Intelligent data-driven adaptive joint observation method and apparatus, and vehicle and medium

Abstract

An intelligent data-driven adaptive joint observation method and apparatus, and a vehicle and a medium. The method comprises: acquiring a yaw rate signal, wheel speed signals, a steering angle signal, a longitudinal acceleration signal, a lateral acceleration signal, driving torque signals and braking torque signals of a vehicle, using one or more signals as data features, and using data features within preset time step sizes to form first to third input features (S101); compensating for wheel speeds on the basis of a yaw rate and a steering angle (S102); calculating a longitudinal reference vehicle speed on the basis of the compensated wheel speeds and a longitudinal acceleration (S103); estimating a lateral reference vehicle speed measurement value on the basis of the first input feature, and respectively estimating a noise covariance of a longitudinal reference vehicle speed measurement value and a noise covariance of the lateral reference vehicle speed measurement value on the basis of the second input feature and the third input feature (S104); and determining a joint observation weight of the longitudinal reference vehicle speed measurement value and lateral reference vehicle speed measurement value on the basis of the noise covariances, and performing joint observation on a longitudinal reference vehicle speed estimation value and a lateral reference vehicle speed estimation value on the basis of the joint observation weight, so as to obtain a vehicle body sideslip angle estimation value (S105). Thus, the problems in the related art such as the applicability and robustness being relatively poor due to the limitations of an MBD method and a DDD method, the accuracy of an observation result being relatively low due to an inaccurate measured value, and the application being limited due to the requirement for a large amount of data and the presence of a black box feature are solved.

Description

Data intelligence-driven adaptive joint observation method, device, vehicle and medium

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application is based on the Chinese patent application with application number 202410017205.4 and application date on January 5, 2024, and claims the priority of the Chinese patent application. The entire content of the Chinese patent application is hereby introduced into this application as a reference. Technical Field

[0003] The present application relates to the field of vehicle dynamics and control technology, and in particular to a data intelligence-driven adaptive joint observation method, device, vehicle, and medium. Background Art

[0004] Key vehicle dynamics states primarily include longitudinal reference speed and side slip angle. Based on different principles, observation methods for these key vehicle dynamics states are categorized into two types: MBD (Model-Based Design) and DDD (Data-Driven Design).

[0005] MBD mainly constructs key state observers based on vehicle dynamics or kinematic models; DDD mainly uses real vehicle test data to train neural networks for key state observations. After the input signal is processed by the neural network, the observation results can be directly output.

[0006] Among related technologies, MBD methods based on vehicle dynamics typically require additional parameters such as tire forces to be known, and place high demands on the accuracy of vehicle dynamics and tire mechanics models, resulting in poor applicability. MBD methods based on vehicle kinematics can employ relatively simple kinematic equations, but their design rules often rely heavily on human experience, resulting in poor robustness. The DDD method overcomes some of the limitations of the MBD method based on vehicle dynamics when entering the nonlinear region, but it cannot process random noise in sensor signals, resulting in low accuracy in observations. Furthermore, the DDD method typically requires large amounts of data and exhibits black-box characteristics, limiting its scope of application.

[0007] Summary of the Invention

[0008] The present application provides a data intelligence-driven adaptive joint observation method, device, vehicle and medium to address the problems of the MBD method and DDD method in related technologies, such as limitations leading to poor applicability and robustness, inaccurate measurement values ​​leading to low accuracy of observation results, and the requirement of large amounts of data and black-box characteristics leading to limited application.

[0009] The first aspect of the present application provides an adaptive joint observation method driven by data intelligence, comprising the following steps: acquiring a vehicle's yaw rate signal, wheel speed signal, steering angle signal, lateral and longitudinal acceleration signals, and driving and braking torque signal, and using one or more signals as data features, and forming the data features of a preset time step into first to third input features respectively; compensating the wheel speed according to the yaw rate and the steering angle; calculating a longitudinal reference vehicle speed measurement value according to the compensated wheel speed and longitudinal acceleration; estimating a lateral reference vehicle speed measurement value according to the first input feature, and estimating the noise covariance of the longitudinal reference vehicle speed measurement value and the lateral reference vehicle speed measurement value according to the second input feature and the third input feature respectively; determining a joint observation weight of the longitudinal reference vehicle speed measurement value and the lateral vehicle speed value based on the noise covariance, and jointly observing the longitudinal reference vehicle speed estimation value and the lateral reference vehicle speed estimation value based on the joint observation weight, thereby obtaining a vehicle body sideslip angle estimation value.

[0010] Optionally, the wheel speed compensation formula is:

[0011] Among them, W ij is the wheel speed signal, ij∈{fl,fr,rl,rr}; W ij c is the compensated wheel speed signal; is the yaw rate signal; δ is the front wheel turning angle signal; k f 、k r They are the track widths of the front and rear axles respectively.

[0012] Optionally, the calculating of the longitudinal reference vehicle speed measurement value based on the compensated wheel speed and the longitudinal acceleration includes: obtaining an upper threshold value and a lower threshold value of the longitudinal acceleration; if the longitudinal acceleration is greater than or equal to the lower threshold value and less than or equal to the upper threshold value, taking the average of the compensated wheel speeds as the longitudinal reference vehicle speed measurement value; if the longitudinal acceleration is less than the lower threshold value, taking the maximum value of the compensated wheel speed and the current longitudinal reference vehicle speed measurement value as the longitudinal reference vehicle speed value when the longitudinal acceleration is greater than the derivative of the current longitudinal reference vehicle speed estimate; in the longitudinal When the acceleration is less than or equal to the derivative of the current longitudinal reference vehicle speed estimate, the maximum value of the compensated wheel speeds is used as the longitudinal reference vehicle speed measurement value; if the longitudinal acceleration is greater than the upper limit threshold, then when the longitudinal acceleration is less than the derivative of the current longitudinal reference vehicle speed estimate, the minimum value between the compensated wheel speed and the current longitudinal reference vehicle speed measurement value is used as the longitudinal reference vehicle speed measurement value; when the longitudinal acceleration is greater than or equal to the derivative of the current longitudinal reference vehicle speed estimate, the minimum value of the compensated wheel speeds is used as the longitudinal reference vehicle speed measurement value.

[0013] Optionally, before calculating the longitudinal reference vehicle speed measurement value based on the compensated wheel speed and longitudinal acceleration, it also includes: obtaining an estimated value of the longitudinal reference vehicle speed at the previous moment; calculating a change in the estimated value based on the time difference between the current moment and the previous moment and the longitudinal acceleration; and calculating the longitudinal reference vehicle speed measurement value at the current moment based on the estimated value of the longitudinal reference vehicle speed at the previous moment and the change in the estimated value.

[0014] Optionally, the calculation formula for the lateral reference vehicle speed and noise covariance is:

[0015] Among them, R x N Expressed as the noise covariance of the longitudinal reference vehicle speed measurement, R y N Expressed as the noise covariance of the lateral reference vehicle speed measurement value, V y N represents the lateral reference vehicle speed measurement value, k represents the current sampling time, X1, X2 and X3 represent different input features, and the input feature refers to the data feature time series within the historical preset time step, f y,vel Refers to the nonlinear fitting function of the lateral reference vehicle speed estimation module, f x,cov Refers to the nonlinear fitting function of the noise covariance estimation module of the longitudinal reference vehicle speed measurement value, f y,cov Refers to the nonlinear fitting function of the noise covariance estimation module of the measured value of the lateral reference vehicle speed.

[0016] Optionally, the lateral reference vehicle speed and noise covariance are fitted using a neural network structure with a similar structure, wherein the neural network structure includes a long-short time neural network layer, a fully connected layer, a batch normalization layer and a multi-layer perceptron processing layer.

[0017] Optionally, determining the joint observation weight of the longitudinal reference vehicle speed measurement value and the lateral reference vehicle speed measurement value based on the noise covariance includes: if the noise covariance increases, reducing the joint observation weight of the longitudinal reference vehicle speed measurement value; if the noise covariance decreases, increasing the joint observation weight of the longitudinal reference vehicle speed measurement value.

[0018] According to a second aspect of the present application, an adaptive joint observation device driven by data intelligence is provided, comprising: a data storage module for acquiring a vehicle's yaw rate signal, a wheel speed signal, a steering angle signal, a lateral and longitudinal acceleration signal, and a driving and braking torque signal, and using one or more signals as data features, and using the data features of a preset time step to form first to third input features, respectively; a yaw compensation module for compensating the wheel speed according to the yaw rate and the steering angle; a wheel speed signal processing module for calculating a longitudinal reference vehicle speed measurement value according to the compensated wheel speed and longitudinal acceleration; a data-driven estimation module for estimating a lateral reference vehicle speed measurement value according to a first input feature, and estimating the noise covariance of the longitudinal reference vehicle speed measurement value and the lateral reference vehicle speed value according to a second input feature and a third input feature, respectively; and an adaptive joint observation module for determining a joint observation weight of the longitudinal reference vehicle speed measurement value and the lateral vehicle speed value based on the noise covariance, and performing a joint observation on the longitudinal reference vehicle speed estimation value and the lateral reference vehicle speed estimation value based on the joint observation weight, thereby obtaining a vehicle body sideslip angle estimation value.

[0019] Optionally, the wheel speed compensation formula is:

[0020] Among them, W ij is the wheel speed signal, ij∈{fl,fr,rl,rr}; W ij c is the compensated wheel speed signal; is the yaw rate signal; δ is the front wheel turning angle signal; k f 、k r They are the track widths of the front and rear axles respectively.

[0021] Optionally, the wheel speed signal processing module is further used to: obtain an upper threshold and a lower threshold of the longitudinal acceleration; if the longitudinal acceleration is greater than or equal to the lower threshold and less than or equal to the upper threshold, use the average of the compensated wheel speeds as the longitudinal reference vehicle speed measurement value; if the longitudinal acceleration is less than the lower threshold, when the longitudinal acceleration is greater than the derivative of the longitudinal reference vehicle speed estimate at the current moment, use the maximum value between the compensated wheel speed and the longitudinal reference vehicle speed measurement value at the current moment as the longitudinal reference vehicle speed measurement value; when the longitudinal acceleration is less than or equal to the derivative of the longitudinal reference vehicle speed estimate at the current moment, When the derivative of the longitudinal reference vehicle speed estimate at the current moment is taken, the maximum value of the compensated wheel speeds is taken as the longitudinal reference vehicle speed measurement value; if the longitudinal acceleration is greater than the upper limit threshold, then when the longitudinal acceleration is less than the derivative of the longitudinal reference vehicle speed estimate at the current moment, the minimum value of the compensated wheel speed and the longitudinal reference vehicle speed measurement value at the current moment is taken as the longitudinal reference vehicle speed measurement value; when the longitudinal acceleration is greater than or equal to the derivative of the longitudinal reference vehicle speed estimate at the current moment, the minimum value of the compensated wheel speeds is taken as the longitudinal reference vehicle speed measurement value.

[0022] Optionally, it also includes: an acquisition module, used to obtain an estimated value of the longitudinal reference vehicle speed at the previous moment before calculating the longitudinal reference vehicle speed measurement value based on the compensated wheel speed and longitudinal acceleration; calculate the estimated value change based on the time difference between the current moment and the previous moment and the longitudinal acceleration; calculate the longitudinal reference vehicle speed measurement value at the current moment based on the estimated value of the longitudinal reference vehicle speed at the previous moment and the estimated value change.

[0023] Optionally, the calculation formula for the lateral reference vehicle speed and noise covariance is:

[0024] Among them, R x N Expressed as the longitudinal reference vehicle speed measurement noise covariance, R y N Expressed as the lateral reference vehicle speed noise covariance, V y N It is represented as the estimated value of the lateral reference vehicle speed measurement value, k represents the current sampling time, X1, X2 and X3 represent different input features, and the input feature refers to the data feature time series within the historical preset time step, f y,vel Refers to the nonlinear fitting function of the lateral reference vehicle speed estimation module, f x,cov Refers to the nonlinear fitting function of the longitudinal reference vehicle speed measurement noise covariance estimation module, f y,cov Refers to the nonlinear fitting function of the lateral reference vehicle speed measurement noise covariance estimation module.

[0025] Optionally, the lateral reference vehicle speed and noise covariance are fitted using a neural network structure with a similar structure, wherein the neural network structure includes a long-short time neural network layer, a fully connected layer, a batch normalization layer and a multi-layer perceptron processing layer.

[0026] Optionally, the adaptive joint observation module is further configured to: reduce the joint observation weight of the longitudinal reference vehicle speed if the noise covariance increases; and increase the joint observation weight of the longitudinal reference vehicle speed if the noise covariance decreases.

[0027] The third aspect of the present application provides a vehicle, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the data intelligence-driven adaptive joint observation method as described in the above embodiment.

[0028] The fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the data intelligence-driven adaptive joint observation method as described in the above embodiments.

[0029] Therefore, this application has at least the following beneficial effects:

[0030] (1) The embodiment of the present application can quickly and conveniently obtain the measurement value of the longitudinal reference vehicle speed, and can ensure that the estimation result has good accuracy under acceleration, deceleration and steady-state conditions.

[0031] (2) The embodiment of the present application can directly fit the measured value of the lateral reference vehicle speed according to the data characteristic time series, avoiding the problem of solving a complex nonlinear vehicle dynamics model, having high applicability, and the estimation result has good accuracy; the noise covariance of the measured values ​​of the longitudinal reference vehicle speed and the lateral reference vehicle speed can be fitted respectively according to the data characteristic time series, which intuitively reflects the accuracy of the measured values.

[0032] (3) The embodiment of the present application takes into account the uncertainty of the measurement values. By estimating the noise covariance of the measurement values ​​of the longitudinal reference vehicle speed and the lateral reference vehicle speed respectively through a data-driven intelligent algorithm, the weights of the key state quantities of the vehicle dynamics can be adaptively adjusted, thereby enhancing the robustness of the joint observation method.

[0033] (4) The data intelligence-driven key state adaptive joint observation method constructed in the embodiment of the present application can be applied to various powertrain models. Compared with traditional observation algorithms, it has the characteristics of dataization, adaptation and high precision.

[0034] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0036] FIG1 is a flow chart of a data intelligence-driven adaptive joint observation method provided according to an embodiment of the present application;

[0037] FIG2 is a processing diagram of a wheel speed signal processing module according to an embodiment of the present application;

[0038] FIG3 is a diagram of an architecture for adaptive joint observation of key vehicle dynamics states driven by data intelligence according to an embodiment of the present application;

[0039] FIG4 is a block diagram of a data intelligence-driven adaptive joint observation device according to an embodiment of the present application;

[0040] FIG5 is a schematic structural diagram of a vehicle according to an embodiment of the present application. DETAILED DESCRIPTION

[0041] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0042] At present, the observation methods of key states of vehicle dynamics are mainly divided into two categories: MBD and DDD. Specifically:

[0043] (1) MBD mainly constructs observers for key states based on vehicle dynamics or kinematic models. MBD methods based on vehicle dynamics can use high-precision observation methods such as KF (Kalman Filter), EKF (Extended Kalman Filter), CKF (Cubature Kalman Filter), and SCKF (Squared-root Cubic Kalman Filter). However, these methods usually require additional parameters such as tire forces to be known, and place high demands on the accuracy of vehicle dynamics and tire mechanics models.

[0044] The MBD method based on vehicle kinematics can use relatively simple kinematic equations. Usually, only wheel speed signals and acceleration signals are needed to achieve high-precision observation. If the robustness of the observation method needs to be improved, the KF-based MMSM (Multi-mode Switching Method) and MSFM (Multi-sensor Fusion Method) can be used, but the design of related rules relies heavily on experience.

[0045] (2) DDD primarily uses real-vehicle test data to train a neural network for key state observations. After the input signal is processed by the neural network, the observation results can be directly output. Neural networks can achieve high-precision fitting of complex nonlinear systems, to some extent overcoming the limitations of MBD based on vehicle dynamics when entering the nonlinear region.

[0046] Commonly used neural networks include FFNNs (Feed-forward Neural Networks) such as FCNN (Fully Connected Neural Network) and RBNN (Radial Basis Neural Network), and RNNs (Recurrent Neural Networks) such as GRUNN (Gate Recurrent Unit Neural Network) and LSTMNN (Long Short-term Memory Neural Network). FFNNs can reflect nonlinear relationships between inputs and outputs but cannot capture temporal information. RNNs, on the other hand, are more suitable for modeling inertial systems because they can map temporal inputs to temporal outputs. Although neural networks have universal fitting properties, their large data requirements and black-box nature limit the widespread application of data-driven approaches. Furthermore, while these approaches circumvent the difficulties of directly solving nonlinear models, they cannot handle random noise in sensor signals.

[0047] The following describes the data intelligent driven adaptive joint observation method, device, vehicle and medium of the embodiment of the present application with reference to the accompanying drawings. Specifically, Figure 1 is a flow chart of a data intelligent driven adaptive joint observation method provided by the embodiment of the present application.

[0048] As shown in Figure 1, the data intelligence-driven adaptive joint observation method includes the following steps:

[0049] In step S101, the vehicle's yaw rate signal, wheel speed signal, steering angle signal, lateral and longitudinal acceleration signals, and driving and braking torque signal are obtained, and one or more signals are used as data features, and the data features of the preset time step are respectively composed of the first to third input features.

[0050] The preset time step may be set according to actual conditions, for example, 10 time steps, without specific limitation.

[0051] It can be understood that the embodiments of the present application can obtain the vehicle's yaw angular velocity signal, wheel speed signal, steering angle signal, lateral and longitudinal acceleration signal and driving braking torque signal through various sensors, etc., use the signals as data features, and use the data features of set time steps to form the first to third input features, based on the above signals as input variables for joint observation, so as to facilitate the subsequent calculation of the longitudinal reference vehicle speed and the vehicle body sideslip angle.

[0052] It should be noted that the first input feature may include the wheel speed of all wheels, the longitudinal acceleration of the vehicle body, the driving torque of all wheels, the braking torque of all wheels, the front wheel angle, and the yaw rate of the vehicle body. The second input feature and the third input feature may both include the longitudinal acceleration of the vehicle body, the driving torque of all wheels, the braking torque of all wheels, the front wheel angle, and the yaw rate of the vehicle body. The three estimation modules, namely, the lateral reference speed, the longitudinal reference speed noise covariance, and the lateral reference speed noise covariance, may adopt different design methods according to actual needs. The input features selected by different design methods are also different, and there is no specific limitation.

[0053] In step S102 , the wheel speed is compensated according to the yaw rate and the steering angle.

[0054] Among them, the compensation formula of wheel speed is:

[0055] Among them, W ij is the wheel speed signal, ij∈{fl,fr,rl,rr}; W ij c is the compensated wheel speed signal; is the yaw rate signal; δ is the front wheel turning angle signal; k f 、k r They are the track widths of the front and rear axles respectively.

[0056] It is understandable that the embodiment of the present application can compensate for the wheel speed through the yaw rate and the turning angle, and calculate the wheel speed values ​​of different wheels of the vehicle to facilitate the subsequent calculation of the longitudinal reference vehicle speed measurement value.

[0057] In step S103 , a longitudinal reference vehicle speed measurement value is calculated based on the compensated wheel speed and longitudinal acceleration.

[0058] It is understandable that the embodiment of the present application can calculate the longitudinal reference vehicle speed measurement value based on the compensated wheel speed and longitudinal acceleration, thereby ensuring the accuracy of the estimated value of the longitudinal reference vehicle speed.

[0059] In the embodiment of the present application, as shown in FIG2 , the longitudinal reference vehicle speed measurement value is calculated based on the compensated wheel speed and the longitudinal acceleration, including: obtaining an upper threshold and a lower threshold of the longitudinal acceleration; if the longitudinal acceleration is greater than or equal to the lower threshold and less than or equal to the upper threshold, then the average of the compensated wheel speeds is used as the longitudinal reference vehicle speed measurement value; if the longitudinal acceleration is less than the lower threshold, then when the longitudinal acceleration is greater than the derivative of the current longitudinal reference vehicle speed estimate, the maximum value of the compensated wheel speed and the current longitudinal reference vehicle speed measurement value is used as the longitudinal reference vehicle speed. measurement value; when the longitudinal acceleration is less than the derivative of the longitudinal reference vehicle speed estimate at the current moment, the maximum value of the compensated wheel speeds is used as the longitudinal reference vehicle speed measurement value; if the longitudinal acceleration is greater than the upper limit threshold, then when the longitudinal acceleration is less than the derivative of the longitudinal reference vehicle speed estimate at the current moment, the minimum value of the compensated wheel speeds and the longitudinal reference vehicle speed measurement value at the current moment is used as the longitudinal reference vehicle speed measurement value; when the longitudinal acceleration is greater than or equal to the derivative of the longitudinal reference vehicle speed estimate at the current moment, the minimum value of the compensated wheel speeds is used as the longitudinal reference vehicle speed measurement value.

[0060] Among them, the upper threshold and the lower threshold can be set according to actual needs without specific restrictions.

[0061] It can be understood that the embodiment of the present application analyzes and calculates three types of operating conditions, namely acceleration, deceleration and steady state, based on kinematic rules, to improve the problem of inaccurate measurement of the longitudinal reference vehicle speed under large slip rates and enhance the accuracy of the longitudinal reference vehicle speed.

[0062] It should be noted that under steady-state conditions, W ij c It is very close to the longitudinal real speed. However, when the vehicle accelerates or decelerates suddenly, the tire will have a large slip rate. The slip phenomenon will be more obvious on the low adhesion road. Therefore, this method is based on the kinematic rules and is aimed at acceleration (A x >T Hup )、Deceleration(A x <TH low ) and steady state (TH low ≤A x ≤TH up ) and other three types of working conditions, which can improve the problem of inaccurate measurement of longitudinal reference speed under large slip rate, as shown in Figure 2. x Represents the longitudinal acceleration signal output by the IMU; T Hup , TH lowRespectively represent the upper and lower thresholds of longitudinal acceleration; V -1 Indicates the estimated value of the longitudinal reference speed at the previous moment; Represents the estimated value of the longitudinal reference speed at the current moment, and is used as the measured value of the longitudinal reference speed in the adaptive joint observation module; express The derivative of , which is the estimated value of the longitudinal acceleration at the current moment.

[0063] Specifically, the current vehicle operating condition information is obtained to determine whether the current vehicle operating condition information belongs to the acceleration operating condition. If it belongs to the acceleration operating condition (A x >T Hup ), then continue to judge whether the derivative of the estimated value of the current longitudinal reference vehicle speed is greater than the longitudinal acceleration signal output by the IMU. If the derivative of the estimated value of the current longitudinal reference vehicle speed is greater than the longitudinal acceleration signal output by the IMU, the compensated wheel speed (W ij c ) and the estimated value of the current longitudinal reference vehicle speed measurement value (V -1 +A x The minimum value of the wheel speed after compensation is used as the longitudinal reference vehicle speed. If the derivative of the estimated value of the current longitudinal reference vehicle speed is less than or equal to the longitudinal acceleration signal output by the IMU, the minimum value of the compensated wheel speed is used as the longitudinal reference vehicle speed measurement value.

[0064] If it does not belong to the processing condition, further determine whether it belongs to the deceleration condition. If it belongs to the deceleration condition, continue to determine whether the derivative of the estimated value of the longitudinal reference vehicle speed at the current moment is less than the longitudinal acceleration signal output by the IMU. If the derivative of the estimated value of the longitudinal reference vehicle speed at the current moment is less than the longitudinal acceleration signal output by the IMU, then the maximum value of the compensated wheel speed and the estimated value of the longitudinal reference vehicle speed at the current moment is used as the longitudinal reference vehicle speed measurement value. If the derivative of the estimated value of the longitudinal reference vehicle speed at the current moment is greater than or equal to the longitudinal acceleration signal output by the IMU, then the maximum value of the compensated wheel speed is used as the longitudinal reference vehicle speed measurement value.

[0065] If it is neither a processing condition nor a deceleration condition, it is a steady-state condition. In this case, W ij c The mean value of is very close to the true longitudinal vehicle speed and is basically not affected by the cumulative error of the sensor, which can ensure the accuracy of the estimated value of the longitudinal reference vehicle speed.

[0066] In an embodiment of the present application, before calculating the longitudinal reference vehicle speed measurement value based on the compensated wheel speed and longitudinal acceleration, it also includes: obtaining an estimated value of the longitudinal reference vehicle speed at the previous moment; calculating the estimated value change based on the time difference between the current moment and the previous moment and the longitudinal acceleration; and calculating the longitudinal reference vehicle speed measurement value at the current moment based on the estimated value of the longitudinal reference vehicle speed at the previous moment and the estimated value change.

[0067] It can be understood that the embodiment of the present application can calculate the change in the estimated value based on the time difference between the previous moment and the current moment and the longitudinal acceleration, and calculate the current moment longitudinal reference vehicle speed measurement value through the estimated value of the longitudinal reference vehicle speed at the previous moment and the change in the estimated value.

[0068] It should be noted that the current longitudinal reference speed estimate is calculated as follows: (V -1 +A x Δt), where V -1 Represents the estimated value of the longitudinal reference speed at the previous moment, A x represents the longitudinal acceleration at the current moment, Δt represents the time difference between the previous moment and the current moment, A x Δt represents the change in the estimated longitudinal acceleration between the previous moment and the current moment.

[0069] In step S104 , a lateral reference vehicle speed measurement value is estimated according to the first input feature, and noise covariances of the longitudinal reference vehicle speed measurement value and the lateral reference vehicle speed measurement value are estimated according to the second input feature and the third input feature, respectively.

[0070] The calculation formula of the lateral reference speed and noise covariance is:

[0071] Among them, R x N Expressed as the longitudinal reference vehicle speed measurement noise covariance, R y N Expressed as the noise covariance of the lateral reference vehicle speed measurement value, V y N is represented by the lateral reference vehicle speed measurement value, k represents the current sampling time, X1, X2 and X3 represent different input features, and the input feature refers to the data feature time series within the historical preset time step, f y,vel Refers to the nonlinear fitting function of the lateral reference vehicle speed estimation module, f x,cov Refers to the nonlinear fitting function of the longitudinal reference vehicle speed measurement noise covariance estimation module, f y,cov Refers to the nonlinear fitting function of the lateral reference vehicle speed measurement noise covariance estimation module.

[0072] It can be understood that the embodiments of the present application can determine the uncertainty of the estimated values ​​of the longitudinal reference speed and the lateral reference speed by calculating the lateral reference speed, the noise covariance of the lateral reference speed measurement value, and the noise covariance of the longitudinal reference speed measurement value, so as to constitute the key component of the noise covariance matrix of the related measurement values ​​of the adaptive joint observation, thereby facilitating subsequent joint observation.

[0073] It should be noted that the noise covariance includes the noise covariance of the lateral reference vehicle speed measurement value and the noise covariance of the longitudinal reference vehicle speed measurement value; the historical preset time step can be set according to actual conditions, for example: 10 historical time steps, without specific limitation.

[0074] In an embodiment of the present application, the lateral reference vehicle speed and the noise covariance are fitted using a neural network structure with a similar structure, wherein the neural network structure includes a long-short time neural network layer, a fully connected layer, a batch normalization layer and a multi-layer perceptron processing layer.

[0075] It can be understood that in the embodiment of the present application, the noise covariance of the lateral reference vehicle speed and the longitudinal reference vehicle speed is fitted using a structurally similar neural network structure to extract the characteristic time series of the input data, determine the inherent characteristics based on the differences between the corresponding estimation modules and experience, and establish a corresponding relationship between input and output. The current state can be predicted based on historical data to improve efficiency.

[0076] Specifically, the noise covariance of the lateral and longitudinal reference speed measurements can be calculated using a similar neural network structure, consisting of two components: a feature extraction algorithm based on an LSTMNN and a post-processing algorithm using an MLP (Multi-layer Perceptron). The LSTMNN is used to extract the input data feature time series (Input Feature 1, Input Feature 2, and Input Feature 3), determine inherent features based on the differences between the corresponding estimation modules and empirically establish a corresponding relationship between input and output.

[0077] Because cars are typical inertial systems, their reference speed doesn't change suddenly, and the current state information is strongly correlated with previous states. Therefore, LSTMNN can predict the current state based on historical data. Following the LSTMNN layer is an MLP post-processing layer consisting of two fully connected layers (FC layers), a batch normalization layer, and a Tanh activation layer. The fully connected layers reduce the dimensionality of the multiple hidden units contained in the LSTMNN output, further integrating the output characteristics of the LSTMNN layer. The batch normalization (quasi-normalization) layer accelerates neural network training and significantly reduces the network's sensitivity to initial values.

[0078] In step S105, a joint observation weight of the longitudinal reference vehicle speed measurement value and the lateral vehicle speed measurement value is determined based on the noise covariance, and the longitudinal reference vehicle speed estimation value and the lateral reference vehicle speed estimation value are jointly observed based on the joint observation weight to obtain a vehicle body sideslip angle estimation value.

[0079] It can be understood that the embodiment of the present application determines the joint observation weight of the longitudinal reference vehicle speed measurement value and the lateral vehicle speed measurement value based on the noise covariance, and jointly observes the longitudinal reference vehicle speed estimation value and the lateral reference vehicle speed estimation value according to the joint observation weight, thereby obtaining the vehicle body sideslip angle estimation value, so that when the accuracy of the input signal value is unstable, the joint observation weight is adaptively adjusted to reduce the impact on the output result of the overall joint observation method, improve the accuracy of the observation result, and enhance the robustness of the joint observation method.

[0080] In an embodiment of the present application, the joint observation weight of the longitudinal reference vehicle speed measurement value and the vehicle body sideslip angle is determined based on the noise covariance, including: if the noise covariance increases, the joint observation weight of the longitudinal reference vehicle speed measurement value is reduced; if the noise covariance decreases, the joint observation weight of the longitudinal reference vehicle speed measurement value is increased.

[0081] It is understandable that the embodiment of the present application can adaptively adjust the joint observation weight of the longitudinal reference vehicle speed according to the noise covariance, thereby enhancing the robustness of the joint observation method.

[0082] According to the data intelligence-driven adaptive joint observation method proposed in the embodiment of the present application, the longitudinal reference vehicle speed is calculated by obtaining one or more signal values, the lateral reference vehicle speed is calculated based on the lateral acceleration, the noise covariance of the lateral reference vehicle speed measurement value and the longitudinal reference vehicle speed measurement value are estimated respectively, and the joint observation weight of the longitudinal reference vehicle speed and the lateral vehicle speed measurement value is determined based on the noise covariance. The longitudinal reference vehicle speed estimation value and the lateral reference vehicle speed estimation value are jointly observed according to the joint observation weight, so that when the accuracy of the input signal value is unstable, the joint observation weight is adaptively adjusted to reduce the impact on the output result of the overall joint observation method, improve the accuracy of the observation result, and enhance the robustness of the joint observation method.

[0083] The following describes the data-intelligence-driven adaptive joint observation method of the present application in detail with reference to FIG3 . The data-intelligence-driven adaptive joint observation architecture for key vehicle dynamics states includes five modules: YC (Yaw Compensation), WSP (Wheel Speed ​​Processing), DS (Data Storage), DDE (Data-driven Estimation), and AJO (Adaptive Joint Observation). The specific modules are as follows:

[0084] (1)DS module

[0085] The output signals of the steering angle sensor, wheel speed sensor, IMU, drive system and braking system are obtained as input variables of the joint observation architecture, including but not limited to: vehicle yaw rate, wheel speed, steering angle, lateral and longitudinal acceleration and driving and braking torque.

[0086] (2) YC module

[0087] The present invention constructs a two-layer signal processing architecture to obtain an estimate of the longitudinal reference vehicle speed. The first layer is the YC module, which is responsible for transferring the four wheel speed signals to the vehicle's center of gravity. The relevant equation is as follows:

[0088] Among them, W ij is the wheel speed signal, ij∈{fl,fr,rl,rr}; W ij c is the compensated wheel speed signal; is the yaw rate signal; δ is the front wheel turning angle signal; k f 、k r They are the track widths of the front and rear axles respectively.

[0089] (3)WSP module

[0090] The second layer of the two-layer signal processing architecture is the WSP module. ij c It is very close to the longitudinal real speed. However, when the vehicle accelerates or decelerates suddenly, the tire will have a large slip rate. The slip phenomenon will be more obvious on the low adhesion road. Therefore, this method is based on the kinematic rules and is aimed at acceleration (A x >T Hup )、Deceleration(A x <TH low ) and steady state (TH low ≤A x ≤TH up) and other three types of working conditions, which can improve the problem of inaccurate estimation of longitudinal reference speed under large slip rate, as shown in Figure 2. x Represents the longitudinal acceleration signal output by the IMU (Inertial Measurement Unit); T Hup , TH low Respectively represent the upper and lower thresholds of longitudinal acceleration; V -1 Indicates the estimated value of the longitudinal reference speed at the previous moment; Represents the estimated value of the longitudinal reference speed at the current moment, and is used as the measured value of the longitudinal reference speed in the adaptive joint observation module; express The derivative of is the estimated value of the longitudinal acceleration at the current moment. Under steady-state conditions, W ij c The mean value of is very close to the true longitudinal vehicle speed and is basically not affected by the cumulative error of the sensor, which can ensure the accuracy of the estimated value of the longitudinal reference vehicle speed.

[0091] (5)DDE module

[0092] The DDE module includes three submodules: longitudinal reference speed noise covariance estimation (RxNet), lateral reference speed estimation (VyNet), and lateral reference speed noise covariance estimation (RyNet). VyNet is used to obtain the estimated value of the lateral reference speed. RxNet and RyNet are used to obtain the uncertainty of the estimated values ​​of the longitudinal reference speed and the lateral reference speed, respectively, including the longitudinal reference speed noise covariance RxN and the lateral reference speed noise covariance RyN. RxN and RyN will serve as key components of the noise covariance matrix of the relevant measurement values ​​in the AJO module.

[0093] The three submodules employ similar neural network structures, each consisting of a LSTMNN-based feature extraction algorithm and an MLP post-processing algorithm. The LSTMNN extracts the inherent features of the input data feature time series (input feature 1, input feature 2, and input feature 3, determined empirically based on the differences between the corresponding estimation modules) and establishes a corresponding relationship between input and output.

[0094] Because cars are typical inertial systems, their reference speed doesn't change suddenly, and the current state information is strongly correlated with previous states. Therefore, LSTMNN can predict the current state based on historical data. Following the LSTMNN layer is an MLP post-processing layer consisting of two fully connected layers (FCLayers), a BatchNormalization layer, and a Tanh activation layer. The fully connected layers reduce the dimensionality of the multiple hidden units contained in the LSTMNN output, further integrating the output characteristics of the LSTMNN layer. The BatchNormalization layer accelerates neural network training and significantly reduces the network's sensitivity to initial values.

[0095] Based on this, the three submodules involved in the DDE module can be expressed as nonlinear functions:

[0096] Among them, R x N Expressed as the longitudinal reference vehicle speed measurement noise covariance, R y N Expressed as the noise covariance of the lateral reference vehicle speed measurement value, V y N is represented by the lateral reference vehicle speed measurement value, k represents the current sampling time, X1, X2 and X3 represent different input features, and the input feature refers to the data feature time series within the historical preset time step, f y,vel Refers to the nonlinear fitting function of the lateral reference vehicle speed estimation module, f x,cov Refers to the nonlinear fitting function of the longitudinal reference vehicle speed measurement noise covariance estimation module, f y,cov Refers to the nonlinear fitting function of the lateral reference vehicle speed measurement noise covariance estimation module.

[0097] (5)AJO module

[0098] The vehicle kinematics equations are used as the design basis for the AJO module:

[0099] in, and Denote the derivatives of the longitudinal velocity and lateral velocity at the center of mass of the vehicle, V x and V y Denote the longitudinal velocity and lateral velocity at the center of mass of the vehicle, A x and A y Respectively represent the longitudinal acceleration signal and lateral acceleration signal output by IMU; b x and b y Respectively represent the longitudinal and lateral acceleration signal deviations, which are generally constants; ε x and ε yRepresent the random noise of longitudinal and lateral acceleration signals respectively.

[0100] Based on formula (1), the discrete state space equation with sampling time Δt can be obtained: x(k)=Ax(k-1)+Bu(k)+e(k) (2)

[0101] Where k represents the current sampling time, x=(V x ,b x ,V y ,b y ) T ,u=(A x ,A y ) T ,e=(ε x ,0,ε y ,0) T ,

[0102] According to formula (3) (4) and KF algorithm, an AKF (Adaptive Kalman Filter) joint observation method based on LSTMNN is proposed. The algorithm steps are shown in Table 1 below.

[0103] Table 1 AKF joint observation method based on LSTMNN

[0104] For the above AKF joint observation method based on LSTMNN (such as the data-driven estimation module and the vehicle dynamics key state adaptive joint observation module in Figure 3), if the data features X1, X2, and X3 are known, the input state And initialize the system matrix A, control matrix B, measurement matrix C, estimated noise covariance P, and process noise covariance Q. Then, V can be obtained through the three estimation modules of the lateral reference speed, longitudinal reference speed noise covariance, and lateral reference speed noise covariance in the data-driven estimation module. y N (k), R x N (k) and R y N (k), (corresponding to steps 2, 3, and 4 in Table 1), based on R x N (k) and R y N (k) can obtain R(k).

[0105] Assuming that x(k-1|k-1) and u(k) are known, x(k|k-1) can be obtained according to step 6, P(k|k-1) can be obtained according to steps 1 and 7, K(k) can be obtained according to steps 1, 5, and 7, x(k|k) can be obtained according to steps 1, 6, and 8, and P(k|k) can be obtained according to steps 1, 7, and 8. At this point, the state estimation x(k|k) of the longitudinal reference speed and the sideslip angle and their covariance matrix P(k|k) at the current sampling time k are obtained.

[0106] As shown in Table 1, in the LSTMNN-based AKF joint observation method, the process noise covariance Q is generally fixed. Because the AKF gain K is directly related to Q and the measurement noise covariance R, adjusting R will directly affect the weights of state prediction and measurement. Therefore, adaptive adjustment of the weights of key state quantities of vehicle dynamics can be achieved, thereby enhancing the robustness of the joint observation method.

[0107] For example, for the longitudinal reference speed observation, when the tire slips significantly, the WSP module obtains The accuracy decreases, the output value RxN of RxNet increases, and the joint observation method will The weight of is reduced accordingly to reduce its impact on the output results of the joint observation method.

[0108] In summary, the data intelligence-driven adaptive joint observation architecture and application method proposed in this application can overcome the original limitations of MBD and DDD to achieve high-precision, adaptive joint observation of key vehicle dynamics states (including longitudinal reference vehicle speed and vehicle side slip angle).

[0109] Next, a data intelligence-driven adaptive joint observation device proposed in an embodiment of the present application will be described with reference to the accompanying drawings.

[0110] FIG4 is a block diagram of a data intelligence-driven adaptive joint observation device according to an embodiment of the present application.

[0111] As shown in FIG4 , the data intelligent driven adaptive joint observation device 10 includes: a data storage module 100 , a yaw compensation module 200 , a wheel speed signal processing module 300 , a data driven estimation module 400 and an adaptive joint observation module 500 .

[0112] Among them, the data storage module 100 is used to obtain the vehicle's yaw rate signal, wheel speed signal, steering angle signal, lateral and longitudinal acceleration signals, and driving and braking torque signal, and use one or more signals as data features, and use the data features of preset time steps to form first to third input features respectively; the yaw compensation module 200 is used to compensate the wheel speed according to the yaw rate and steering angle; the wheel speed signal processing module 300 is used to calculate the longitudinal reference vehicle speed based on the compensated wheel speed and longitudinal acceleration; the data-driven estimation module 400 is used to estimate the lateral reference vehicle speed measurement value based on the first input feature, and estimate the noise covariance of the longitudinal reference vehicle speed measurement value and the lateral reference vehicle speed measurement value based on the second input feature and the third input feature respectively; the adaptive joint observation module 500 is used to determine the joint observation weight of the longitudinal reference vehicle speed measurement value and the lateral reference vehicle speed value based on the noise covariance, and jointly observe the longitudinal reference vehicle speed estimation value and the lateral reference vehicle speed estimation value based on the joint observation weight to obtain the vehicle body sideslip angle estimation value.

[0113] In the embodiment of the present application, the wheel speed compensation formula is:

[0114] Among them, W ij is the wheel speed signal, ij∈{fl,fr,rl,rr}; W ij c is the compensated wheel speed signal; is the yaw rate signal; δ is the front wheel turning angle signal; k f 、k r They are the track widths of the front and rear axles respectively.

[0115] In the embodiment of the present application, the wheel speed signal processing module 300 is further used to: obtain an upper threshold and a lower threshold of the longitudinal acceleration; if the longitudinal acceleration is greater than or equal to the lower threshold and less than or equal to the upper threshold, use the average of the compensated wheel speeds as the longitudinal reference vehicle speed measurement value; if the longitudinal acceleration is less than the lower threshold, then when the longitudinal acceleration is greater than the derivative of the current longitudinal reference vehicle speed estimate, use the maximum of the compensated wheel speeds and the current longitudinal reference vehicle speed measurement value as the longitudinal reference vehicle speed measurement value; when the longitudinal acceleration is less than or equal to the derivative of the current longitudinal reference vehicle speed estimate, use the maximum of the compensated wheel speeds as the longitudinal reference vehicle speed measurement value; if the longitudinal acceleration is greater than the upper threshold, then when the longitudinal acceleration is less than the derivative of the current longitudinal reference vehicle speed estimate, use the minimum of the compensated wheel speeds and the current longitudinal reference vehicle speed measurement value as the longitudinal reference vehicle speed measurement value; when the longitudinal acceleration is less than or equal to the derivative of the current longitudinal reference vehicle speed estimate, use the minimum of the compensated wheel speeds as the longitudinal reference vehicle speed measurement value.

[0116] In an embodiment of the present application, it also includes: an acquisition module, which is used to obtain an estimated value of the longitudinal reference vehicle speed at the previous moment before calculating the longitudinal reference vehicle speed measurement value based on the compensated wheel speed and longitudinal acceleration; calculate the estimated value change based on the time difference between the current moment and the previous moment and the longitudinal acceleration; and calculate the longitudinal reference vehicle speed measurement value at the current moment based on the estimated value of the longitudinal reference vehicle speed at the previous moment and the estimated value change.

[0117] In the embodiment of the present application, the calculation formula of the lateral reference vehicle speed and the noise covariance is:

[0118] Among them, R x N Expressed as the longitudinal reference vehicle speed measurement noise covariance, R y N Expressed as the noise covariance of the lateral reference vehicle speed measurement value, V y N is represented by the lateral reference vehicle speed measurement value, k represents the current sampling time, X1, X2 and X3 represent different input features, and the input feature refers to the data feature time series within the historical preset time step, f y,vel Refers to the nonlinear fitting function of the lateral reference vehicle speed estimation module, f x,cov Refers to the nonlinear fitting function of the longitudinal reference vehicle speed measurement noise covariance estimation module, f y,cov Refers to the nonlinear fitting function of the lateral reference vehicle speed measurement noise covariance estimation module.

[0119] In an embodiment of the present application, the lateral reference vehicle speed and the noise covariance are fitted using a neural network structure with a similar structure, wherein the neural network structure includes a long-short time neural network layer, a fully connected layer, a batch normalization layer and a multi-layer perceptron processing layer.

[0120] In the embodiment of the present application, the adaptive joint observation module 500 is further configured to: reduce the joint observation weight of the longitudinal reference vehicle speed measurement value if the noise covariance increases; and increase the joint observation weight of the longitudinal reference vehicle speed measurement value if the noise covariance decreases.

[0121] It should be noted that the above explanation of the embodiment of the data intelligence-driven adaptive joint observation method is also applicable to the data intelligence-driven adaptive joint observation device of this embodiment, and will not be repeated here.

[0122] According to the data intelligence-driven adaptive joint observation device proposed in the embodiment of the present application, the longitudinal reference vehicle speed is calculated by obtaining one or more signal values, the lateral reference vehicle speed is calculated based on the lateral acceleration, the noise covariance of the lateral reference vehicle speed measurement value and the longitudinal reference vehicle speed measurement value are estimated respectively, and the joint observation weight of the longitudinal reference vehicle speed and the lateral vehicle speed measurement value is determined based on the noise covariance. The longitudinal reference vehicle speed estimation value and the lateral reference vehicle speed estimation value are jointly observed according to the joint observation weight, so that when the accuracy of the input signal value is unstable, the joint observation weight is adaptively adjusted to reduce the impact on the output result of the overall joint observation method, improve the accuracy of the observation result, and enhance the robustness of the joint observation method.

[0123] FIG5 is a schematic diagram of the structure of a vehicle provided in an embodiment of the present application. The vehicle may include:

[0124] Memory 501 , processor 502 , and computer programs stored in the memory 501 and executable on the processor 502 .

[0125] When the processor 502 executes the program, the data intelligence-driven adaptive joint observation method provided in the above embodiment is implemented.

[0126] Furthermore, the vehicle further comprises:

[0127] The communication interface 503 is used for communication between the memory 501 and the processor 502 .

[0128] The memory 501 is used to store computer programs that can be run on the processor 502 .

[0129] The memory 501 may include a high-speed RAM (Random Access Memory) memory, and may also include a non-volatile memory, such as at least one disk memory.

[0130] If the memory 501, processor 502, and communication interface 503 are implemented independently, the communication interface 503, memory 501, and processor 502 can be interconnected via a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. Buses can be divided into address buses, data buses, control buses, etc. For ease of illustration, FIG5 shows only one thick line, but this does not mean that there is only one bus or only one type of bus.

[0131] Optionally, in a specific implementation, if the memory 501, the processor 502 and the communication interface 503 are integrated on a chip, the memory 501, the processor 502 and the communication interface 503 can communicate with each other through an internal interface.

[0132] The processor 502 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0133] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned data intelligence-driven adaptive joint observation method.

[0134] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0135] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of this application, "N" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0136] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing a custom logical function or process step, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed in a different order than shown or discussed, including performing functions in a substantially simultaneous manner or in a reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application pertain.

[0137] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiment, the N steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array, a field programmable gate array, etc.

[0138] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0139] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A data intelligence-driven adaptive joint observation method, characterized in that: The following steps are involved: Acquire a yaw rate signal, a wheel speed signal, a steering angle signal, a lateral and longitudinal acceleration signal, and a driving and braking torque signal of the vehicle, and use one or more of the signals as data features, and use the data features of a preset time step to form first to third input features respectively; compensating the wheel speed according to the yaw rate and the steering angle; Calculating the longitudinal reference vehicle speed based on the compensated wheel speed and longitudinal acceleration; estimating a lateral reference vehicle speed measurement value based on the first input feature, and estimating noise covariances of the longitudinal reference vehicle speed measurement value and the lateral reference vehicle speed measurement value based on the second input feature and the third input feature, respectively; A joint observation weight of the longitudinal reference vehicle speed measurement value and the lateral vehicle speed measurement value is determined based on the noise covariance, and the longitudinal reference vehicle speed estimation value and the lateral reference vehicle speed estimation value are jointly observed based on the joint observation weight to obtain a vehicle body sideslip angle estimation value.

2. The data intelligence-driven adaptive joint observation method according to claim 1 is characterized in that: The wheel speed compensation formula is: Among them, W ij is the wheel speed signal, ij∈{fl,fr,rl,rr}; W ij c is the compensated wheel speed signal; is the yaw rate signal; δ is the front wheel turning angle signal; k f 、k r They are the track widths of the front and rear axles respectively.

3. The data intelligence-driven adaptive joint observation method according to claim 1, characterized in that: Calculating the longitudinal reference vehicle speed measurement value based on the compensated wheel speed and longitudinal acceleration includes: Obtaining an upper threshold value and a lower threshold value of the longitudinal acceleration; If the longitudinal acceleration is greater than or equal to the lower threshold and less than or equal to the upper threshold, the average of the compensated wheel speeds is used as the longitudinal reference vehicle speed measurement value; If the longitudinal acceleration is less than the lower threshold, then when the longitudinal acceleration is greater than the derivative of the current longitudinal reference vehicle speed estimate, the maximum of the compensated wheel speed and the current longitudinal reference vehicle speed measurement is used as the longitudinal reference vehicle speed measurement; and when the longitudinal acceleration is less than or equal to the derivative of the current longitudinal reference vehicle speed estimate, the maximum of the compensated wheel speed is used as the longitudinal reference vehicle speed measurement. If the longitudinal acceleration is greater than the upper limit threshold, then when the longitudinal acceleration is less than the derivative of the current longitudinal reference vehicle speed estimate, the minimum of the compensated wheel speed and the current longitudinal reference vehicle speed measurement value is used as the longitudinal reference vehicle speed measurement value; when the longitudinal acceleration is greater than or equal to the derivative of the current longitudinal reference vehicle speed estimate, the minimum of the compensated wheel speeds is used as the longitudinal reference vehicle speed measurement value.

4. The data intelligence-driven adaptive joint observation method according to claim 3 is characterized in that: Before calculating the longitudinal reference vehicle speed measurement value based on the compensated wheel speed and longitudinal acceleration, it also includes: Get the estimated value of the longitudinal reference speed at the previous moment; Calculating an estimated value change based on a time difference between the current moment and the previous moment and the longitudinal acceleration; The current moment longitudinal reference vehicle speed measurement value is calculated according to the estimated value of the previous moment longitudinal reference vehicle speed and the change in the estimated value.

5. The data intelligence-driven adaptive joint observation method according to claim 1, characterized in that: in, The calculation formula of the lateral reference speed and noise covariance is: Among them, R x N Expressed as the longitudinal reference vehicle speed measurement noise covariance, R y N Expressed as the noise covariance of the lateral reference vehicle speed measurement value, V y N is the lateral reference vehicle speed measurement value, k is the current sampling time, X1, X2 and X3 are different input features, which are the data feature time series within the historical preset time step, and f y,vel Refers to the nonlinear fitting function of the lateral reference vehicle speed estimation module, f x,cov Refers to the nonlinear fitting function of the longitudinal reference vehicle speed measurement noise covariance estimation, f y,cov Refers to the nonlinear fitting function for estimating the noise covariance of the lateral reference vehicle speed measurement value.

6. The data intelligence-driven adaptive joint observation method according to claim 1 or 5, characterized in that: The lateral reference vehicle speed and noise covariance are fitted using a neural network structure with a similar structure, wherein the neural network structure includes a long-short time neural network layer, a fully connected layer, a batch normalization layer and a multi-layer perceptron processing layer.

7. The data intelligence-driven adaptive joint observation method according to claim 1 or 5, characterized in that: The determining, based on the noise covariance, a joint observation weight of the longitudinal reference vehicle speed measurement value and the lateral reference vehicle speed measurement value includes: If the noise covariance increases, reducing the joint observation weight of the longitudinal reference vehicle speed measurement value; If the noise covariance decreases, the joint observation weight of the longitudinal reference vehicle speed measurement value is increased.

8. A data intelligence-driven adaptive joint observation device, characterized in that: include: The data storage module is used to obtain the vehicle's yaw rate signal, wheel speed signal, steering angle signal, lateral and longitudinal acceleration signal and driving and braking torque signal, and use one or more signals as data features to store the data features of the preset time step. The features constitute the first to third input features respectively; a yaw compensation module, configured to compensate the wheel speed according to the yaw angular velocity and the turning angle; A wheel speed signal processing module, used to calculate a longitudinal reference vehicle speed measurement value based on the compensated wheel speed and longitudinal acceleration; a data-driven estimation module configured to estimate a lateral reference vehicle speed measurement value based on a first input feature, and to estimate noise covariances of the longitudinal reference vehicle speed measurement value and the lateral reference vehicle speed measurement value based on a second input feature and a third input feature, respectively; an adaptive joint observation module for determining a joint observation weight for the longitudinal reference vehicle speed measurement value and the lateral vehicle speed measurement value based on the noise covariance, and performing a joint observation on the longitudinal reference vehicle speed estimate value and the lateral reference vehicle speed estimate value based on the joint observation weight, thereby obtaining a vehicle body sideslip angle estimate value.

9. A vehicle, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the data intelligence-driven adaptive joint observation method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the data intelligence-driven adaptive joint observation method according to any one of claims 1 to 7.

Images