processed vehicle state parameters, perform semantic interaction optimization based on eigen-diagonal double mediation on the vehicle state local key semantic information encoding vector corresponding to the channel feature vector between the to-be-processed vehicle state parameters to obtain an optimized vehicle state local key semantic information encoding vector of the channel feature vector between the to-be-processed vehicle state parameters; Calculate the Mahalanobis distance between the global temporal key semantic information encoding vector among the vehicle state parameters and the optimized vehicle state local key semantic information encoding vector of the channel feature vector between the to-be-processed vehicle state parameters as the vehicle state semantic mapping guidance factor of the channel feature vector between the to-be-processed vehicle state parameters.

8. The intelligent control system of the air suspension shock absorber for new energy vehicles according to claim 7, characterized in that, The pressure value adjustment module is configured to: Input the enhanced temporal correlation encoding feature map between the vehicle state parameters into a model prediction controller to obtain suspension optimization parameters, where the suspension optimization parameters include the pressure value of the air spring; Send the suspension optimization parameters to an actuator layer, and the actuator layer sends a control instruction to the air spring based on the comparison between the suspension optimization parameters and the real-time suspension parameters to adjust the pressure value of the air spring.

9. An intelligent control method for an air suspension shock absorber of a new energy vehicle, characterized in that, Includes: Obtain a set of vehicle state data collected by a sensor assembly deployed on a new energy vehicle to obtain a set of vehicle state data. The sensor assembly includes an acceleration sensor, a vehicle speed sensor, a height sensor, and a vibration frequency sensor, and the vehicle state data includes a vehicle acceleration value, a vehicle speed value, a vehicle height value, and a vehicle vibration frequency value; Perform data regularization and temporal feature extraction on the set of vehicle state data to obtain a vehicle acceleration temporal correlation feature implicit encoding vector, a vehicle speed temporal correlation feature implicit encoding vector, a vehicle height temporal change feature implicit encoding vector, and a vehicle vibration frequency temporal correlation feature implicit encoding vector; Perform multi-dimensional temporal correlation on the vehicle acceleration temporal correlation feature implicit encoding vector, the vehicle speed temporal correlation feature implicit encoding vector, the vehicle height temporal change feature implicit encoding vector, and the vehicle vibration frequency temporal correlation feature implicit encoding vector to obtain a temporal correlation encoding feature map between vehicle state parameters; Perform global-local vehicle state semantic mapping enhancement on the temporal correlation encoding feature map between vehicle state parameters to obtain an enhanced temporal correlation encoding feature map between vehicle state parameters; Based on the enhanced temporal correlation encoding feature map between vehicle state parameters, adjust the pressure value of the air spring.

10. The intelligent control method of the air suspension shock absorber for new energy vehicles according to claim 9, characterized in that, Based on the enhanced temporal correlation encoding feature map between vehicle state parameters, adjusting the pressure value of the air spring includes: Input the encoded feature map with enhanced temporal correlation between vehicle state parameters into the model predictive controller to obtain suspension optimization parameters, where the suspension optimization parameters include the pressure value of the air spring; Send the suspension optimization parameters to the actuator layer, and the actuator layer sends a control instruction to the air spring based on the comparison between the suspension optimization parameters and the real-time suspension parameters to adjust the pressure value of the air spring.

Images