Adjusting method and device for vehicle air suspension and vehicle

Abstract

The invention relates to the technical field of vehicle control, and discloses an adjusting method and device for a vehicle air suspension and a vehicle. Based on a deep learning model, the vehicle state is analyzed according to the vehicle running state parameters, and an air suspension adjusting instruction is generated; the adjustment instruction comprises an adjustment allowing instruction and an adjustment locking instruction; under the condition that the adjusting instruction is an allowed adjusting instruction, the air suspension is adjusted according to the allowed adjusting instruction; and under the condition that the adjusting instruction is the locking adjusting instruction, the air suspension is locked to be in the current state according to the locking adjusting instruction. The deep learning model can generate a corresponding adjustment instruction according to the vehicle operation condition with complex change, so that the response to various complex working conditions is realized, the performance of the air suspension is improved, and the stability and comfort of the vehicle are kept.

Description

Technical Field

[0001] This application relates to the technical field of vehicle control, for example, to a method and device for adjusting a vehicle air suspension and a vehicle. Background Art

[0002] The air suspension system is one of the important technologies for modern vehicles to improve driving comfort and handling stability. The air suspension system adjusts the suspension height and stiffness to adapt to different loads, road conditions, and driving requirements.

[0003] In the related art, an air suspension control method based on a preset rule logic is disclosed, where the preset rule logic is preset according to the engineer's experience and experimental data.

[0004] In the process of implementing the embodiments of the present disclosure, it is found that at least the following problems exist in the related art:

[0005] The preset rule logic is usually static and difficult to be adjusted in real time according to dynamic changes such as vehicle body posture and load, and cannot adapt to dynamic changes in real time.

[0006] It should be noted that the information disclosed in the above background art section is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0007] To have a basic understanding of some aspects of the disclosed embodiments, a simple summary is given below. The summary is not a general review, nor is it intended to identify key / important elements or delineate the protection scope of these embodiments, but rather serves as a preface to the subsequent detailed description.

[0008] The embodiments of the present disclosure provide a method and device for adjusting a vehicle air suspension and a vehicle to adapt to complex dynamic working conditions and improve the performance of the air suspension.

[0009] In some embodiments, the method for adjusting a vehicle air suspension includes: collecting vehicle operation state parameters; analyzing the vehicle state based on a deep learning model according to the vehicle operation state parameters to generate an adjustment instruction for the air suspension; the adjustment instruction includes an allowable adjustment instruction and a locking adjustment instruction; when the adjustment instruction is an allowable adjustment instruction, adjusting the air suspension according to the allowable adjustment instruction; when the adjustment instruction is a locking adjustment instruction, locking the air suspension to the current state according to the locking adjustment instruction.

[0010] Optionally, the deep learning model generates adjustment instructions for the air suspension according to the following process: perform correlation analysis based on the temporal characteristics of vehicle operation state parameters to generate multiple feature vectors; calculate the attention weights of each feature vector to obtain weighted feature vectors; determine the decision result according to the weighted feature vectors; the decision result includes allowing adjustment or locking adjustment; splice the weighted feature vectors and the decision result for analysis to generate an adjustment allowed instruction or a locking adjustment instruction for the air suspension; the adjustment allowed instruction includes the target height and direction for which the air suspension is to be adjusted, and the locking adjustment instruction includes the locking reason.

[0011] Optionally, determining the decision result according to the weighted feature vectors includes: determining the weight matrix corresponding to the weighted feature vectors; based on the weight matrix, performing a fully connected layer mapping on the weighted feature vectors to obtain two-dimensional raw scores; performing normalization processing on the two-dimensional raw scores to obtain an adjustment allowed probability and a locking adjustment probability; selecting the higher value of the adjustment allowed probability and the locking adjustment probability as the decision result.

[0012] Optionally, splicing the weighted feature vectors and the decision result for analysis to generate an adjustment allowed instruction or a locking adjustment instruction for the air suspension includes: obtaining the joint feature vectors after splicing the weighted feature vectors and the decision result; in the case where the decision result is allowing adjustment, inputting the joint feature vectors into a fully connected regression sub-network to obtain the target height and direction for which the air suspension is to be adjusted; in the case where the decision result is locking adjustment, inputting the joint feature vectors into a fully connected classification sub-network to obtain the locking reason.

[0013] Optionally, the deep learning model is trained in the following manner: build a neural network model; train the neural network model based on a training set, update the parameters of the neural network model to minimize the target loss function, and obtain the trained deep learning model; where the target loss function is the sum of the first loss function, the second loss function, and the third loss function, and is used to optimize the decision-making of the deep learning model; the first loss function is used to optimize the decision of determining whether the air suspension is adjusted or locked; the second loss function is used to optimize the decision of determining the locking reason; the third loss function is used to optimize the decision of determining the adjustment height and direction.

[0014] Optionally, adjusting the air suspension according to the adjustment allowed instruction includes: obtaining the height difference between the current height of the air suspension and the target height in the adjustment allowed instruction; according to the height difference and the current driving state of the vehicle, controlling the power and working duration of the vehicle's compressor, and the opening degrees of the vehicle's intake valve and exhaust valve to achieve the adjustment of the height and direction of the air suspension.

[0015] Optionally, lock the air suspension in the current state according to the lock adjustment instruction, including: controlling the compressor of the vehicle to stop running, and closing the intake valve or exhaust valve of the vehicle to lock the air pressure and height of the air suspension in the current state.

[0016] Optionally, the adjustment method further includes: during the adjustment of the air suspension, monitoring the change of the vehicle state; in the case of abnormal change of the vehicle state, suspending the adjustment operation of the air suspension; and resuming the adjustment operation of the air suspension after the vehicle state is stable again.

[0017] Optionally, the adjustment method further includes: recording the vehicle state information when the air suspension is locked and adjusted; after the air suspension is unlocked, according to the recorded vehicle state information, adopting a progressive adjustment strategy to adjust the valve opening and compressor power to make the air suspension return to the state before locking.

[0018] Optionally, the adjustment method further includes: during the adjustment of the air suspension, feedback the adjustment state of the air suspension to the user through the user interface; and / or, during the locking process of the air suspension, feedback the locking reason to the user through the user interface.

[0019] In some embodiments, an adjustment device for a vehicle air suspension includes a processor and a memory storing program instructions, and the processor is configured to execute the adjustment method for the vehicle air suspension as described above when running the program instructions.

[0020] In some embodiments, a vehicle includes: a vehicle body provided with an air suspension; the adjustment device for the vehicle air suspension as described above, which is installed on the vehicle body and connected to the air suspension.

[0021] The adjustment method, device and vehicle for a vehicle air suspension provided by the embodiments of the present disclosure can achieve the following technical effects:

[0022] In the embodiments of the present disclosure, by analyzing the vehicle operation state parameters through a deep learning model, it is possible to achieve real-time and accurate identification of the vehicle state, and then achieve the adjustment of the air suspension. According to the specific working conditions of the vehicle, the deep learning model can also identify potential risks and generate a lock adjustment instruction to lock the air suspension in the current state, preventing unnecessary adjustment of the air suspension, thereby avoiding sudden changes in the vehicle posture. Therefore, the deep learning model can generate corresponding adjustment instructions according to the complex and changing vehicle operation conditions, realize the response to various complex working conditions, improve the performance of the air suspension, and maintain the stability and comfort of the vehicle.

[0023] The above general description and the following description are only exemplary and explanatory, and are not used to limit this application. Description of the Drawings

[0024] One or more embodiments are exemplarily illustrated by corresponding accompanying drawings. These exemplary illustrations and the accompanying drawings do not constitute a limitation on the embodiments. Elements with the same reference numerals in the accompanying drawings are shown as similar elements. The accompanying drawings do not constitute a scale limitation, and wherein:

[0025] Figure 1 is a schematic diagram of a method for adjusting a vehicle air suspension provided by an embodiment of the present disclosure;

[0026] Figure 2 is a schematic diagram of a system for adjusting a vehicle air suspension provided by an embodiment of the present disclosure;

[0027] Figure 3 is a schematic diagram of a method for generating an adjustment instruction for an air suspension provided by an embodiment of the present disclosure;

[0028] Figure 4 is a schematic diagram of another method for adjusting a vehicle air suspension provided by an embodiment of the present disclosure;

[0029] Figure 5 is a schematic diagram of another method for adjusting a vehicle air suspension provided by an embodiment of the present disclosure;

[0030] Figure 6 is a schematic diagram of a device for adjusting a vehicle air suspension provided by an embodiment of the present disclosure. Detailed Description of the Embodiments

[0031] In order to be able to understand the features and technical content of the embodiments of the present disclosure in more detail, the implementation of the embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for reference and illustration only and are not used to limit the embodiments of the present disclosure. In the following technical description, for the sake of explanation, numerous details are provided to give a thorough understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be shown in a simplified manner to simplify the accompanying drawings.

[0032] The terms "first", "second", etc. in the technical solutions described in this application are used to distinguish similar objects and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so as to implement the embodiments of the present disclosure described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion.

[0033] Unless otherwise specified, the term "plurality" means two or more.

[0034] In the embodiments of the present disclosure, the character " / " indicates an "or" relationship between the front and rear objects. For example, A / B means: A or B.

[0035] The term "and / or" is an associative relationship describing an object, indicating that three relationships can exist. For example, A and / or B means: A or B, or, the three relationships of A and B.

[0036] The term "corresponding" can refer to an associative relationship or a binding relationship. A corresponding to B means that there is an associative relationship or a binding relationship between A and B.

[0037] Combined with Figure 1 As shown, the embodiments of the present disclosure provide an adjustment method for a vehicle air suspension. The execution subject of the adjustment method can be a processor. The adjustment method includes:

[0038] S101, the processor collects vehicle operation state parameters.

[0039] S102, the processor analyzes the vehicle state based on the deep learning model according to the vehicle operation state parameters, and generates an adjustment instruction for the air suspension. Among them, the adjustment instruction includes an allow adjustment instruction and a lock adjustment instruction.

[0040] S103, when the adjustment instruction is an allow adjustment instruction, the processor adjusts the air suspension according to the allow adjustment instruction.

[0041] S104, when the adjustment instruction is a lock adjustment instruction, the processor locks the air suspension to the current state according to the lock adjustment instruction.

[0042] In the embodiments of the present disclosure, by analyzing the vehicle operation state parameters through the deep learning model, it is possible to achieve real-time and accurate identification of the vehicle state, and then achieve the adjustment of the air suspension. According to the specific working conditions of the vehicle, the deep learning model can also identify potential risks and generate a lock adjustment instruction to lock the air suspension in the current state, preventing unnecessary adjustment of the air suspension, thereby avoiding sudden changes in the vehicle posture. Therefore, the deep learning model can generate corresponding adjustment instructions according to the complex and changing vehicle operation conditions, achieve the response to various complex working conditions, improve the performance of the air suspension, and maintain the stability and comfort of the vehicle.

[0043] Combined with Figure 2 As shown, the embodiments of the present disclosure provide an adjustment system for a vehicle air suspension, including a data acquisition module, a data preprocessing module, a deep learning model, an actuator, and a feedback and interaction module. The adjustment system will be described below in combination with the adjustment method.

[0044] Optionally, combined with Figure 2As shown, the vehicle operating state parameters include one or more of: air suspension height adjustment mode, actuator working duration, left front suspension height, right front suspension height, left rear suspension height, right rear suspension height, door state, slope value, steering wheel angle, steering wheel angle rate, pitch angle, roll angle, compressor temperature, lateral acceleration, longitudinal acceleration, ABS (Anti-lock Braking System) state, TCS (Traction Control System) state, and ESC (Electronic Stability Control) state. After the vehicle operating state parameters are collected by the data acquisition module, they are transmitted via the CAN bus to the data preprocessing module for preprocessing and encoding, and then input into the deep learning model for generating adjustment instructions for the suspension system.

[0045] In this embodiment, after the vehicle is started, an initialization operation is first performed on the entire air suspension system. Specifically, it includes self-checking each sensor in the data acquisition module to ensure its normal operation and the accuracy of data acquisition. Then, the vehicle operating state parameters are collected by each sensor at the set sampling frequency. Among them, in combination with Figure 2 As shown, the sensors include: ECU (Electronic Control Unit), 4 height sensors, door lock contacts, gyroscopes, IMU (Inertial Measurement Unit), steering wheel angle sensors, temperature sensors, and accelerometers. After the collection is completed, the collected vehicle operating state parameter data is transmitted via the CAN bus and temporarily stored in the buffer for further processing.

[0046] Optionally, the steering wheel angle and the steering wheel angle rate are collected by the steering wheel angle and angle rate sensors provided on the steering column, and their sampling frequency is 100Hz.

[0047] Optionally, a high-speed CAN line is directly connected to the vehicle electronic stability system to obtain the working modes and states of ABS, TCS, and ESP in real time.

[0048] Optionally, the lateral acceleration and longitudinal acceleration of the vehicle are collected by the lateral / longitudinal acceleration sensors provided in the vehicle centroid area.

[0049] Optionally, the compressor temperature is collected by the compressor temperature sensor provided at the key part of the compressor housing, and its sampling frequency is 100Hz.

[0050] Optionally, the pitch angle and roll angle of the vehicle are collected by the IMU, and the IMU is installed at the center of the vehicle chassis to accurately sense the body attitude, and its sampling frequency is 100Hz.

[0051] Optionally, the door state is collected by a door state sensor provided at the connection between the door and the vehicle body.

[0052] Optionally, the suspension height is collected by suspension height sensors provided at both ends of the suspension strut, and its sampling frequency is 100 Hz.

[0053] Optionally, the air suspension height adjustment mode request and the actuator working duration are collected by the ECU.

[0054] Optionally, preprocessing is performed on the collected vehicle operating state parameters, including noise removal and dimension unification.

[0055] Optionally, an adaptive filtering algorithm is used to remove noise interference in the collected data. Taking the vehicle pitch angle data as an example, the adaptive filtering algorithm automatically adjusts the parameters of the filter according to the real-time change characteristics of the sensor data.

[0056] Optionally, a standardization method is adopted to ensure the comparability of data with different physical dimensions when entering the deep learning model. In this way, the input range of data in the deep learning model can be made more reasonable, avoiding the influence of too large or too small data on the learning effect of the model, and facilitating the deep learning model to capture the correlation between different dimensions. For discrete data, such as the four air suspension height adjustment modes of raising, lowering, maintaining, and adaptive, one-hot encoding is used to encode them as [1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0], and [0, 0, 0, 1] respectively. For continuous data, such as the vehicle lateral acceleration, after dimension standardization, linear scaling processing is performed to map it to the [-1, 1] interval.

[0057] Optionally, the deep learning model generates an adjustment instruction for the air suspension according to the following process: performing correlation analysis based on the temporal characteristics of the vehicle operating state parameters to generate multiple feature vectors; calculating the attention weights of each feature vector to obtain weighted feature vectors; determining a decision result according to the weighted feature vectors; the decision result includes allowing adjustment or locking adjustment; splicing the weighted feature vectors and the decision result for analysis to generate an allowable adjustment instruction or a locking adjustment instruction for the air suspension; the allowable adjustment instruction includes the target height and direction to be adjusted for the air suspension, and the locking adjustment instruction includes the locking reason.

[0058] Combined with Figure 3 As shown, the embodiments of the present disclosure provide a method for generating an adjustment instruction for an air suspension. The execution subject of the generation method is a deep learning model, and the generation method includes:

[0059] S301, the deep learning model performs correlation analysis based on the temporal characteristics of the vehicle operating state parameters to generate multiple feature vectors.

[0060] In S302, the deep learning model calculates the attention weights of each feature vector to obtain the weighted feature vectors.

[0061] In S303, the deep learning model determines the decision result based on the weighted feature vectors. The decision result includes allowing adjustment or locking the adjustment.

[0062] In S304, the deep learning model splices the weighted feature vectors and the decision result for analysis to generate an adjustment permission instruction or a locking adjustment instruction for the air suspension. The adjustment permission instruction includes the target height and direction for the air suspension to be adjusted, and the locking adjustment instruction includes the locking reason.

[0063] In this embodiment, based on the temporal features of the vehicle operating state parameters, the deep learning model can more comprehensively understand the vehicle operating state. For example, the continuous steering angle change of the steering wheel can reflect the vehicle's steering trend, and the vehicle's acceleration change can reflect the vehicle's dynamic behavior. Conducting a correlation analysis on the temporal features of multiple parameters, such as combining the steering wheel angle with the vehicle roll angle for analysis, can more accurately determine the vehicle's stability and suspension requirements during turning, thereby generating multiple feature vectors. Under different working conditions, the importance of certain features for decision-making may be higher. For example, when driving at high speed, the vehicle's lateral acceleration and pitch angle may be more critical for suspension adjustment decisions; while when driving at low speed, the suspension height and road surface slope may be more important. By calculating the attention weights of each feature vector, the deep learning model can dynamically highlight important features and improve the accuracy of decision-making. The weighted feature vectors are a comprehensive representation of the vehicle's current state by the deep learning model, containing key information about the vehicle operating state, such as the weighted values of parameters such as the vehicle's speed, acceleration, steering wheel angle, and suspension height, which reflect the importance of each parameter in the current decision. Based on the weighted feature vectors, the deep learning model can more accurately determine the decision result, that is, allowing adjustment or locking the adjustment. By splicing the weighted feature vectors and the decision result, the deep learning model can further analyze and optimize the adjustment instruction, generate the target height and direction of the adjustment permission instruction, making the suspension adjustment more precise, and the locking reason of the locking adjustment instruction, so as to clearly understand the specific situation of the suspension locking and enhance the driving safety and transparency.

[0064] Optionally, determining the decision result according to the weighted feature vectors includes: determining the weight matrix corresponding to the weighted feature vectors; based on the weight matrix, performing a fully connected layer mapping on the weighted feature vectors to obtain two-dimensional raw scores; normalizing the two-dimensional raw scores to obtain an adjustment permission probability and a locking adjustment probability; and selecting the higher of the adjustment permission probability and the locking adjustment probability as the decision result.

[0065] Optionally, the weight matrix is used to map the weighted feature vector to the decision space. The determination of the weight matrix is completed through training. During the training process, based on a large number of sample data, including the operating states of the vehicle under different working conditions and the corresponding adjustment instructions, the influence degree of each feature on the decision is learned, and these influence degrees are stored in the weight matrix in the form of weights.

[0066] In this embodiment, by determining the weight matrix corresponding to the weighted feature vector, the deep learning model can quantify the importance of each feature, and then through the fully connected layer mapping, obtain the two-dimensional raw score. The two-dimensional raw score includes the allowable adjustment score and the locked adjustment score. Through normalization processing, the two-dimensional raw score is converted into the allowable adjustment probability and the locked adjustment probability within the range of [0, 1], and it is ensured that the sum of the two probability values is 1. For example, if the allowable adjustment probability is 0.8 and the locked adjustment probability is 0.2, it can be determined that the possibility of allowing adjustment is greater in the current state.

[0067] Optionally, after splicing the weighted feature vector and the decision result for analysis, an allowable adjustment instruction or a locked adjustment instruction for the air suspension is generated, including: obtaining the joint feature vector after splicing the weighted feature vector and the decision result; in the case where the decision result is allowable adjustment, inputting the joint feature vector into the fully connected regression sub-network to obtain the target height and direction to be adjusted for the air suspension; in the case where the decision result is locked adjustment, inputting the joint feature vector into the fully connected classification sub-network to obtain the locking reason.

[0068] In this embodiment, the joint feature vector not only contains the feature information of the vehicle state, but also adds the context information of the decision result, providing a more comprehensive input for the subsequent instruction generation. When the decision result is allowable adjustment, the joint feature vector is input into the fully connected regression sub-network to predict the target height and direction to be adjusted for the air suspension according to the joint feature vector. The fully connected regression sub-network consists of multiple fully connected layers, and each layer contains a large number of neurons, which are used to learn the mapping relationship between the input features and the output target, that is, the continuous mapping relationship between the input features and the target height. When the decision result is locked adjustment, the joint feature vector is input into the fully connected classification sub-network to predict the reason for the locked adjustment according to the joint feature vector. The fully connected classification sub-network also consists of multiple fully connected layers, but its output is a classification result indicating the specific reason for the locked adjustment, and its learning process is the discrete mapping relationship between the input features and the locking reason. Among them, the locking reason may be "vehicle speed is too high", "road surface slope is too large" or "suspension system failure", etc.

[0069] Optionally, the deep learning model is trained as follows: build a neural network model; train the neural network model based on the training set, update the parameters of the neural network model to minimize the objective loss function, and obtain the trained deep learning model; wherein, the objective loss function is the sum of the first loss function, the second loss function, and the third loss function, and is used to optimize the decision-making of the deep learning model; the first loss function is used to optimize the decision of determining whether the air suspension is adjusted or locked; the second loss function is used to optimize the decision of determining the locking reason; the third loss function is used to optimize the decision of determining the adjustment height and direction.

[0070] In this embodiment, after collecting the state parameters of the vehicle under different working conditions and dividing the training set, validation set, and test set proportionally, a neural network model is built, and then the trained deep learning model is obtained. As shown in Figure 2 The deep learning model includes an input layer, a hidden layer, an attention network layer, and an output layer.

[0071] The input layer uses 18 neurons to connect to 18 preprocessed vehicle running state parameters respectively. After receiving the input data, these neurons pass it to the hidden layer.

[0072] The hidden layer uses a 4-layer LSTM (Long Short-Term Memory Network) network, with 512, 256, 128, and 64 neurons set in each layer respectively. The number of neurons in the first layer of LSTM is 512, and its main function is to initially extract the temporal features in the data. For example, for the steering wheel angle data sequence at consecutive sampling moments of the vehicle, the first layer of LSTM can learn the trend and periodic features of the angle change. The 256 neurons in the second layer of LSTM further explore deeper feature combinations on this basis, such as correlating the steering wheel angle feature with the vehicle's lateral acceleration feature to judge the vehicle's steering stability. The 128 neurons in the third layer of LSTM focus on fusing more different types of feature information, such as combining the vehicle body attitude feature with the suspension height adjustment historical data to predict the reasonable adjustment amount and direction of the current suspension. The 64 neurons in the fourth layer of LSTM refine and compress the complex features extracted previously, and extract the key feature information for subsequent decision-making. Some common neural networks can be used in the hidden layer, such as convolutional neural networks, recurrent neural networks, etc. Preferably, considering the temporal information in the air suspension system, a recurrent neural network or its variant is used as the hidden layer. To further extract the correlation between different states, on the premise of sufficient computing power, the number of layers of the neural network can also be deepened, such as a 2-layer recurrent neural network or a 4-layer recurrent neural network, etc. The number of neurons in each layer of the neural network can be set differently according to actual needs and complexity.

[0073] The function of the attention network layer is to calculate the attention weights of each feature vector based on the attention mechanism according to the feature vectors output by the LSTM layer, so as to highlight the importance of certain key features for decision-making. Under different working conditions of the vehicle, such as high-speed driving and low-speed driving, the importance of different features for the final decision of the neural network is very different. For example, in the case of high-speed driving of the vehicle, features such as the lateral / longitudinal acceleration of the vehicle and the road surface slope have a greater impact on the decision of whether the air suspension allows adjustment. The attention network layer will assign higher weights to these features, while the weights of other features are relatively low. The LSTM network obtains the weighted feature vectors after passing through the attention layer.

[0074] The output layer contains two neurons, and its function is to map the hidden state of the neural network to two results: allowing suspension adjustment or pausing suspension adjustment. Subsequently, the decision result is concatenated with the weighted feature vector output by the attention network to form a joint feature vector. Finally, the joint feature vector is input into the fully connected network to determine the reason for locking in the locked state or the target height and direction of the suspension to be adjusted in the non-locked state.

[0075] After the neural network model is built, the collected vehicle state parameter data is divided into a training set, a validation set, and a test set according to the ratio of 8:1:1. The training set is used for learning and updating the model parameters, the validation set is used to monitor the model performance during the training process, and the test set is used to finally evaluate the generalization ability and accuracy of the deep learning model.

[0076] During the model training process, for the determination of whether to pause adjustment and the classification of specific locking reasons, the cross-entropy loss function is adopted, as follows:

[0077]

[0078] Among them, L lack is the first loss function, is the determination result of whether to lock and adjust output by the neural network model, y i is the true label of locking or not locking in the training set, and N is the number of samples. L lack_reason is the second loss function, is the reason for locking and adjusting output by the neural network model, C i is the true reason for locking and adjusting in the training set, and M is the number of samples.

[0079] For the target height decision, the mean square error loss function is adopted, as follows:

[0080]

[0081] Among them, L hei is the third loss function, is the target height value output by the neural network model, H i is the true target height value in the training set, and K is the number of samples. The training data sample size is equal to the sum of the locked adjustment sample size and the allowed adjustment sample size, i.e., N = M + K.

[0082] Take the sum of the first loss function, the second loss function, and the third loss function as the total loss function, specifically as follows:

[0083] Loss = αL lack + βL lack_reason + γL hei

[0084] where α, β, and γ are the weights of the three loss functions respectively. Optionally, α = β = γ = 1 / 3.

[0085] Use the adaptive moment estimation algorithm to update the parameters of the neural network model, aiming to minimize the value of the total loss function Loss.

[0086] During the model performance evaluation process, in addition to common metrics such as accuracy, specific metrics for the air suspension system are also introduced, such as the locked decision error rate, etc. At the same time, visualization tools are used to monitor the training process of the model, such as plotting the loss function curve, accuracy curve, etc., to intuitively display the performance changes during the model training process.

[0087] During the actual operation of the deep learning model, the deep learning model continuously receives data collected in real time by the vehicle. The data flows through each layer in turn. The model will perform real-time state analysis and feature extraction on the input data, perform state recognition and decision-making, accurately identify the current state information of the vehicle, judge whether there are potential operation risks, generate adjustment or locked adjustment instructions and send them to the actuator to achieve intelligent control of the air suspension system. At the same time, the deep learning model can also continuously collect data during the vehicle operation and perform further optimization and update through algorithms such as semi-supervised learning, unsupervised learning, and transfer learning to adapt to the continuously changing operation conditions and environment of the vehicle and continuously improve the performance and reliability of the model.

[0088] Optionally, after obtaining the trained deep learning model, deploy the deep learning model to the vehicle.

[0089] During the deployment of a deep learning model, it is first necessary to determine the hardware selection. Subsequently, proceed to build the operating environment and correctly configure the software and hardware drivers, install a stable operating system, and remove redundant functions to reserve sufficient computing power and memory space for the deep learning model. Secondly, perform lightweight processing on the parameters of the deep learning model. The volume of the trained original model is often large, and direct deployment will exceed the carrying limit of in-vehicle hardware. Therefore, model compression technology should be used to reduce the redundant parameters of the model. For example, adopt a pruning strategy to cut off neuron connections and parameters that contribute little to the model's decision-making or convert high-precision floating-point parameters into lower-precision integer numbers. After lightweight processing, the memory occupancy and computational volume are greatly reduced, making the deep learning model compact and efficient, so that it can be easily deployed to the in-vehicle hardware side. Thirdly, it is necessary to establish a communication link between the deep learning model and the vehicle's electronic architecture. The core of the deployment lies in deeply integrating the deep learning model with the existing vehicle electronic architecture, closely connecting the deep learning model with sensors scattered throughout the vehicle through the CAN bus, and building a real-time and stable data transmission link. Fourthly, establish an interaction mechanism between the deep learning model and the vehicle actuator. When the deep learning model outputs a suspension locking or adjustment decision, ensure that the signal output from the deep learning model can be quickly fed back to the vehicle's actuator to form a closed-loop information flow. Finally, conduct on-vehicle testing and optimization calibration. After the deployment is completed, conduct a comprehensive test on the vehicle, simulate various driving scenarios, and collect deviation data between the output results of the deep learning model and the actual suspension adjustment state. If there is a deviation, the weights of the relevant layers of the deep learning model need to be fine-tuned to make the output conform to the true mechanical characteristics of the vehicle and ensure a stable driving posture.

[0090] Optionally, adjusting the air suspension according to an allowable adjustment instruction includes: obtaining the height difference between the current height of the air suspension and the target height in the allowable adjustment instruction; controlling the power and working duration of the vehicle's compressor and the opening degrees of the vehicle's intake valve and exhaust valve according to the height difference and the current driving state of the vehicle to achieve the adjustment of the height and direction of the air suspension.

[0091] Combined Figure 4 As shown, another method for adjusting a vehicle air suspension provided by an embodiment of the present disclosure includes:

[0092] S401, the processor collects vehicle operation state parameters.

[0093] S402, the processor analyzes the vehicle state based on the deep learning model according to the vehicle operation state parameters and generates an adjustment instruction for the air suspension. Among them, the adjustment instruction includes an allowable adjustment instruction and a locking adjustment instruction.

[0094] S403, when the adjustment instruction is an allowable adjustment instruction, the processor obtains the height difference between the current height of the air suspension and the target height in the allowable adjustment instruction.

[0095] S404. The processor controls the power and working duration of the vehicle's compressor, as well as the opening degrees of the vehicle's intake valve and exhaust valve, according to the height difference and the current driving state of the vehicle, so as to adjust the height and direction of the air suspension.

[0096] S405. When the adjustment instruction is a locking adjustment instruction, the processor controls the vehicle's compressor to stop running and closes the vehicle's intake valve or exhaust valve, so as to lock the air pressure and height of the air suspension in the current state.

[0097] Optionally, the adjustment method further includes: during the adjustment process of the air suspension, monitoring the change of the vehicle state; in the case of abnormal change of the vehicle state, pausing the adjustment operation of the air suspension; and continuing the adjustment operation of the air suspension after the vehicle state is stable again.

[0098] In this embodiment, in combination with Figure 2 As shown, when the deep learning model determines that the air suspension is in an unlocked state, it simultaneously sends the output target height data to the actuator. Based on the current suspension height and the target height, according to the preset height difference and compressor power curve, it calculates the working power and working duration of the compressor, and combines the current driving state of the vehicle, such as the dynamic and static states of the vehicle, vehicle speed, etc., to adjust the opening degrees of the intake and exhaust valves, and orderly adjusts the intake and exhaust flow rates, ensuring that the suspension height adjustment can change according to the predetermined adjustment curve, making the suspension height gradually approach the target height, rather than mutating or oscillating repeatedly, so as to achieve stable adjustment of the suspension height. For example, when the vehicle is in an accelerating state, in order to ensure the stability of the vehicle, the actuator will appropriately reduce the opening degree of the intake or exhaust valve to achieve stable adjustment of the suspension. During the adjustment process of the air suspension, the deep learning model monitors the change of the vehicle state in real time. For example, when the vehicle suddenly brakes, the deep learning model will immediately control the actuator to pause the suspension adjustment action and wait for the vehicle state to be stable before continuing the adjustment, avoiding affecting safety and comfort due to the conflict between suspension adjustment and vehicle driving state. If the suspension adjustment requirement changes due to external interference, such as sudden acceleration, sudden deceleration or road surface bumps, the working parameters of the compressor and the opening degree of the valve are adjusted in time to ensure the dynamic and static adaptability and stability of the suspension adjustment.

[0099] Optionally, locking the air suspension in the current state according to the locking adjustment instruction includes: controlling the vehicle's compressor to stop running and closing the vehicle's intake valve or exhaust valve, so as to lock the air pressure and height of the air suspension in the current state.

[0100] Optionally, the adjustment method also includes: recording vehicle status information during air suspension lock adjustment; after the air suspension is unlocked, adjusting the valve opening and compressor power using a progressive adjustment strategy based on the recorded vehicle status information to restore the air suspension to the state before locking.

[0101] In this embodiment, when the deep learning model determines that the air suspension is in a locked adjustment state, a locking adjustment instruction is immediately issued to the actuator. After receiving the locking adjustment, the actuator quickly stops the operation of the compressor and closes the intake and exhaust valves within 100 milliseconds to lock the air pressure and height of the suspension system in the current state. The actuator also records the vehicle state information during the air suspension locking adjustment, such as vehicle speed, acceleration, suspension height, etc. in the cache, so that a smooth transition can be made based on this information when the lock is released. For example, when the vehicle is driving at high speed and encounters a turning road, the deep learning model determines that the suspension adjustment needs to be locked to maintain the driving stability of the vehicle. After the actuator performs the locking operation, it records the vehicle speed and lateral / longitudinal acceleration information at that time. After the deep learning model issues a lock release instruction, the actuator uses a progressive adjustment strategy based on these recorded information to slowly adjust the valve opening and compressor power, so that the suspension gradually returns to a normal adjustment state, avoiding the sudden change of the vehicle posture at the maximum compressor power and the maximum valve opening at the moment of lock release, thereby affecting ride comfort and safety.

[0102] Optionally, the adjustment method further includes: during the adjustment process of the air suspension, feeding back the adjustment state of the air suspension to the user through the user interface; and / or, during the locking process of the air suspension, feeding back the locking reason to the user through the user interface.

[0103] Combination Figure 5 As shown, the embodiment of the present disclosure provides another adjustment method for a vehicle air suspension, comprising:

[0104] S501, the processor collects vehicle operating status parameters.

[0105] S502: The processor analyzes the vehicle state based on the deep learning model according to the vehicle running state parameters and generates an adjustment instruction for the air suspension, wherein the adjustment instruction includes an allow adjustment instruction and a lock adjustment instruction.

[0106] S503: When the adjustment instruction is an allowable adjustment instruction, the processor adjusts the air suspension according to the allowable adjustment instruction.

[0107] S504: The processor feeds back the adjustment state of the air suspension to the user through the user interface.

[0108] S505: When the adjustment instruction is a locking adjustment instruction, the processor locks the air suspension to a current state according to the locking adjustment instruction.

[0109] S506, the processor feeds back the locking reason to the user through the user interface.

[0110] In this embodiment, in combination with Figure 2 As shown, in order to improve the user's intuitive perception of the working state of the air suspension system, the present invention designs a suspension system state feedback and interaction module. During the suspension adjustment process, the air suspension system communicates with the vehicle computer in real time via the CAN bus. The air suspension system feeds back information such as the target adjustment height and the current height of the suspension to the vehicle computer, and the user interface of the vehicle computer system displays the adjustment process of the suspension in the form of a 3D animation. By building a dynamic suspension model, as the height of the suspension changes, the suspension components in the model expand and contract accordingly. At the same time, the real-time height value, adjustment progress percentage, and estimated remaining adjustment time of the suspension are displayed on the interface, allowing the user to intuitively understand the dynamic situation of the suspension adjustment. When the air suspension pauses adjustment, the vehicle computer system pops up a prompt box, which displays the reason for pausing adjustment in prominent font, such as "The compressor is overheated, the air suspension pauses adjustment, please try again after the compressor cools down!". In addition to popping up the prompt box to remind, the vehicle computer system can also provide a voice broadcast reminder function to ensure that the user can timely know the state change of the suspension system, enhance the user's perception of the vehicle state, and improve driving pleasure and safety.

[0111] In combination with Figure 6 As shown, an adjustment device 600 for a vehicle air suspension provided by an embodiment of the present disclosure includes a processor 700 and a memory 701, and may further include a communication interface 702 and a bus 703. Among them, the processor 700, the communication interface 702, and the memory 701 can complete mutual communication through the bus 703. The communication interface 702 can be used for information transmission. The processor 700 can call the logical instructions in the memory 701 to execute the adjustment method for the vehicle air suspension in the above embodiment.

[0112] In addition, when the logical instructions in the above-mentioned memory 701 are implemented in the form of a software functional unit and sold or used as an independent product, they can be stored in a computer-readable storage medium.

[0113] The memory 701, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as the program instructions / modules corresponding to the methods in the embodiments of the present disclosure. The processor 700 executes functional applications and data processing by running the program instructions / modules stored in the memory 701, that is, implements the adjustment method for the vehicle air suspension in the above method embodiments.

[0114] The memory 701 may include a program storage area and a data storage area. Among them, the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created according to the use of the terminal device, etc. In addition, the memory 701 may include a high-speed random access memory and may also include a non-volatile memory.

[0115] An embodiment of the present disclosure provides a vehicle, including: a vehicle body provided with an air suspension; an adjusting device for the vehicle air suspension as described above, which is installed on the vehicle body and connected to the air suspension. The installation relationship described here is not limited to being placed inside the vehicle, but also includes installation connections with other components of the vehicle, including but not limited to physical connections, electrical connections, or signal transmission connections, etc.

[0116] An embodiment of the present disclosure provides a computer-readable storage medium storing computer-executable instructions, and the computer-executable instructions are set to execute the above-mentioned adjusting method for the vehicle air suspension.

[0117] An embodiment of the present disclosure provides a computer program product. The computer program product includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions that, when executed by a computer, cause the computer to execute the above-mentioned adjusting method for the vehicle air suspension.

[0118] The above-mentioned computer-readable storage medium may be a transient computer-readable storage medium or a non-transient computer-readable storage medium.

[0119] The technical solution of the embodiment of the present disclosure may be embodied in the form of a software product. The computer software product is stored in a storage medium and includes one or more instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiment of the present disclosure. The foregoing storage medium may be a non-transient storage medium, including: various media that can store program codes such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc, or may also be a transient storage medium.

[0120] The above description and the accompanying drawings fully illustrate the embodiments of the present disclosure, enabling those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, process, and other changes. The embodiments merely represent possible variations. Unless explicitly required, individual components and functions are optional, and the order of operations may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. The scope of the embodiments of the present disclosure includes the entire scope of the claims and all available equivalents of the claims. When used in this application, although terms such as "first", "second", etc. may be used in this application to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without changing the meaning of the description, the first element may be called the second element, and similarly, the second element may be called the first element, as long as all occurrences of "the first element" are consistently renamed and all occurrences of "the second element" are consistently renamed. The first element and the second element are both elements, but they may not be the same element. Moreover, the terms used in this application are only used to describe the embodiments and do not limit the claims. As used in the description of the embodiments and the claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to also include the plural forms. Similarly, as used in this application, the term "and / or" refers to any and all possible combinations including one or more of the associated listed items. Additionally, when used in this application, the term "comprise" and its variants "comprises" and / or "comprising", etc. mean the presence of the stated features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or groupings thereof. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, or device including the element. In this document, what each embodiment focuses on may be the differences from other embodiments, and the same or similar parts among the various embodiments may be referred to each other. For the methods, products, etc. disclosed in the embodiments, if they correspond to the method parts disclosed in the embodiments, the relevant parts may refer to the description of the method parts.

[0121] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner may depend on the specific application and design constraints of the technical solution. The skilled person can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the embodiments of the present disclosure. The skilled person can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0122] In the embodiments disclosed herein, the disclosed methods, products (including but not limited to devices, equipment, etc.) can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units can be merely a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in an electrical, mechanical, or other form. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to implement this embodiment. In addition, in the embodiments of the present disclosure, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0123] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a portion of code, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions noted in the blocks may occur in a different order than noted in the accompanying drawings. For example, two consecutive blocks may in fact be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. In the description corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may in fact be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

Claims

1. A method for adjusting a vehicle air suspension, characterized in that, Including: Collecting vehicle operating state parameters; Based on a deep learning model, analyzing the vehicle state according to the vehicle operating state parameters, and generating an adjustment instruction for the air suspension; The adjustment instruction includes an allowable adjustment instruction and a locking adjustment instruction; In the case where the adjustment instruction is an allowable adjustment instruction, adjusting the air suspension according to the allowable adjustment instruction; In the case where the adjustment instruction is a locking adjustment instruction, locking the air suspension to the current state according to the locking adjustment instruction.

2. The adjustment method according to claim 1, characterized in that The deep learning model generates an adjustment instruction for the air suspension according to the following process: Performing correlation analysis according to the temporal characteristics of the vehicle operating state parameters to generate multiple feature vectors; Calculating the attention weight of each feature vector to obtain a weighted feature vector; Determining a decision result according to the weighted feature vector; the decision result includes allowable adjustment or locking adjustment; Analyzing after splicing the weighted feature vector and the decision result to generate an allowable adjustment instruction or a locking adjustment instruction for the air suspension; the allowable adjustment instruction includes the target height and direction to be adjusted for the air suspension, and the locking adjustment instruction includes the locking reason.

3. The adjustment method according to claim 2, characterized in that Determining a decision result according to the weighted feature vector, including: Determining the weight matrix corresponding to the weighted feature vector; Based on the weight matrix, performing a fully connected layer mapping on the weighted feature vector to obtain a two-dimensional raw score; Performing normalization processing on the two-dimensional raw score to obtain an allowable adjustment probability and a locking adjustment probability; Selecting the higher one of the allowable adjustment probability and the locking adjustment probability as the decision result.

4. The adjustment method according to claim 2, characterized in that Analyzing after splicing the weighted feature vector and the decision result to generate an allowable adjustment instruction or a locking adjustment instruction for the air suspension, including: Obtaining a joint feature vector after splicing the weighted feature vector and the decision result; In the case where the decision result is allowable adjustment, inputting the joint feature vector into a fully connected regression sub-network to obtain the target height and direction to be adjusted for the air suspension; In the case where the decision result is locking adjustment, inputting the joint feature vector into a fully connected classification sub-network to obtain the locking reason.

5. The adjustment method according to claim 2, characterized in that Training the deep learning model in the following manner: Building a neural network model; Training the neural network model based on a training set, updating the parameters of the neural network model to minimize the target loss function, and obtaining a trained deep learning model; Wherein, the target loss function is the sum of a first loss function, a second loss function, and a third loss function, and is used to optimize the decision of the deep learning model; the first loss function is used to optimize the decision of determining whether the air suspension is adjusted or locked; the second loss function is used to optimize the decision of determining the locking reason; the third loss function is used to optimize the decision of determining the adjustment height and direction.

6. The adjustment method according to claim 1, wherein Adjusting the air suspension according to the allowable adjustment instruction includes: Obtaining the height difference between the current height of the air suspension and the target height in the allowable adjustment instruction; According to the height difference and the current driving state of the vehicle, controlling the power and working duration of the vehicle's compressor, and the opening degrees of the vehicle's intake valve and exhaust valve, so as to realize the adjustment of the height and direction of the air suspension; And / or Lock the air suspension in the current state according to the locking adjustment instruction, including: Control the compressor of the vehicle to stop running, and close the intake valve or exhaust valve of the vehicle to lock the air pressure and height of the air suspension in the current state.

7. The adjustment method according to claim 6, characterized in that, It also includes: During the adjustment process of the air suspension, monitor the changes in the vehicle state; In the case of abnormal changes in the vehicle state, suspend the adjustment operation of the air suspension; After the vehicle state stabilizes again, continue the adjustment operation of the air suspension; and / or, Record the vehicle state information when the air suspension is locked and adjusted; after the air suspension is unlocked, according to the recorded vehicle state information, adopt a progressive adjustment strategy to adjust the valve opening and compressor power to make the air suspension return to the state before locking.

8. The adjustment method according to any one of claims 1 to 7, characterized in that, It also includes: During the adjustment process of the air suspension, feedback the adjustment state of the air suspension to the user through the user interface; and / or, During the locking process of the air suspension, feedback the locking reason to the user through the user interface.

9. An adjustment device for a vehicle air suspension, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to execute the adjustment method for the vehicle air suspension according to any one of claims 1 to 8 when running the program instructions.

10. A vehicle, characterized in that, Including: The vehicle body is provided with an air suspension; The adjustment device for the vehicle air suspension according to claim 9 is installed on the vehicle body and connected to the air suspension.

Images