Vehicle state fault-tolerant estimation method and device based on Bayesian BiLSTM

Abstract

The invention relates to the technical field of intelligent automobile control, in particular to a vehicle state fault-tolerant estimation method and device based on Bayesian BiLSTM, which can solve the problem that a model based on a physical model method is easy to mismatch to a certain extent and cope with the problem of neural network hyper-parameter optimization. The vehicle state fault-tolerant estimation method based on the Bayesian BiLSTM comprises the following steps: establishing a vehicle kinematics model, a dynamics model and related sub-models, wherein the related sub-models comprise a wheel model and a tire model; designing a vehicle state fault-tolerant estimation method based on the kinematic model; designing a vehicle state fault-tolerant estimation method based on the dynamic model; and designing the vehicle state fault-tolerant estimation method based on the Bayesian BiLSTM model according to the two vehicle state fault-tolerant estimation methods.

Description

Technical Field

[0001] The present application relates to the field of intelligent vehicle control technology, and more specifically, to a vehicle state fault-tolerant estimation method and device based on Bayesian BiLSTM. Background Art

[0002] With the rapid development of advanced driver assistance and autonomous driving, vehicle active safety control systems have become a research hotspot. Accurate vehicle state parameters, especially longitudinal velocity, yaw rate, and sideslip angle, are essential components of vehicle stability control systems.

[0003] Longitudinal vehicle speed can be obtained using GPS or the product of wheel speed and radius. However, GPS signal interruption, combined with wheel slip or locking, can render these methods ineffective. Yaw rate can be measured using a gyroscope, but this can fail during actual driving. Side slip angle can be measured using an optical sensor, but this is limited by cost and installation space. Therefore, studying vehicle state parameter estimation in the presence of a yaw rate sensor failure is crucial.

[0004] Although existing estimation methods have made some progress, they are often based on the premise that the yaw rate sensor works normally, and often ignore the fact that the state estimation results based on the physical model can reflect the true trend of the state parameters to a certain extent. Summary of the Invention

[0005] In order to solve the problem of model mismatch based on physical model methods and to address the problem of neural network hyperparameter optimization, this application provides a vehicle state fault-tolerant estimation method and device based on Bayesian BiLSTM.

[0006] The embodiment of the present application is implemented as follows:

[0007] In a first aspect, the present application provides a vehicle state fault-tolerant estimation method based on Bayesian BiLSTM, comprising:

[0008] Establishing a vehicle kinematic model, a dynamic model, and related sub-models, wherein the related sub-models include a wheel model and a tire model;

[0009] Based on the kinematic model, a vehicle state fault-tolerant estimation method is designed;

[0010] Based on the dynamic model, a vehicle state fault-tolerant estimation method is designed;

[0011] According to two vehicle state fault-tolerant estimation methods, a vehicle state fault-tolerant estimation method is designed based on the Bayesian BiLSTM model.

[0012] In a possible implementation, the kinematic model is expressed as:

[0013]

[0014] Among them, a x and a y Represent the longitudinal and lateral acceleration of the vehicle respectively; v x and v y are the longitudinal and lateral velocities of the vehicle, respectively; ω r and β are the yaw rate and sideslip angle of the center of mass, respectively.

[0015] In a possible implementation, the kinetic model is expressed as:

[0016]

[0017] Where m represents the vehicle mass; I z Indicates the yaw moment of the vehicle; F Xfl ,F Xfr ,F Xrl and F Xrr Represents the longitudinal force of the tire; F Yfl ,F Yfr ,F Yrl and F Yrr represents the lateral force of the tire; B is the wheelbase; a and b are the distances from the center of mass of the vehicle to the front and rear axles, respectively; δ f is the front wheel steering angle.

[0018] In one possible implementation, the vehicle state fault-tolerant estimation method designed based on the kinematic model further includes:

[0019] Calculate the longitudinal speed of the four wheel centers according to the TCS and ABS triggering conditions;

[0020] The yaw rate is calculated using the longitudinal velocities of the wheel centers of the front left and right wheels;

[0021] Calculate the vehicle's longitudinal velocity and lateral velocity through integration operations;

[0022] The sideslip angle of the center of mass is calculated based on the lateral acceleration and yaw rate.

[0023] In one possible implementation, calculating the longitudinal speeds of the four wheel centers according to the TCS and ABS triggering conditions may further include:

[0024] When the TCS is triggered, the wheel rotates linearly with acceleration Less than or equal to the vehicle's longitudinal acceleration When the duration t1 of the motion feature is less than or equal to a certain threshold Δt1, the longitudinal velocity of the four wheel centers is estimated as the product of the wheel angular velocity obtained by the wheel speed sensor and the wheel rolling radius, that is: Otherwise, the estimated longitudinal speeds of the four wheel centers are According to the estimated longitudinal velocity of the vehicle's center of mass at the previous moment Center of mass longitudinal acceleration a x (k) and time increment Δt are calculated, that is:

[0025] When ABS is triggered, it is the same as when TCS is triggered;

[0026] When TCS and ABS are not triggered, the longitudinal speed of the four wheel centers is

[0027] In one possible implementation, the vehicle state fault-tolerant estimation method designed based on the dynamic model further includes:

[0028] Initialize the state vector estimate and covariance matrix P0;

[0029] The unscented Kalman filter (UKF) algorithm is used to iteratively update the system state equation and measurement equation. The formula is:

[0030] System state equation:

[0031]

[0032] Measurement equation:

[0033]

[0034] In one possible implementation, the Bayesian BiLSTM model includes:

[0035] The input layer receives estimates of the kinematic and dynamic models;

[0036] The BiLSTM layer captures temporal features through forward and backward LSTM, and the output sequence is the fusion result of the forward hidden sequence and the backward hidden sequence;

[0037] The output layer generates the final estimates of the vehicle state parameters.

[0038] In one possible implementation, the BiLSTM unit state update formula is:

[0039]

[0040] The forward hidden sequence and reverse hidden sequence Fusion gets the output sequence y t , the update process of BiLSTM can be expressed as:

[0041]

[0042] In one possible implementation, the Bayesian optimization process includes:

[0043] A surrogate model of the objective function is constructed using Gaussian process (GP), whose mean and covariance functions are m(u) and k(u,u') respectively;

[0044] The hyperparameter combination is selected based on the expected improvement (EI) acquisition function, and the mathematical expression is:

[0045]

[0046] Among them, L(u * ) represents the current optimal function, and Φ(·) is the probability density function of the standard normal distribution.

[0047] In a second aspect, the present application provides a vehicle state fault-tolerant estimation device based on Bayesian BiLSTM, comprising:

[0048] A model building module, used to build a vehicle kinematic model, a dynamic model and related sub-models, wherein the related sub-models include a wheel model and a tire model;

[0049] A first sub-method design module is used to design a vehicle state fault-tolerant estimation method based on the kinematic model;

[0050] A second sub-method design module is used to design a vehicle state fault-tolerant estimation method based on the dynamic model;

[0051] The main method design module is used to design a vehicle state fault-tolerant estimation method based on two vehicle state fault-tolerant estimation methods and the Bayesian BiLSTM model.

[0052] The technical solution provided by this application can achieve at least the following beneficial effects:

[0053] The present application provides a Bayesian BiLSTM-based vehicle state fault-tolerant estimation method and device, which integrates the physical interpretability of the kinematic-dynamic model and the nonlinear time series regression capability of deep learning to solve the problem of inaccurate state parameter estimation caused by yaw rate sensor failure, improves depth, and combines the bidirectional information capture capability of BiLSTM with the global optimization of Bayesian optimization hyperparameters to improve the model's prediction accuracy and generalization ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0055] Figure 1 This is a flowchart of a vehicle state fault-tolerant estimation method based on Bayesian BiLSTM, shown in an exemplary embodiment of the present application;

[0056] Figure 2 This is a schematic diagram of a specific implementation process of another vehicle state fault-tolerant estimation method based on Bayesian BiLSTM, shown in an exemplary embodiment of the present application;

[0057] Figure 3 Schematic diagram of LSTM and BiLSTM structures shown in an exemplary embodiment of the present application;

[0058] Figure 4 It is a structural diagram of a vehicle state fault-tolerant estimation device based on Bayesian BiLSTM, shown as an exemplary embodiment of the present application.

[0059] Reference numerals:

[0060] 1. Model building module; 2. First sub-method design module; 3. Second sub-method design module; 4. Main method design module. DETAILED DESCRIPTION

[0061] In order to make the purpose, implementation methods and advantages of the present application clearer, the exemplary implementation methods of the present application will be clearly and completely described below in conjunction with the drawings in the exemplary embodiments of the present application. Obviously, the described exemplary embodiments are only part of the embodiments of the present application, not all of the embodiments. It should be understood that the specific embodiments described here are only used to explain the present application and are not used to limit the present application.

[0062] It should be noted that the brief descriptions of terms in this application are only for the purpose of facilitating the understanding of the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise specified, these terms should be understood according to their ordinary and usual meanings.

[0063] In the specification and claims of this application and the accompanying drawings, the terms "first," "second," "third," etc. are used to distinguish similar or similar objects or entities, and are not necessarily intended to limit a particular order or sequence, unless otherwise noted. It should be understood that the terms used in this manner are interchangeable under appropriate circumstances.

[0064] The terms "comprise," "include," and "have," and any variations thereof, are intended to cover but not exclude inclusion; for example, a product or device comprising a list of components is not necessarily limited to all the components expressly listed but may include other components not expressly listed or inherent to such product or device.

[0065] Before explaining the Bayesian BiLSTM-based vehicle state fault-tolerant estimation method provided in the embodiment of the present application, the application scenario and implementation environment of the embodiment of the present application are first introduced.

[0066] With the rapid development of advanced driver assistance and autonomous driving, vehicle active safety control systems have become a research hotspot. Accurate vehicle state parameters, especially longitudinal velocity, yaw rate, and sideslip angle, are essential components of vehicle stability control systems.

[0067] Longitudinal vehicle speed can be obtained using GPS or the product of wheel speed and radius. However, GPS signal interruption, combined with wheel slip or locking, can render these methods ineffective. Yaw rate can be measured using a gyroscope, but this can fail during actual driving. Side slip angle can be measured using an optical sensor, but this is limited by cost and installation space. Therefore, studying vehicle state parameter estimation in the presence of a yaw rate sensor failure is crucial.

[0068] Although existing estimation methods have made some progress, they are often based on the premise that the yaw rate sensor works normally, and often ignore the fact that the state estimation results based on the physical model can reflect the true trend of the state parameters to a certain extent.

[0069] Based on this, this application provides a Bayesian BiLSTM-based vehicle state fault-tolerant estimation method and device, which improves the robustness and accuracy of vehicle state estimation by fusing kinematic models, dynamic models, and Bayesian-optimized bidirectional long short-term memory networks (BiLSTM). Specifically, it includes: establishing a vehicle kinematic model and a dynamic model, respectively designing a kinematic-based fault-tolerant estimation method and a dynamic fault-tolerant estimation method based on unscented Kalman filtering; proposing a Bayesian BiLSTM fusion model, using kinematic and dynamic estimates as prior inputs, combining the BiLSTM's bidirectional time series feature extraction capabilities and Bayesian-optimized hyperparameter global optimization, to improve the model's convergence efficiency and generalization capabilities.

[0070] Next, the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems will be described in detail through embodiments and in conjunction with the accompanying drawings. The various embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, not all of them.

[0071] Figure 1 This is a flowchart of a vehicle state fault-tolerant estimation method based on Bayesian BiLSTM, shown as an exemplary embodiment of the present application.

[0072] In an exemplary embodiment, Figure 1 As shown, a vehicle state fault-tolerant estimation method based on Bayesian BiLSTM is provided. In this embodiment, the method may include the following steps:

[0073] Step S1: establishing a vehicle kinematic model, a dynamic model, and related sub-models, wherein the related sub-models include a wheel model and a tire model;

[0074] Step S2: designing a vehicle state fault-tolerant estimation method based on the kinematic model;

[0075] Step S3: designing a vehicle state fault-tolerant estimation method based on the dynamic model;

[0076] Step S4: According to the two vehicle state fault-tolerant estimation methods, a vehicle state fault-tolerant estimation method is designed based on the Bayesian BiLSTM model.

[0077] Figure 2 This is a schematic diagram of a specific implementation process of another vehicle state fault-tolerant estimation method based on Bayesian BiLSTM shown in an exemplary embodiment of the present application.

[0078] In one possible implementation, Figure 2 As shown in Figure 2, the specific implementation process of the estimation method is as follows:

[0079] Step S1: Establish kinematic model, dynamic model and related models;

[0080] S11: Establish kinematic model:

[0081]

[0082] Among them, a x and a y Represent the longitudinal and lateral acceleration of the vehicle respectively; v x and v y are the longitudinal and lateral velocities of the vehicle, respectively; ω r and β are the yaw rate and sideslip angle of the center of mass, respectively.

[0083] S12: Establishing the kinetic model:

[0084]

[0085] Where m represents the vehicle mass; I z Indicates the yaw moment of the vehicle; F Xfl ,F Xfr ,F Xrl and F Xrr Represents the longitudinal force of the tire; F Yfl ,F Yfr ,F Yrl and F Yrr represents the lateral force of the tire; B is the wheelbase; a and b are the distances from the center of mass of the vehicle to the front and rear axles, respectively; δ f is the front wheel steering angle.

[0086] S13: Build wheel model:

[0087]

[0088] Among them, J i represents the moment of inertia of the wheel, T di and T bi represents the driving torque and braking torque, R represents the rolling radius of the tire, ω i Indicates the angular velocity of the wheel.

[0089] S14: Establish tire model:

[0090] Y=Dsin(Ctan -1 (BX-E(BX-tan -1 BX)));

[0091] Where X represents the sideslip angle or slip rate, and Y represents the tire lateral force or longitudinal force.

[0092] Step S2: Based on the kinematic model, a vehicle state fault-tolerant estimation method is designed.

[0093] S21: Calculate the longitudinal speed of the four wheel centers:

[0094] Case A: TCS is triggered; Case B: ABS is triggered; Case C: Neither TCS nor ABS is triggered.

[0095] Method for estimating the longitudinal velocity of the four-wheel center in case A: When the wheel rotation linear acceleration Less than or equal to the vehicle's longitudinal acceleration When the duration of the motion feature t1 is less than or equal to a certain threshold Δt1, the longitudinal velocity of the four-wheel center is estimated The wheel angular velocity obtained by the wheel speed sensor The product of the wheel rolling radius R is: If the above conditions are not met, the estimated longitudinal speed of the four wheel centers According to the estimated longitudinal velocity of the vehicle's center of mass at the previous moment Center of mass longitudinal acceleration a x (k) and time increment Δt are calculated, that is: If the estimated value of the wheel center longitudinal velocity is greater than the wheel rotation linear velocity, then let

[0096] Method for estimating the longitudinal speed of the four-wheel center in case B: The principle is the same as that in case A.

[0097] Method for estimating the longitudinal velocity of the four-wheel center in case C:

[0098] S22: Calculate the yaw rate:

[0099] Calculate using the front left wheel center longitudinal velocity and the front right wheel center longitudinal velocity:

[0100]

[0101] S23: Calculate the vehicle's longitudinal and lateral speeds:

[0102]

[0103] S24: Calculate the sideslip angle of the center of mass:

[0104]

[0105] Step S3: Design a vehicle state fault-tolerant estimation method based on the dynamic model:

[0106] Initialize the state vector estimate And the covariance matrix P0:

[0107]

[0108] State vector x, measurement vector z, and input u:

[0109]

[0110] System state equation:

[0111]

[0112] Measurement equation:

[0113]

[0114] The Unscented Kalman Filter (UKF) algorithm approximates the probability density distribution of a nonlinear function through the unscented transform (UT), and uses the "sigma point" samples selected by the unscented transform to ensure that their mean and covariance accurately reflect the statistical characteristics of the original state distribution.

[0115] Generate sigma points:

[0116]

[0117] in, represents the Cholesky decomposition of the covariance matrix, is the scaling factor, α1 determines the range of sigma point distribution, usually needs to satisfy α1∈[1e -4 ,1]; α2 is the tuning parameter, usually taken as α2=3-n.

[0118] The mean weight of each sigma point W i (m) and covariance weights W i ( c ) They are:

[0119]

[0120] Among them, α3 represents the parameter that suppresses the errors of high-order terms.

[0121] calculate Prediction Status and the covariance matrix

[0122]

[0123]

[0124] Computing observational predictions The predicted mean of the observations Prediction covariance P zz , cross covariance matrix P xz and the Kalman gain K k+1 .

[0125]

[0126] System state update and covariance matrix update:

[0127]

[0128] Step S4: Design a vehicle state fault-tolerant estimation method based on the Bayesian BiLSTM model.

[0129] Figure 3 This is a schematic diagram of the LSTM and BiLSTM structures shown in an exemplary embodiment of the present application.

[0130] Step S41: Figure 3 As shown, establish a BiLSTM network:

[0131] Input data: Output data:

[0132] LSTM introduces a storage unit with a gating mechanism, which is connected to the input gate i t 、Forget Gate t , output gate o t For the unit state c at the previous moment t-1 and hidden state h t-1 Make dynamic adjustments and output the new unit state c at the current moment t and the new hidden unit h t During temporal processing, each memory cell performs the following iterative update at time step :

[0133] Determine the historical unit state c through the forget gate t-1 Information that should be ignored.

[0134] f t Is to use the current input information x t , the hidden state output h at the previous moment t-1 , forget gate bias term b f , and the weight matrix W f,x and W f,h Obtained and calculated by the activation function σ.

[0135] f t =σ(W f,x x t +W f,h h t-1 +b f );

[0136] Get the input gate value i at time step t t and the new candidate value

[0137] i t =σ(W i,x x t +W i,h h t-1 +b i );

[0138]

[0139] Calculate the new cell state based on the results of the previous steps:

[0140]

[0141] The LSTM layer storage unit output is calculated as follows:

[0142] o t =σ(W o,x x t +W o,h h t-1 +b o );

[0143] h t =o t ⊙tanh(c t );

[0144] The BiLSTM neural network architecture includes an input layer, a forward LTSM layer, a reverse LSTM layer, and an output layer. The input data is processed by the forward LSTM to obtain a forward hidden sequence. Reverse LSTM processing to obtain reverse hidden sequence Then and Fusion gets the output sequence y t The neural network update process can be expressed as:

[0145]

[0146] Among them, W j (j=1,2,3,4,5,6) is the weight coefficient, and b y is the bias vector.

[0147] Step S42: Bayesian algorithm optimizes the hyperparameters of the BiLSTM network;

[0148] Step S42: Bayesian algorithm optimizes the hyperparameters of the BiLSTM network.

[0149] S421: Constructing a probabilistic agent model:

[0150] Bayesian optimization is an efficient global optimization method. Its core is to use Bayes' theorem to approximate the posterior distribution of the black-box function or objective function to be optimized. The next hyperparameter combination to be sampled is then selected based on the posterior distribution, thereby effectively exploring the hyperparameter space within a limited number of iterations. Bayes' theorem is as follows:

[0151]

[0152] Where w is the unknown target function and D is the set of observations. p(w) and p(w|D) are the prior and posterior probabilities, respectively. P(D|w) is the likelihood function, and P(D) is the marginal probability.

[0153] Defined as u1,u2,…,u n A set of hyperparameter combinations, and the corresponding search space is defined as U1,U2,…,U n L(u) is the objective function to be optimized, i.e., the performance u of the neural network model after training on the training set under the hyperparameter configuration. It can be the loss function of the trained model on the validation set or the error rate on the validation set. The mathematical expression is as follows:

[0154] u * =arg u∈U minL(u);

[0155] A surrogate model of the objective function is constructed using Gaussian process (GP) as follows:

[0156]

[0157] Where m(u) and k(u,u') are the mean function and covariance function respectively. The probability density function of any u in L(u) is the normal distribution function. Add a new set of evaluation pairs to (u n+1 ,L(u n+1 )) The historical evaluation set is used and the posterior probability distribution L estimated using the updated covariance matrix n+1 , the expression is as follows:

[0158]

[0159] S422: Sampling strategy based on acquisition function:

[0160] By using the acquisition function to determine the next observation location, the optimal predicted observation attribute can be obtained. The posterior distribution of the optimization objective function can be obtained through the probabilistic surrogate model. The function α(u|D) is obtained by minimizing it, and the next sampling point is calculated. The formula is as follows:

[0161] u n+1 =arg u∈U minα(u|D);

[0162] Expected Improvement (EI) is used as an acquisition function to select the next evaluation point in order to quickly approximate the global optimal solution. The mathematical expression of EI is as follows:

[0163]

[0164] Among them, L(u *) represents the current optimal function, and Φ(·) is the probability density function of the standard normal distribution.

[0165] It should be understood that, although the various steps in the flowcharts involved in the above-described embodiments are displayed in sequence according to the instructions, these steps are not necessarily executed in the order indicated. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0166] Corresponding to the aforementioned embodiment of the vehicle state fault-tolerant estimation method based on Bayesian BiLSTM, adopting the same technical concept, the present application also provides an embodiment of a vehicle state fault-tolerant estimation device based on Bayesian BiLSTM.

[0167] Figure 4 It is a structural diagram of a vehicle state fault-tolerant estimation device based on Bayesian BiLSTM, shown as an exemplary embodiment of the present application.

[0168] In an exemplary embodiment, Figure 4 As shown, the vehicle state fault-tolerant estimation device based on Bayesian BiLSTM includes:

[0169] A model building module, used to build a vehicle kinematic model, a dynamic model and related sub-models, wherein the related sub-models include a wheel model and a tire model;

[0170] A first sub-method design module is used to design a vehicle state fault-tolerant estimation method based on the kinematic model;

[0171] A second sub-method design module is used to design a vehicle state fault-tolerant estimation method based on the dynamic model;

[0172] The main method design module is used to design a vehicle state fault-tolerant estimation method based on two vehicle state fault-tolerant estimation methods and the Bayesian BiLSTM model.

[0173] For the specific limitations of the Bayesian BiLSTM-based vehicle state fault-tolerant estimation device, please refer to the limitations of the Bayesian BiLSTM-based vehicle state fault-tolerant estimation method above, which will not be repeated here. The various modules in the above-mentioned Bayesian BiLSTM-based vehicle state fault-tolerant estimation device can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0174] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0175] The embodiments described above merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A vehicle state fault-tolerant estimation method based on Bayesian BiLSTM, characterized in that: include: Establishing a vehicle kinematic model, a dynamic model, and related sub-models, wherein the related sub-models include a wheel model and a tire model; Based on the kinematic model, a vehicle state fault-tolerant estimation method is designed; Based on the dynamic model, a vehicle state fault-tolerant estimation method is designed; According to two vehicle state fault-tolerant estimation methods, a vehicle state fault-tolerant estimation method is designed based on the Bayesian BiLSTM model.

2. The vehicle state fault-tolerant estimation method based on Bayesian BiLSTM according to claim 1, characterized in that: The kinematic model is expressed as: Among them, a x and a y Represent the longitudinal and lateral acceleration of the vehicle respectively; v x and v y are the longitudinal and lateral velocities of the vehicle, respectively; ω r and β are the yaw rate and sideslip angle of the center of mass, respectively.

3. The vehicle state fault-tolerant estimation method based on Bayesian BiLSTM according to claim 1, characterized in that: The kinetic model is expressed as: Where m represents the vehicle mass; I z Indicates the yaw moment of the vehicle; F Xfl ,F Xfr ,F Xrl and F Xrr Represents the longitudinal force of the tire; F Yfl ,F Yfr ,F Yrl and F Yrr represents the lateral force of the tire; B is the wheelbase; a and b are the distances from the center of mass of the vehicle to the front and rear axles, respectively; δ f is the front wheel steering angle.

4. The vehicle state fault-tolerant estimation method based on Bayesian BiLSTM as claimed in claim 3, characterized in that: Based on the kinematic model, the designed vehicle state fault-tolerant estimation method further includes: Calculate the longitudinal speed of the four wheel centers according to the TCS and ABS triggering conditions; The yaw rate is calculated using the longitudinal velocities of the wheel centers of the front left and right wheels; Calculate the vehicle's longitudinal velocity and lateral velocity through integration operations; The sideslip angle of the center of mass is calculated based on the lateral acceleration and yaw rate.

5. The vehicle state fault-tolerant estimation method based on Bayesian BiLSTM according to claim 1, characterized in that: The calculating of the longitudinal speeds of the four wheel centers according to the TCS and ABS triggering conditions further includes: When TCS is triggered, the wheel rotates linearly with acceleration Less than or equal to the vehicle's longitudinal acceleration When the duration t1 of the motion feature is less than or equal to a certain threshold Δt1, the longitudinal velocity of the four wheel centers is estimated as the product of the wheel angular velocity obtained by the wheel speed sensor and the wheel rolling radius, that is: Otherwise, the estimated longitudinal speeds of the four wheel centers are According to the estimated longitudinal velocity of the vehicle's center of mass at the previous moment Center of mass longitudinal acceleration a x (k) and time increment Δt are calculated, that is: When ABS is triggered, it is the same as when TCS is triggered; When TCS and ABS are not triggered, the longitudinal speed of the four wheel centers is 6. The vehicle state fault-tolerant estimation method based on Bayesian BiLSTM according to claim 5, characterized in that: Based on the dynamic model, the designed vehicle state fault-tolerant estimation method further includes: Initialize the state vector estimate and covariance matrix P0; The unscented Kalman filter (UKF) algorithm is used to iteratively update the system state equation and measurement equation. The formula is: System state equation: Measurement equation:

7. The vehicle state fault-tolerant estimation method based on Bayesian BiLSTM according to claim 1, characterized in that: The Bayesian BiLSTM model includes: The input layer receives estimates of the kinematic and dynamic models; The BiLSTM layer captures temporal features through forward and backward LSTM, and the output sequence is the fusion result of the forward hidden sequence and the backward hidden sequence; The output layer generates the final estimates of the vehicle state parameters.

8. The vehicle state fault-tolerant estimation method based on Bayesian BiLSTM according to claim 1, characterized in that: The unit state update formula of the BiLSTM is: The forward hidden sequence and reverse hidden sequence Fusion gets the output sequence y t , the update process of BiLSTM can be expressed as:

9. The vehicle state fault-tolerant estimation method based on Bayesian BiLSTM according to claim 1, characterized in that: The Bayesian optimization process includes: A surrogate model of the objective function is constructed using Gaussian process (GP), whose mean and covariance functions are m(u) and k(u,u') respectively; The hyperparameter combination is selected based on the expected improvement (EI) acquisition function, and the mathematical expression is: Among them, L(u * ) represents the current optimal function, and Φ(·) is the probability density function of the standard normal distribution.

10. A vehicle state fault-tolerant estimation device based on Bayesian BiLSTM, characterized in that: include: A model building module, used to build a vehicle kinematic model, a dynamic model and related sub-models, wherein the related sub-models include a wheel model and a tire model; A first sub-method design module is used to design a vehicle state fault-tolerant estimation method based on the kinematic model; A second sub-method design module is used to design a vehicle state fault-tolerant estimation method based on the dynamic model; The main method design module is used to design a vehicle state fault-tolerant estimation method based on the Bayesian BiLSTM model.

Images