Drive-by-wire four-wheel independent steering system and roll risk assessment method thereof

Abstract

The invention discloses a drive-by-wire four-wheel independent steering system and a roll risk assessment method thereof, and the method comprises the steps: building a vertical load distribution model, so as to monitor and feed back the time-varying property and nonlinear change of a vertical load in real time; establishing a real-time roll risk assessment model, designing a dynamic compensation algorithm to correct the model, and generating a roll risk prediction value sequence of a future working condition through a time sequence analysis and prediction algorithm; and establishing a roll risk early warning model, and generating different levels of safety early warning signals based on a risk level judgment mechanism. According to the invention, accurate measurement of the vertical load can be realized under the limit working condition, and real-time roll risk early warning is provided in combination with the dynamic behavior of the vehicle, so that the safety and reliability of the intelligent driving system are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of automotive chassis control, and particularly relates to a steer-by-wire four-wheel independent steering system and a roll risk assessment method thereof. Background Art

[0002] With the development of intelligent driving technology and electric vehicles, the steer-by-wire four-wheel independent steering system (4WIS) has shown great potential in improving vehicle maneuverability, stability, and responsiveness. Especially under high-dynamic driving conditions, 4WIS can enhance vehicle handling performance through precise angle control and provide higher safety in complex road environments. At the same time, accurate measurement of vertical load and assessment of roll risk are crucial for improving vehicle driving stability and safety under complex working conditions.

[0003] The steer-by-wire four-wheel independent steering system independently controls the steering angle of each wheel, so that the connection between the wheels and the vehicle body is no longer restricted by the traditional steering mechanism. However, this independent control causes significant changes in the vertical load distribution between the wheels and the ground when the vehicle is facing high-intensity turning, braking, or accelerating, which poses higher requirements for vehicle stability and safety. Especially in working conditions with a high roll risk, how to real-time monitor and evaluate the change of vertical load, and then effectively predict the possible rollover accidents of the vehicle, is a major challenge in current research.

[0004] Currently, most existing research focuses on the establishment of vehicle dynamics models and the optimization of control strategies, especially in the design and control of four-wheel independent steering systems. Although these studies have made certain progress, most of them have not effectively combined vertical load measurement and roll risk assessment. Especially during the dynamic change process of the vehicle, they have not been able to realize real-time assessment and early warning of safety performance under complex working conditions.

[0005] In the existing Chinese invention patent application No. CN201710756283.6, titled "Vehicle rollover warning control method, system, and vehicle implementing the control method", it is proposed to improve the reliability of rollover warning by combining with the actual lateral acceleration. However, this method mainly focuses on the calculation and warning of lateral acceleration and lacks real-time monitoring of vertical load and quantitative assessment of roll risk. In Chinese invention patent application No. CN201611254306.5, titled "A rollover warning system and its warning method", the vertical load of the tire is calculated through multiple sensor modules (such as roll angle sensors, pitch angle sensors, etc.) and the Kalman filtering algorithm, and the roll risk is judged according to whether the vertical load of the tire is zero. This method can real-time monitor the change of the vertical load of the tire, but in terms of roll risk assessment, it mainly conducts hierarchical warning through the signal of the tire leaving the ground and lacks continuous quantitative assessment of roll risk.

[0006] There are still two major problems in the prior art:

[0007] First, although certain progress has been made in vertical load measurement in existing research, most of the work is limited to static tests or simplified dynamic models, and these methods fail to fully consider the non-linear variation of the vertical load distribution under complex dynamic conditions of the vehicle. Specifically, under high-speed driving, sharp turning, or extreme driving conditions, the change of the vertical load is not only jointly affected by vehicle speed, vehicle body attitude, load, and road conditions, but also the interaction between these factors results in a highly time-varying and non-linear characteristic of the spatial distribution of the load. In the prior art, the measurement accuracy and response speed of many mechanical models or sensors cannot meet the requirements of real-time dynamic changes. Especially when the vehicle encounters emergencies or extreme conditions, traditional methods often cannot accurately capture the change of the load distribution in a timely manner. Therefore, how to achieve accurate measurement of the vertical load in a highly complex and dynamically changing driving environment and effectively evaluate its potential impact on vehicle stability and safety remains an important challenge in the current technology.

[0008] Second, existing roll risk assessment methods mostly rely on static analysis or experience-based models, and these methods fail to fully consider the complex behavior of the vehicle in a real dynamic environment. Specifically, existing roll risk assessment models usually ignore the combined effects of factors such as vehicle body attitude, tire lateral force, and suspension system response in dynamic conditions, and these factors show significant dynamic changes when the vehicle makes sharp turns, accelerates suddenly, or encounters an uneven road surface. Existing roll risk assessment methods often rely on simplified assumptions, such as estimating the roll risk through single parameters such as turning radius and vehicle speed, and this method fails to fully reflect the role of multi-variable and non-linear systems in complex driving scenarios. Therefore, how to construct a roll risk assessment method based on real-time dynamic monitoring data, which can evaluate the roll risk of the vehicle in real time through data fusion and high-precision algorithms in a complex and changeable driving environment, has become an important technical problem for improving vehicle dynamic stability and safety. Summary of the Invention

[0009] Aiming at the deficiencies of the above prior art, the purpose of the present invention is to provide a steer-by-wire four-wheel independent steering system and its roll risk assessment method to solve the problems in the prior art of difficult accurate measurement of vertical load changes and real-time assessment of roll risk under complex dynamic conditions. The present invention can achieve accurate measurement of the vertical load under extreme conditions, provide real-time roll risk warnings in combination with the dynamic behavior of the vehicle, thereby improving the safety and reliability of the intelligent driving system.

[0010] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0011] A steer-by-wire four-wheel independent steering system of the present invention includes: a wheel corner module, a vehicle frame, a vehicle controller, and a power supply.

[0012] The wheel corner module is composed of four corner modules: a left front corner module, a right front corner module, a left rear corner module, and a right rear corner module. Each corner module has the same structural design and is respectively installed at the left front, right front, left rear, and right rear of the vehicle frame, hinged to the vehicle frame. The corner module includes: a worm and worm gear reduction mechanism, a steering arm column, a steering angle sensor, a steering arm, a steering column, a braking system assembly, a hub motor, a lower control arm, a tire, a high-voltage wire harness, an upper control arm, a shock absorber, a high-voltage power supply, a mounting bracket, a low-voltage power supply, a low-voltage wire harness, a steering motor controller, a steering motor, a vehicle frame connecting piece, and a reducer; the steering motor controller and the housing of the steering motor are integrally designed. The steering motor controller outputs a signal to control the rotation of the steering motor. The output shaft of the steering motor is connected to the reducer, the reducer is connected to the worm and worm gear reduction mechanism, and the worm and worm gear reduction mechanism is rigidly connected to the steering arm column and transmits the torque of the steering motor to the steering arm; the lower end of the steering arm column is rigidly connected to the short side of the steering arm. The steering angle sensor is installed on the steering arm column to collect the steering angle signal of the steering arm column in real time and send the steering angle signal to the steering motor controller; the steering arm is connected to the steering column through a shock absorber, an upper control arm, and a lower control arm. The upper end of the steering column is respectively connected to the upper control arm and the lower end of the shock absorber, and the lower end of the steering column is connected to the upper end of the lower control arm. The upper end of the upper control arm, the upper end of the shock absorber, and the lower end of the lower control arm are respectively connected to the steering arm; the center position of the steering column is respectively connected to the braking system assembly, the hub motor, and the tire; the steering motor controller and the steering motor are both electrically connected to the low-voltage power supply through the low-voltage wire harness; the hub motor is electrically connected to the high-voltage power supply through the high-voltage wire harness; the high-voltage power supply and the low-voltage power supply are both fixed on the mounting bracket, and one end of the mounting bracket is rigidly connected to the short side of the steering arm through the vehicle frame connecting piece;

[0013] The vehicle controller is installed in the front middle part of the vehicle, respectively connecting and controlling the left front wheel corner module, the right front wheel corner module, the left rear wheel corner module, and the right rear wheel corner module, and communicating with the steering motor controllers in the four corner modules respectively.

[0014] The power supply module is fixedly installed in the middle rear part of the vehicle frame, electrically connected to the high-voltage power supply and the low-voltage power supply, and respectively realizing the detection and control of the power of the high-voltage power supply and the low-voltage power supply.

[0015] Further, the steering motor uses a small-torque low-voltage motor.

[0016] The present invention also provides a method for evaluating the roll risk of a steer-by-wire four-wheel independent steering system. Based on the above system, the steps are as follows:

[0017] 1) Build a vertical load distribution model for the steer-by-wire four-wheel independent steering system based on vehicle speed, vehicle body attitude, load, and road surface adhesion coefficient to monitor and feedback the time-varying and non-linear changes of the vertical load in real time;

[0018] 2) Based on the vertical load calculation results in step 1), establish a real-time roll risk assessment model that integrates vehicle longitudinal acceleration, lateral acceleration, suspension dynamic response, and tire lateral force characteristics, design a dynamic compensation algorithm to correct the model, and generate a sequence of roll risk prediction values for future working conditions through time series analysis and prediction algorithms;

[0019] 3) Combine the roll risk prediction value sequence in step 2) with the change of vertical load distribution to establish a roll risk warning model, and generate safety warning signals of different levels based on the risk level determination mechanism.

[0020] Further, step 1) specifically includes:

[0021] 11) Build a vehicle roll dynamics model to describe the influence of vehicle speed, vehicle body attitude, load, and road surface adhesion coefficient on the vertical load; the vehicle roll dynamics model has the following force and moment balance equations;

[0022] The lateral force balance equation is:

[0023]

[0024] In the formula, m is the total vehicle mass, v x is the vehicle longitudinal driving speed, is the vehicle center of mass sideslip angular velocity, β and γ are respectively the vehicle center of mass sideslip angle and yaw angular velocity, m s is the unsprung mass of the vehicle, h s is the center of mass height of the unsprung mass, is the vehicle body roll angular acceleration, k f 、k r are respectively the front tire lateral stiffness and rear tire lateral stiffness of the vehicle, l f 、l r are respectively the distance from the center of mass to the front axle and the distance from the center of mass to the rear axle, R f 、R r are respectively the roll stiffness of the vehicle front axle and the roll stiffness of the vehicle rear axle, is the vehicle body roll angle, δ f 、δ r are respectively the vehicle front wheel steering angle and the vehicle rear wheel steering angle;

[0025] The yaw moment balance equation is:

[0026]

[0027] In the formula, Iz is the moment of inertia of the vehicle about the z-axis, is the yaw angular velocity change rate of the vehicle, I xz is the product of inertia of the vehicle about the x-axis and z-axis;

[0028] The balance equation of the sprung mass roll moment is:

[0029]

[0030] In the formula, I x is the moment of inertia of the vehicle about the x-axis, is the roll damping coefficient, is the body roll stiffness coefficient, g is the acceleration due to gravity;

[0031] The balance equation of the sprung mass pitch moment is:

[0032]

[0033] In the formula, I y is the moment of inertia about the y-axis, is the body pitch angular acceleration, e is the height from the center of mass to the roll center, is the longitudinal velocity change rate of the vehicle, v y is the lateral driving speed of the vehicle, k θ is the pitch stiffness coefficient of the vehicle, c θ is the pitch damping coefficient, θ is the body pitch angle, is the body pitch angular velocity;

[0034] The balance equation of the sprung mass taking moment about the roll center is:

[0035]

[0036] In the formula, a ys is the lateral acceleration of the sprung mass, is the body roll angular velocity;

[0037] The balance equation of the unsprung mass about the central axis of the tire contact point is:

[0038]

[0039] In the formula, F zfl is the vertical load of the left front wheel of the vehicle, F zfr is the vertical load of the right front wheel of the vehicle, F zrl is the vertical load of the left rear wheel of the vehicle, F zrr is the vertical load of the right rear wheel of the vehicle, B is the wheelbase of the vehicle, m d is the unsprung mass of the vehicle, h d is the center of mass height of the unsprung mass, h sris the height from the roll center to the ground, a y is the lateral acceleration of the vehicle, is the rate of change of the vehicle's lateral velocity;

[0040] Vertical force balance equation:

[0041]

[0042] In the formula, is the rate of change of the vehicle's vertical velocity;

[0043] 12) Based on the vehicle rollover dynamics model in step 11), design a vertical load distribution model applicable to dynamic conditions, calculate and predict the vertical load distribution under different conditions in real time. The calculation formulas for the vertical loads of each tire are as follows:

[0044]

[0045] In the formula, l is the distance from the front axle to the rear axle, l = l f + l r , B is the vehicle track width, a x is the longitudinal acceleration of the vehicle, is the rate of change of the vehicle's longitudinal velocity;

[0046] 13) Based on the vehicle rollover dynamics model in step 11) and the vertical load distribution model in step 12), monitor the change of the vehicle's vertical load, and use model predictive control to adjust and optimize the calculation result of the vertical load to ensure real-time performance and accuracy;

[0047] Select the longitudinal vehicle speed, lateral vehicle speed, vertical vehicle speed, roll angular velocity, pitch angular velocity and yaw angular velocity as state variables, expressed as Regard the six state variables as the controlled outputs, expressed as Regard the longitudinal resultant force, lateral resultant force, vertical resultant force, roll resultant moment, pitch resultant moment and yaw resultant moment as control inputs, expressed as u = [F xd , F yd , F zd , M xd , M yd , M zd , where

[0048]

[0049] Conduct a force analysis on the vehicle to obtain the pitch moment M xd of the vehicle, roll moment M yd and yaw moment M zd The relationship with the tire forces is:

[0050]

[0051] The linear two-degree-of-freedom vehicle model is adopted as the reference model for the ideal motion of the vehicle body. It is assumed that the driver input directly acts on the steering wheels to generate the expectation of the driver's operation, and the influence of the suspension system and load transfer during steering is ignored. The left and right sides of the vehicle are combined and simplified, and at the same time, the change in the longitudinal speed of the vehicle body is ignored. The state matrix of the linear two-degree-of-freedom vehicle model is as follows:

[0052]

[0053] In the formula, are the change rates of the expected values of the vehicle lateral speed and yaw angular velocity respectively;

[0054] The expected value of the longitudinal speed is determined by using the given initial speed initial value and acceleration curve. The expected value v of the longitudinal speed xd is the expected value a of the longitudinal acceleration xd integrated over time t, as follows:

[0055]

[0056] In the formula, v x0 is the initial longitudinal speed of the vehicle, t is the integration time, and a xd (t) is the expected longitudinal acceleration;

[0057] Take the pitch angle, pitch angular velocity, roll angle, and roll angular velocity expected values of the vehicle as 0, as follows:

[0058]

[0059] In the formula, θ d 、 are the pitch angle and roll angle of the vehicle respectively, are the pitch angular velocity and roll angular velocity of the vehicle respectively;

[0060] Write the expected longitudinal, lateral, and vertical speeds of the vehicle, as well as the yaw angular velocity, expected roll angular velocity, and expected pitch angular velocity of the vehicle as state variables in the form of;

[0061] Write the vehicle roll dynamics model in the form of a continuous-time state-space equation, which can be expressed as:

[0062]

[0063] In the formula, x(t) is the state vector of the system at time t, u(t) is the input vector of the system at time t, and f(x(t), u(t))

[0064] is the state equation, g(x(t), u(t)) is the output equation, and y(t) is the output vector of the system at time t;

[0065] Taking the sampling time, the continuous state-space model is discretized, and the discretized state-space model is:

[0066] x(k + 1) = F(x(k), u(k))

[0067] y(k) = G(x(k), u(k))

[0068] Where x(k) represents the state vector at discrete time k, x(k + 1) represents the state vector at the next time, u(k) is the input vector at discrete time k, F(x(k), u(k)) is the state equation at discrete time k, G(x(k), u(k)) is the output equation at discrete time k, and y(k) is the output vector at discrete time k;

[0069] Defining the prediction horizon as p and the control horizon as m, where p > m, the state of the vehicle within the prediction horizon [k + 1, k + p] can be obtained based on the current state of the vehicle and the model predictive control process. At time k + p, the vehicle state is:

[0070] x(k + p) = F(x(k), u(k), u(k + 1), …, u(k + m), …, u(k + p - 1))

[0071] Where x(k) represents the state vector at discrete time k, u(k) is the input vector at discrete time k, u(k + 1) is the input vector at discrete time k + 1, u(k + m) is the input vector at discrete time k + m, u(k + p - 1) is the input vector at discrete time k + p - 1, and x(k + p) represents the state vector at the prediction horizon p;

[0072] During the model predictive control process, the state and control input of the system are optimized within a prediction horizon to minimize a certain performance index. When the prediction horizon p is greater than the control horizon m, that is, when the sampling time is greater than the control horizon, the control input remains unchanged until the prediction horizon:

[0073] u(k + m - 1) = u(k + m) = u(k + m + 1) = … = u(k + p - 1)

[0074] Where u(k + m - 1) is the input vector at discrete time k + m - 1, u(k + m) is the input vector at discrete time k + m, u(k + m + 1) is the input vector at discrete time k + m + 1, and u(k + p - 1) is the input vector at discrete time k + p - 1;

[0075] Let the optimal control input and predicted output at time k be as follows:

[0076]

[0077] where \(U(k)\) is the optimal control input vector at discrete time \(k\), which contains the optimal control input sequence from the current time \(k\) to \(k + m - 1\), \(u(k|k)\) is the optimal control input based on the information at discrete time \(k\), \(u(k + 1|k)\) is the optimal control input based on the information at discrete time \(k\) for time \(k + 1\), \(u(k + m - 1|k)\) is the optimal control input based on the information at discrete time \(k\) for time \(k + m - 1\), \(Y(k + 1|k)\) is the predicted output vector at discrete time \(k\), which contains the predicted output sequence from \(k + 1\) to \(k + p\), \(r(k + 1|k)\) is the predicted output based on the information at discrete time \(k\) for time \(k + 1\), \(r(k + 2|k)\) is the predicted output based on the information at discrete time \(k\) for time \(k + 2\), and \(r(k + p|k)\) is the predicted output based on the information at discrete time \(k\) for time \(k + p\);

[0078] Considering the vehicle body attitude and suspension response, while realizing real-time monitoring and optimization of the vehicle vertical load, the corresponding formula for the tire vertical load is:

[0079]

[0080] where \(a\) xs is the longitudinal acceleration of the unsprung mass, and \(h\) sp is the height from the pitch center to the ground.

[0081] Furthermore, step 2) specifically includes:

[0082] 21) Based on the vertical load calculation result in step 1), fuse the longitudinal acceleration, lateral acceleration, suspension dynamic response, and tire lateral force characteristics to establish a real-time roll risk assessment model;

[0083] The vehicle lateral load transfer ratio \(LTR\) is used as the vehicle roll risk assessment index. The vehicle lateral load transfer ratio is defined as the ratio of the absolute value of the difference in the vertical loads of the left and right tires of the vehicle to the total vertical load of the tires, as follows:

[0084]

[0085] where \(F\) zl and \(F\) zr respectively represent the total vertical loads of the left and right tires of the vehicle;

[0086] 22) Based on the real-time roll risk assessment model in step 21), for the non-linear changes under high-speed driving, sharp turning and extreme driving conditions, the unscented Kalman filter is used to estimate the body roll angle, roll angle speed and lateral acceleration, and a vehicle lateral load transfer rate adaptive dynamic compensation algorithm is designed to correct the model;

[0087] According to the established vehicle roll dynamics model, a body roll angle estimation model based on unscented Kalman filter is constructed. Let the state vector be The input vector is u1 = [δ f δ r T , and the output vector is

[0088] 23) Based on the corrected roll risk assessment model in step 22), the long short-term memory model is used for time series analysis and prediction to generate a sequence of roll risk prediction values for future working conditions;

[0089] The long short-term memory model consists of an input layer, a fully connected layer, an output layer and two LSTM layers. The input layer is X = (X t-i+1 ,..., X t-2 , X t-1 , X t ), where X is the state vector containing the states from i moments before t. represents the body roll angle at time t body roll angle speed lateral acceleration a y , the lateral acceleration a of the sprung mass ys four state variables;

[0090] The output layer is represents the predicted output sequence of the LTR containing the states from i moments before t;

[0091] Each node in the LSTM layer is selected as an LSTM cell. The structure of the LSTM cell includes four components: a forget gate, an input gate, a cell state, and an output gate;

[0092] When the X vector is used as the unit input at time t, the formula for the forget gate is as follows:

[0093] f t = σ(W f [h t-1 , X t + b f )

[0094] In the formula, f t is the output, W f is the input weight, b f ​is the bias, and σ(·) represents the sigmoid activation function;

[0095] When the X vector is used as the unit input at time t, the input gate formula is as follows:

[0096] i t = σ(W i [h t-1 , X t + b i )

[0097] In the formula, i t is the output, W i is the input weight, b i is the bias, and σ(·) represents the sigmoid activation function;

[0098] When the X vector is used as the unit input at time t, the cell state formula is as follows:

[0099]

[0100] In the formula, is the candidate cell state, W c is the input weight of the cell state, b c is the bias of the cell state, c t is the cell state at time t, c t-1 is the cell state at time t-1, and tanh(·) represents the hyperbolic tangent function;

[0101] When the X vector is used as the unit input at time t, the output gate formula is as follows:

[0102] o t = σ(W o [h t-1 , X t + b o )

[0103] h t = o t tanh(c t )

[0104] In the formula, o t is the output, W o is the input weight, b o is the bias; h t and h t-1 represent the hidden states at the current time and the previous time respectively; c t and c t-1 represent the cell states at the current time and the previous time respectively; σ(·) represents the sigmoid activation function, and tanh(·) represents the hyperbolic tangent function;

[0105] The regression output layer needs to calculate the root mean square error loss of the regression. For a single observation, the root mean square error e RMSE The calculation formula is as follows:

[0106]

[0107] In the formula, η is the number of responses, t i is the target output, and Y i is the predicted output.

[0108] Furthermore, the unscented Kalman filtering process in step 22) is as follows:

[0109] 221) Use unscented Kalman filtering to generate an importance density function for each particle, and construct a 2n + 1 Sigma point sampling distribution χ for each particle i and the weight assignment formula is:

[0110]

[0111] In the formula, n is the dimension of the state quantity; λ is the scaling factor, λ = α 2 (n + κ) - n, α is a small positive value, which determines the distribution state of the Sigma points; δ is a non - negative weight coefficient, which can combine the dynamic errors of the high - order terms in the equation; is the state quantity of the i - th Sigma point at time k - 1; is the covariance of the i - th Sigma at time k - 1; is the weight assignment coefficient for approximating the mean; is the weight assignment coefficient for approximating the covariance; κ ∈ R, which characterizes the dispersion degree of the Sigma points relative to the mean. R is a real number. The larger κ is, the farther the Sigma points at non - mean positions are from the mean and the smaller the weights are. Take κ = 3 - n;

[0112] 222) Calculate the prior distribution of the state quantity at time k by weighted calculation, as follows:

[0113]

[0114] In the formula, is the prior estimate value of the state quantity at time k, is the state quantity at time k, is the prior estimate value of the covariance of the state quantity at time k, is the weight of the mean of the state quantities of each Sigma point, is the weight of the covariance of the state quantities of each Sigma point, is the variance of the system noise at time k;

[0115] 223) For the state quantity at time k Perform Sigma point sampling on the prior distribution and the observed quantity The probability distribution is as follows:

[0116]

[0117] In the formula, is Gaussian white noise with zero mean and covariance R of the observed quantity, is the weight distribution coefficient of the approximate mean, 2n is the number of Sigma points, is the observed quantity at time k, is the weighted mean of the observed quantity at time k;

[0118] 224) Calculate the cross-covariance of the state quantity and the observed quantity and the autocovariance of the observed quantity as follows:

[0119]

[0120] In the formula, is the cross-covariance of the observed quantity, is the autocovariance of the observed quantity, is the weight distribution coefficient of the approximate mean square error, is the observed quantity at time k, is the weighted mean of the observed quantity at time k, is the variance of the observation noise;

[0121] 225) Calculate the Kalman gain of the state quantity at time k and update the state quantity as follows:

[0122]

[0123] In the formula, K i is the Kalman gain of the i-th Sigma point, is the mean of the state quantity at time k, z k is the value of the observed quantity at time k, is the weighted mean of the observed quantity at time k, is the covariance estimate value of the state quantity at time k;

[0124] According to the roll angle, roll rate and lateral acceleration of the vehicle body estimated by the unscented Kalman filter in steps 221)-225), the lateral load transfer ratio LTR is expressed as:

[0125]

[0126] Select the vehicle lateral acceleration and the roll angle of the vehicle body as the threshold parameters of the adaptive compensation strategy. The formula is as follows:

[0127]

[0128] LTR * = LTR + ΔLTR

[0129] where ΔLTR is the compensation amount of the lateral load transfer ratio, k ay 、 are respectively the lateral acceleration compensation coefficient and the vehicle body roll angle compensation coefficient, a y* is the vehicle lateral acceleration threshold, is the vehicle body roll angle threshold.

[0130] Furthermore, step 3) specifically includes:

[0131] 31) Based on the roll risk prediction value sequence generated in step 23), construct a roll risk warning model based on the rollover warning time TTR. The rollover warning time TTR is defined as the time when the vehicle runs to rollover with the past period and the current moment state as the initial conditions. The roll risk warning model can be expressed as:

[0132] TTR = t r - t0

[0133] In the formula, t r is the vehicle rollover time, and t0 is the current moment;

[0134] 32) Based on the roll risk warning model constructed in step 31), perform dynamic adjustment. Set the step size of the roll risk warning model as T s = the workload of a0, and the upper limit value of the rollover warning time TTR is T up = a. When the vehicle rollover warning time is within the upper limit value T up and the lateral load transfer ratio is not higher than the maximum value LTR max or lower than the minimum value LTR min , the vehicle is driving safely and will not roll over, and the next prediction is carried out; According to the roll risk warning model step size T s and the vehicle rollover dynamics model in step 1), calculate the lateral load transfer ratio of the vehicle. When the lateral load transfer ratio LTR reaches the rollover condition, obtain the running step number N, and obtain the rollover warning time as TTR = N * T s , where N is the running step number and T s is the warning step size;

[0135] Use the lateral load transfer ratio LTR to dynamically adjust the threshold value of the rollover warning time TTR. The specific adjustment method is as follows:

[0136] When the lateral load transfer ratio LTR exceeds the maximum value LTR maxWhen it indicates that the vehicle load transfer is severe, the roll angle speed increases rapidly, and the vehicle rollover risk is high, it is necessary to adjust the TTR threshold downward from a seconds to a - a h seconds, triggering the rollover warning earlier, where a h is the downward adjustment amount of the TTR threshold when the vehicle rollover risk increases; when the lateral load transfer rate LTR of the vehicle is lower than the minimum value LTR min the vehicle rollover risk is small, and the TTR threshold is adjusted upward to a + a l seconds to avoid the possibility of false alarms for the vehicle, where a l is the upward adjustment amount of the TTR threshold when the vehicle rollover risk decreases; when the lateral load transfer rate LTR of the vehicle is between the maximum value LTR max and the minimum value LTR min the TTR threshold remains unchanged at a;

[0137] 33) Establish a risk level determination mechanism that fuses multiple indicators based on the lateral load transfer rate LTR and the rollover warning time TTR, and generate safety warning signals of different levels;

[0138] According to the combined values of LTR and TTR, set the following different warning level mechanisms:

[0139] Low risk: LTR < LTR min and TTR > a + a l ;

[0140] Medium risk: LTR min ≤ LTR ≤ LTR max or TTR < a;

[0141] High risk: LTR > LTR max and TTR < a - a h ;

[0142] Send different risk warning signals according to the risk warning results.

[0143] Advantages of the present invention:

[0144] 1. By designing a system that can monitor and calculate the vertical load of the steer-by-wire four-wheel independent steering system in real time, and combining the multi-dimensional dynamic information of the vehicle (such as acceleration, suspension response, etc.), the present invention establishes a highly accurate rollover risk assessment model. At the same time, by introducing a dynamic compensation algorithm and time series prediction technology, the real-time response ability and prediction accuracy of the model are further improved.

[0145] 2. By real-time monitoring and predicting the roll risk of the vehicle, the present invention generates safety warning signals of different levels, provides decision-making support for the intelligent driving system, helps to take preventive measures in a timely manner under extreme conditions such as high-speed driving and sharp turns, avoid rollover accidents, and significantly improve the safety performance of the vehicle.

[0146] 3. The present invention integrates multi-source information such as the steering state, load distribution change, and environmental disturbance of the vehicle into the roll risk warning model in real time, improves the adaptability of the system to complex environments, not only optimizes the safety performance of the vehicle, but also provides important support for the development of future intelligent transportation and autonomous driving technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0147] Figure 1 is a schematic structural diagram of the system of the present invention;

[0148] Figure 2 is a schematic structural diagram of the wheel corner module in the present invention;

[0149] Figure 3 is a schematic diagram of the principle of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0150] For the convenience of understanding by those skilled in the art, the present invention will be further described below in conjunction with the embodiments and the drawings. The content mentioned in the embodiments does not limit the present invention.

[0151] Referring to Figure 1 、 Figure 2 As shown, a steer-by-wire four-wheel independent steering system of the present invention includes: a wheel corner module, a vehicle frame 5, a vehicle controller (ECU) 6, and a power supply 7;

[0152] The wheel corner module is composed of four corner modules: the left front corner module 1, the right front corner module 2, the left rear corner module 3, and the right rear corner module 4. Each corner module adopts the same structural design and is respectively installed at the left front, right front, left rear, and right rear of the vehicle frame 5, and is hinged to the vehicle frame 5. The corner module includes: a worm and gear reduction mechanism 11, a steering arm column 12, a corner sensor 13, a steering arm 14, a steering column 15, a brake system assembly 16, a hub motor 17, a lower fork arm 18, a tire 19, a high-voltage wire harness 20, an upper fork arm 21, a shock absorber 22, a high-voltage power supply 23, a mounting bracket 24, a low-voltage power supply 25, a low-voltage wire harness 26, a steering motor controller 27, a steering motor 28, a vehicle frame connecting piece 29, and a reducer 30; the steering motor controller 27 and the housing of the steering motor 28 adopt an integrated design. The steering motor controller 27 outputs a signal to control the rotation of the steering motor 28. The output shaft of the steering motor 28 is connected to the reducer 30. The reducer 30 is connected to the worm and gear reduction mechanism 11. The worm and gear reduction mechanism 11 is rigidly connected to the steering arm column 12 and transmits the torque of the steering motor 28 to the steering arm 14; the lower end of the steering arm column 12 is rigidly connected to the short side of the steering arm 14. The corner sensor 13 is installed on the steering arm column 12 to collect the corner signal of the steering arm column 12 in real time and send the signal to the steering motor controller 27; the steering arm 14 is connected to the steering column 15 through the shock absorber 22, the upper fork arm 21, and the lower fork arm 18. The upper end of the steering column 15 is respectively connected to the lower ends of the upper fork arm 21 and the shock absorber 22. The lower end of the steering column 15 is connected to the upper end of the lower fork arm 18. The upper ends of the upper fork arm 21, the shock absorber 22, and the lower end of the lower fork arm 18 are respectively connected to the steering arm 14; the central position of the steering column 15 is respectively connected to the brake system assembly 16, the hub motor 17, and the tire 19; the steering motor controller 27 and the steering motor 28 are both electrically connected to the low-voltage power supply 25 through the low-voltage wire harness 26; the hub motor 17 is electrically connected to the high-voltage power supply 23 through the high-voltage wire harness 20; the high-voltage power supply 23 and the low-voltage power supply 25 are both fixed on the mounting bracket 24. One end of the mounting bracket 24 is rigidly connected to the short side of the steering arm 14 through the vehicle frame connecting piece 29;

[0153] The vehicle controller 6 is installed in the front middle part of the vehicle, and is respectively connected to and controls the left front wheel corner module 1, the right front wheel corner module 2, the left rear wheel corner module 3, and the right rear wheel corner module 4, and communicates with the steering motor controllers in the four corner modules respectively;

[0154] The power supply module 7 is fixedly installed at the middle rear of the vehicle frame, and is electrically connected to the high-voltage power supply 23 and the low-voltage power supply 25 to respectively detect and control the power of the high-voltage power supply 23 and the low-voltage power supply 25.

[0155] In the example, the steering motor 28 adopts a small torque and low voltage motor.

[0156] Referring to Figure 3 as shown, the present invention also provides a method for evaluating the roll risk of a steer-by-wire four-wheel independent steering system, and the steps are as follows:

[0157] 1) Construct a vertical load distribution model of the steer-by-wire four-wheel independent steering system based on vehicle speed, vehicle body attitude, load, and road adhesion coefficient to monitor and feedback the time-varying and non-linear changes of the vertical load in real time;

[0158] Among them, the specific steps of step 1) include:

[0159] 11) Construct a vehicle roll dynamics model to describe the influence of vehicle speed, vehicle body attitude, load, and road adhesion coefficient on the vertical load; the vehicle roll dynamics model has the following force and moment balance equations;

[0160] The lateral force balance equation is:

[0161]

[0162] In the formula, m is the total vehicle mass, v x is the longitudinal driving speed of the vehicle, is the angular velocity of the vehicle's center of mass side slip, β and γ are respectively the center of mass side slip angle and yaw angular velocity of the vehicle, m s is the sprung mass of the vehicle, h s is the height of the center of mass of the sprung mass, is the angular acceleration of the vehicle body roll, k f and k r are respectively the cornering stiffness of the front tires and the cornering stiffness of the rear tires of the vehicle, l f and l r are respectively the distance from the center of mass to the front axle and the distance from the center of mass to the rear axle, R f and R r are respectively the roll stiffness of the front axle and the roll stiffness of the rear axle of the vehicle, is the roll angle of the vehicle body, δ f and δ r are respectively the front wheel steering angle and the rear wheel steering angle of the vehicle;

[0163] The yaw moment balance equation is:

[0164]

[0165] In the formula, I z is the moment of inertia of the vehicle about the z-axis, is the change rate of the vehicle's yaw angular velocity, I xz is the product of inertia of the vehicle about the x-axis and z-axis;

[0166] The roll moment balance equation of the sprung mass is:

[0167]

[0168] In the formula, I x is the moment of inertia of the vehicle about the x-axis, is the roll damping coefficient, is the body roll stiffness coefficient, and g is the acceleration due to gravity;

[0169] The pitching moment balance equation of the sprung mass is:

[0170]

[0171] In the formula, I y is the moment of inertia about the y-axis, is the angular acceleration of the body pitch, e is the height from the center of mass to the roll center, is the longitudinal speed change rate of the vehicle, v y is the lateral driving speed of the vehicle, k θ is the pitching stiffness coefficient of the vehicle, c θ is the pitching damping coefficient, θ is the body pitch angle, is the angular velocity of the body pitch;

[0172] The moment balance equation of the sprung mass about the roll center is:

[0173]

[0174] In the formula, a ys is the lateral acceleration of the sprung mass, is the angular velocity of the body roll;

[0175] The moment balance equation of the unsprung mass about the center axis of the tire contact point is:

[0176]

[0177] In the formula, F zfl is the vertical load of the left front wheel of the vehicle, F zfr is the vertical load of the right front wheel of the vehicle, F zrl is the vertical load of the left rear wheel of the vehicle, F zrr is the vertical load of the right rear wheel of the vehicle, B is the wheelbase of the vehicle, m d is the unsprung mass of the vehicle, h d is the center of mass height of the unsprung mass, h sr is the height from the roll center to the ground, a y is the lateral acceleration of the vehicle, is the lateral speed change rate of the vehicle;

[0178] The force balance equation in the vertical direction:

[0179]

[0180] In the formula, is the vertical velocity change rate of the vehicle;

[0181] 12) Based on the vehicle roll dynamics model in step 11), design a vertical load distribution model applicable to dynamic conditions, calculate and predict the vertical load distribution under different conditions in real time. The calculation formulas for the vertical load of each tire are as follows:

[0182]

[0183] In the formula, l is the distance from the front axle to the rear axle, l = l f + l r , B is the wheelbase of the vehicle, a x is the longitudinal acceleration of the vehicle, is the longitudinal velocity change rate of the vehicle;

[0184] 13) Based on the vehicle roll dynamics model in step 11) and the vertical load distribution model in step 12), monitor the change of the vertical load of the vehicle, and use model predictive control to adjust and optimize the calculation result of the vertical load to ensure real-time performance and accuracy;

[0185] Select the longitudinal vehicle speed, lateral vehicle speed, vertical vehicle speed, roll angular velocity, pitch angular velocity and yaw angular velocity as state variables, expressed as Regard the six state variables as the controlled outputs, expressed as Regard the longitudinal resultant force, lateral resultant force, vertical resultant force, roll resultant moment, pitch resultant moment and yaw resultant moment as control inputs, expressed as u = [F xd , F yd , F zd , M xd , M yd , M zd , where

[0186]

[0187] Conduct a force analysis on the vehicle to obtain the relationship between the pitch moment M xd of the vehicle, the roll moment M yd and the yaw moment M zd of the vehicle and the tire forces as:

[0188]

[0189] A linear two-degree-of-freedom vehicle model is used as the reference model for the ideal motion of the vehicle body. It is assumed that the driver input directly acts on the steering wheels to generate the expected driver operation, and the effects of the suspension system and load transfer during steering are ignored. The left and right sides of the vehicle are combined and simplified, and at the same time, the change in the longitudinal speed of the vehicle body is ignored. The state matrix of the linear two-degree-of-freedom vehicle model is as follows:

[0190]

[0191] In the formula, are the change rates of the expected values of the vehicle lateral speed and yaw angular velocity respectively;

[0192] The expected value of the longitudinal speed is determined by using the given initial speed initial value and acceleration curve. The expected value of the longitudinal speed v xd is the integral of the expected value of the longitudinal acceleration a xd over time t, as follows:

[0193]

[0194] In the formula, v x0 is the initial longitudinal speed of the vehicle, t is the integration time, and a xd (t) is the expected longitudinal acceleration;

[0195] The pitch angle, pitch angular velocity, roll angle, and roll angular velocity expected values of the vehicle are taken as 0, as follows:

[0196]

[0197] In the formula, θ d 、 are the pitch angle and roll angle of the vehicle respectively, are the pitch angular velocity and roll angular velocity of the vehicle respectively;

[0198] The expected longitudinal, lateral, and vertical speeds of the vehicle, as well as the yaw angular velocity, expected roll angular velocity, and expected pitch angular velocity of the vehicle, are written in the form of state variables ;

[0199] The vehicle roll dynamics model is written in the form of a continuous-time state-space equation, which can be expressed as:

[0200]

[0201] In the formula, x(t) is the state vector of the system at time t, u(t) is the input vector of the system at time t, f(x(t), u(t))

[0202] is the state equation, g(x(t), u(t)) is the output equation, and y(t) is the output vector of the system at time t;

[0203] Take the sampling time to discretize the continuous state - space model. The discretized state - space model is as follows:

[0204] x(k + 1)=F(x(k),u(k))

[0205] y(k)=G(x(k),u(k))

[0206] In the formula, x(k) represents the state vector at discrete time k, x(k + 1) represents the state vector at the next time, u(k) is the input vector at discrete time k, F(x(k),u(k)) is the state equation at discrete time k, G(x(k),u(k)) is the output equation at discrete time k, and y(k) is the output vector at discrete time k;

[0207] Define the prediction horizon as p and the control horizon as m, where p>m. The state of the vehicle within the prediction horizon [k + 1,k + p] can be obtained based on the current state of the vehicle and the model predictive control process. At time k + p, the vehicle state is:

[0208] x(k + p)=F(x(k),u(k),u(k + 1),…,u(k + m),…,u(k + p - 1))

[0209] In the formula, x(k) represents the state vector at discrete time k, u(k) is the input vector at discrete time k, u(k + 1) is the input vector at discrete time k + 1, u(k + m) is the input vector at discrete time k + m, u(k + p - 1) is the input vector at discrete time k + p - 1, and x(k + p) represents the state vector at the prediction - horizon time p;

[0210] During the model predictive control process, the state and control input of the system are optimized within a prediction horizon to minimize a certain performance index. When the prediction horizon p is greater than the control horizon m, that is, when the sampling time is greater than the control horizon, the control input remains unchanged until the prediction horizon:

[0211] u(k + m - 1)=u(k + m)=u(k + m + 1)=…=u(k + p - 1)

[0212] In the formula, u(k + m - 1) is the input vector at discrete time k + m - 1, u(k + m) is the input vector at discrete time k + m, u(k + m + 1) is the input vector at discrete time k + m + 1, and u(k + p - 1) is the input vector at discrete time k + p - 1;

[0213] Let the optimal control input and predicted output at time k be:

[0214]

[0215] In the formula, U(k) is the optimal control input vector at discrete time k, which contains the optimal control input sequence from the current time k to k+m-1. u(k|k) is the optimal control input based on the information at discrete time k. u(k+1|k) is the optimal control input based on the information at discrete time k+1. u(k+m-1|k) is the optimal control input based on the information at discrete time k+m-1. Y(k+1|k) is the predicted output vector at discrete time k, which contains the predicted output sequence from k+1 to k+p. r(k+1|k) is the predicted output based on the information at discrete time k+1. r(k+2|k) is the predicted output based on the information at discrete time k+2. r(k+p|k) is the predicted output based on the information at discrete time k+p;

[0216] Considering the vehicle body attitude and suspension response, while realizing the real-time monitoring and optimization of the vehicle vertical load, the corresponding formula for the tire vertical load is:

[0217]

[0218] In the formula, a xs is the longitudinal acceleration of the unsprung mass, h sp is the height from the pitch center to the ground.

[0219] 2) Based on the vertical load calculation result in step 1), establish a real-time roll risk assessment model that integrates the vehicle longitudinal acceleration, lateral acceleration, suspension dynamic response, and tire lateral force characteristics, design a dynamic compensation algorithm to correct the model, and generate a sequence of roll risk prediction values for future working conditions through time series analysis and prediction algorithms;

[0220] Among them, the specific content of step 2) includes:

[0221] 21) Based on the vertical load calculation result in step 1), integrate the longitudinal acceleration, lateral acceleration, suspension dynamic response, and tire lateral force characteristics to establish a real-time roll risk assessment model;

[0222] Use the vehicle lateral load transfer rate LTR as the vehicle roll risk assessment index. The vehicle lateral load transfer rate is defined as the ratio of the absolute value of the difference in the vertical loads of the tires on the left and right sides of the vehicle to the total tire vertical load, as follows:

[0223]

[0224] In the formula, F zl 、F zrrespectively represent the total vertical load of the left-side tires and the total vertical load of the right-side tires of the vehicle;

[0225] 22) Based on the real-time roll risk assessment model in step 21), for the non-linear changes under high-speed driving, sharp turning and extreme driving conditions, use the unscented Kalman filter to estimate the body roll angle, roll angular velocity and lateral acceleration, and design an adaptive dynamic compensation algorithm for the vehicle lateral load transfer ratio to correct the model;

[0226] According to the established vehicle roll dynamics model, construct a body roll angle estimation model based on the unscented Kalman filter, and set the state vector as The input vector is u1 = [δ f δ r T , and the output vector is

[0227] 23) Based on the corrected roll risk assessment model in step 22), use the long short-term memory model to perform time series analysis and prediction, and generate a sequence of roll risk prediction values for future working conditions;

[0228] The long short-term memory model consists of an input layer, a fully connected layer, an output layer and two LSTM layers. The input layer is X = (X t-i+1 ,..., X t-2 , X t-1 , X t ), where X is the state vector containing the states from i moments before t, represents the body roll angle at time t body roll angular velocity lateral acceleration a y , the lateral acceleration a of the sprung mass ys four state variables;

[0229] The output layer is represents the predicted output sequence of the LTR containing the states from i moments before t;

[0230] Each node in the LSTM layer is selected as an LSTM cell, and the structure of the LSTM cell includes four components: a forget gate, an input gate, a cell state, and an output gate;

[0231] When the X vector is used as the unit input at time t, the formula for the forget gate is as follows:

[0232] f t = σ(W f [h t-1 , X t + b f )

[0233] In the formula, ft is the output, W f is the input weight, b f is the bias, and σ(·) represents the sigmoid activation function;

[0234] When the X vector is used as the cell input at time t, the input gate formula is as follows:

[0235] i t = σ(W i [h t-1 , X t + b i )

[0236] In the formula, i t is the output, W i is the input weight, b i is the bias, and σ(·) represents the sigmoid activation function;

[0237] When the X vector is used as the cell input at time t, the cell state formula is as follows:

[0238]

[0239] In the formula, is the candidate cell state, W c is the input weight of the cell state, b c is the bias of the cell state, c t is the cell state at time t, c t-1 is the cell state at time t-1, and tanh(·) represents the hyperbolic tangent function;

[0240] When the X vector is used as the cell input at time t, the output gate formula is as follows:

[0241] o t = σ(W o [h t-1 , X t + b o )

[0242] h t = o t tanh(c t )

[0243] In the formula, o t is the output, W o is the input weight, b o is the bias; h t and h t-1 represent the hidden states at the current time and the previous time respectively; c t and c t-1respectively represent the cell states at the current moment and the previous moment; σ(·) represents the sigmoid activation function, and tanh(·) represents the hyperbolic tangent function;

[0244] For the regression output layer, the root mean square error loss of the regression needs to be calculated. For a single observation, the root mean square error e RMSE The calculation formula is as follows:

[0245]

[0246] In the formula, η is the number of responses, t i is the target output, Y i is the predicted output.

[0247] Specifically, the unscented Kalman filtering process in step 22) is as follows:

[0248] 221) Use unscented Kalman filtering to generate an importance density function for each particle, and construct a 2n + 1 Sigma point sampling distribution χ for each particle i and the weight assignment formula is:

[0249]

[0250] In the formula, n is the dimension of the state quantity; λ is the scaling factor, λ = α 2 (n + κ) - n, α is a small positive value, which determines the distribution state of the Sigma points; δ is a non - negative weight coefficient, which can combine the dynamic errors of the high - order terms in the equation; is the state quantity of the i - th Sigma point at the (k - 1) - th moment; is the covariance of the i - th Sigma at the (k - 1) - th moment; is the weight assignment coefficient for approximating the mean; is the weight assignment coefficient for approximating the covariance; κ ∈ R, which characterizes the spread degree of the Sigma points relative to the mean. The larger κ is, the farther the Sigma points at non - mean positions are from the mean, and the smaller the weights are. Take κ = 3 - n;

[0251] 222) Calculate the prior distribution of the state quantity at time k by weighted calculation, as follows:

[0252]

[0253] In the formula, is the prior estimated value of the state quantity at time k, is the state quantity at time k, is the prior estimated value of the covariance of the state quantity at time k, is the weight of the mean of the state quantities of each Sigma point, is the weight of the state quantity covariance of each Sigma point. is the variance of the system noise at time k;

[0254] 223) Sample the prior distribution of the state quantity at time k and the probability distribution of the observed quantity as follows:

[0255]

[0256] where is Gaussian white noise with zero mean and covariance R of the observed quantity, is the weight distribution coefficient of the approximate mean, 2n is the number of Sigma points, is the observed quantity at time k, is the weighted mean of the observed quantity at time k;

[0257] 224) Calculate the cross-covariance between the state quantity and the observed quantity and the auto-covariance of the observed quantity as follows:

[0258]

[0259] where is the cross-covariance of the observed quantity, is the auto-covariance of the observed quantity, is the weight distribution coefficient of the approximate mean square error, is the observed quantity at time k, is the weighted mean of the observed quantity at time k, is the variance of the observation noise;

[0260] 225) Calculate the Kalman gain of the state quantity at time k and update the state quantity as follows:

[0261]

[0262] where K i is the Kalman gain of the i-th Sigma point, is the mean of the state quantity at time k, z k is the value of the observed quantity at time k, is the weighted mean of the observed quantity at time k, is the covariance estimate value of the state quantity at time k;

[0263] According to the roll angle, roll rate and lateral acceleration of the vehicle body estimated by the unscented Kalman filter in steps 221)-225), the lateral load transfer ratio LTR is expressed as:

[0264]

[0265] The vehicle lateral acceleration and the vehicle body roll angle are selected as the threshold parameters of the adaptive compensation strategy, and the formula is as follows:

[0266]

[0267] Among them, ΔLTR is the lateral load transfer ratio compensation amount, and k ay 、 are the lateral acceleration compensation coefficient and the vehicle body roll angle compensation coefficient respectively, is the vehicle lateral acceleration threshold, is the vehicle body roll angle threshold.

[0268] 3) Combine the roll risk prediction value sequence in step 2) with the change of the vertical load distribution to establish a roll risk warning model, and generate safety warning signals of different levels based on the risk level determination mechanism;

[0269] Among them, step 3) specifically includes:

[0270] 31) Based on the roll risk prediction value sequence generated in step 23), construct a roll risk warning model based on the time to roll (TTR). The time to roll (TTR) is defined as the time for the vehicle to roll over starting from the state of the past period and the current moment. The roll risk warning model can be expressed as:

[0271] TTR = t r - t0

[0272] In the formula, t r is the vehicle rollover time, and t0 is the current moment;

[0273] 32) Dynamically adjust based on the roll risk warning model constructed in step 31). Set the step size of the roll risk warning model as T s = the workload of a0, and the upper limit value of the time to roll (TTR) is T up = a. When the vehicle time to roll (TTR) is within the upper limit value T up and the lateral load transfer ratio is not higher than the maximum value LTR max or lower than the minimum value LTR min , the vehicle is driving safely and will not roll over, and the next prediction is carried out; According to the step size T s of the roll risk warning model and the vehicle rollover dynamics model in step 1), calculate the lateral load transfer ratio of the vehicle. When the lateral load transfer ratio LTR reaches the rollover condition, obtain the number of running steps N, and obtain the time to roll (TTR) = N * T s , where N is the number of running steps and T s is the warning step size;

[0274] Dynamically adjust the threshold of the rollover warning time TTR using the lateral load transfer rate LTR. The specific adjustment method is as follows:

[0275] When the lateral load transfer rate LTR exceeds the maximum value LTR max it indicates that the vehicle's load transfer is severe, the roll angle speed increases, and the vehicle has a high risk of rollover. It is necessary to adjust the TTR threshold downward from a seconds to a - a h seconds to trigger the rollover warning earlier, where a h is the downward adjustment amount of the TTR threshold when the vehicle's rollover risk increases; when the lateral load transfer rate LTR of the vehicle is lower than the minimum value LTR min the vehicle has a low rollover risk, and the TTR threshold is adjusted upward to a + a l seconds to avoid the possibility of false alarms for the vehicle, where a l is the upward adjustment amount of the TTR threshold when the vehicle's rollover risk decreases; when the lateral load transfer rate LTR of the vehicle is between the maximum value LTR max and the minimum value LTR min the TTR threshold remains unchanged at a;

[0276] 33) Establish a risk level determination mechanism that fuses multiple indicators based on the lateral load transfer rate LTR and the rollover warning time TTR to generate safety warning signals of different levels;

[0277] According to the combined values of LTR and TTR, set the following different warning level mechanisms:

[0278] Low risk: LTR < LTR min and TTR > a + a l ;

[0279] Medium risk: LTR min ≤ LTR ≤ LTR max or TTR < a;

[0280] High risk: LTR > LTR max and TTR < a - a h ;

[0281] Send different risk warning signals according to the risk warning results.

[0282] The specific application ways of the present invention are numerous. The above description is only the preferred implementation manner of the present invention. It should be noted that for those of ordinary skill in the art of this technology, without departing from the principle of the present invention, several improvements can still be made, and these improvements should also be regarded as the protection scope of the present invention.

Claims

1. A wire-controlled four-wheel independent steering system, characterized in that, Including: Wheel corner modules, a vehicle frame, a vehicle controller, and a power supply; The wheel corner module is composed of four corner modules: a left front corner module, a right front corner module, a left rear corner module, and a right rear corner module. Each corner module adopts the same structural design and is respectively installed at the left front, right front, left rear, and right rear of the vehicle frame, hinged to the vehicle frame. The corner module includes: a worm and worm gear reduction mechanism, a steering arm column, a steering angle sensor, a steering arm, a steering column, a brake system assembly, a hub motor, a lower control arm, a tire, a high-voltage wire harness, an upper control arm, a shock absorber, a high-voltage power supply, a mounting bracket, a low-voltage power supply, a low-voltage wire harness, a steering motor controller, a steering motor, a vehicle frame connecting piece, and a reducer; The steering motor controller and the housing of the steering motor are integrally designed. The steering motor controller outputs a signal to control the rotation of the steering motor. The output shaft of the steering motor is connected to the reducer, and the reducer is connected to the worm and worm gear reduction mechanism. The worm and worm gear reduction mechanism is rigidly connected to the steering arm column and transmits the torque of the steering motor to the steering arm; The lower end of the steering arm column is rigidly connected to the short side of the steering arm. The steering angle sensor is installed on the steering arm column to collect the steering angle signal of the steering arm column in real time and send the steering angle signal to the steering motor controller; The steering arm is connected to the steering column through a shock absorber, an upper control arm, and a lower control arm. The upper end of the steering column is respectively connected to the upper control arm and the lower end of the shock absorber, and the lower end of the steering column is connected to the upper end of the lower control arm. The upper end of the upper control arm, the upper end of the shock absorber, and the lower end of the lower control arm are respectively connected to the steering arm; The center position of the steering column is respectively connected to the brake system assembly, the hub motor, and the tire; The steering motor controller and the steering motor are both electrically connected to the low-voltage power supply through the low-voltage wire harness; The hub motor is electrically connected to the high-voltage power supply through the high-voltage wire harness; The high-voltage power supply and the low-voltage power supply are both fixed on the mounting bracket, and one end of the mounting bracket is rigidly connected to the short side of the steering arm through the vehicle frame connecting piece; The vehicle controller is installed in the front middle part of the vehicle, respectively connecting and controlling the left front wheel corner module, the right front wheel corner module, the left rear wheel corner module, and the right rear wheel corner module, and communicating with the steering motor controllers in the four corner modules respectively; The power supply module is fixedly installed at the middle rear of the vehicle frame, electrically connected to the high-voltage power supply and the low-voltage power supply, and respectively realizing the detection and control of the power of the high-voltage power supply and the low-voltage power supply.

2. The steer-by-wire four-wheel independent steering system according to claim 1, wherein The steering motor adopts a small torque and low voltage motor.

3. A method for evaluating the roll risk of a steer-by-wire four-wheel independent steering system, based on the system described in claim 1, characterized in that, The method steps are as follows: 1) Construct a vertical load distribution model of the steer-by-wire four-wheel independent steering system based on vehicle speed, vehicle body attitude, load, and road surface adhesion coefficient to monitor and feedback the time-varying and non-linear changes of the vertical load in real time; 2) Based on the vertical load calculation results in step 1), establish a real-time roll risk assessment model that integrates vehicle longitudinal acceleration, lateral acceleration, suspension dynamic response, and tire lateral force characteristics, design a dynamic compensation algorithm to correct the model, and generate a sequence of roll risk prediction values for future working conditions through time series analysis and prediction algorithms; 3) Combine the roll risk prediction value sequence in step 2) with the vertical load distribution change to establish a roll risk warning model, and generate safety warning signals of different levels based on the risk level determination mechanism.

4. The roll risk assessment method for the steer-by-wire four-wheel independent steering system according to claim 3, characterized in that, The specific content of step 1) includes: 11) Construct a vehicle roll dynamics model to describe the influence of vehicle speed, body attitude, load, and road adhesion coefficient on the vertical load; the vehicle roll dynamics model has the following force and moment balance equations; The lateral force balance equation is: where m is the vehicle mass, v x is the longitudinal driving speed of the vehicle, is the yaw rate of the vehicle's center of mass, β and γ are the sideslip angle and yaw rate of the vehicle's center of mass respectively, m s is the unsprung mass of the vehicle, h s is the height of the center of mass of the unsprung mass, is the roll acceleration of the vehicle body, k f and k r are the cornering stiffnesses of the front and rear tires of the vehicle respectively, l f and l r are the distances from the center of mass to the front axle and the rear axle respectively, R f and R r are the roll stiffnesses of the front and rear axles of the vehicle respectively, is the roll angle of the vehicle body, δ f and δ r are the steering angles of the front and rear wheels of the vehicle respectively; The yaw moment balance equation is: where I z is the moment of inertia of the vehicle about the z-axis, is the yaw rate of change of the vehicle, and I xz is the product of inertia of the vehicle about the x-axis and z-axis; The roll moment balance equation of the sprung mass is: where I x is the moment of inertia of the vehicle about the x-axis, is the roll damping coefficient, is the body roll stiffness coefficient, and g is the acceleration due to gravity; The pitch moment balance equation of the sprung mass is: Where, I y is the moment of inertia about the y-axis, is the vehicle body pitch angular acceleration, e is the height from the center of mass to the roll center, is the longitudinal speed change rate of the vehicle, v y is the lateral driving speed of the vehicle, k θ is the vehicle pitch stiffness coefficient, c θ is the pitch damping coefficient, θ is the vehicle body pitch angle, is the vehicle body pitch angular velocity; The moment balance equation of the sprung mass about the roll center is: where a ys is the lateral acceleration of the sprung mass, and is the angular velocity of vehicle body roll; The moment balance equation of the unsprung mass about the center axis of the tire contact point is: where F zfl is the vertical load of the left front wheel of the vehicle, F zfr is the vertical load of the right front wheel of the vehicle, F zrl is the vertical load of the left rear wheel of the vehicle, F zrr is the vertical load of the right rear wheel of the vehicle, B is the wheelbase of the vehicle, m d is the unsprung mass of the vehicle, h d is the centroid height of the unsprung mass, h sr is the height from the roll center to the ground, a y is the lateral acceleration of the vehicle, is the lateral speed change rate of the vehicle; The vertical force balance equation: In the formula, is the vertical vehicle speed change rate; 12) Based on the vehicle roll dynamics model in step 11), design a vertical load distribution model applicable to dynamic conditions, calculate and predict the vertical load distribution under different conditions in real time, and the calculation formulas for the vertical load of each tire are as follows: wherein, l is the distance from the front axle to the rear axle, l = l f + l r , B is the wheelbase of the vehicle, a x is the longitudinal acceleration of the vehicle, is the longitudinal speed change rate of the vehicle; 13) Based on the vehicle roll dynamics model in step 11) and the vertical load distribution model in step 12), monitor the change of the vertical load of the vehicle, and use model predictive control to adjust and optimize the calculation result of the vertical load to ensure real-time performance and accuracy; Select the longitudinal vehicle speed, lateral vehicle speed, vertical vehicle speed, roll angular velocity, pitch angular velocity, and yaw angular velocity as state variables, denoted as Regard the six state variables as the controlled outputs, denoted as Take the longitudinal resultant force, lateral resultant force, vertical resultant force, roll resultant moment, pitch resultant moment, and yaw resultant moment as the control inputs, denoted as u = [F xd , F yd , F zd , M xd , M yd , M zd , where Perform a force analysis on the vehicle to obtain the pitch moment M of the vehicle xd , the roll moment M yd , and the yaw moment M zd . The relationship with the tire forces is as follows: M yd = l f (F yfr + F yrr ) - l r (F yfl + F yrl ) Adopt a linear two-degree-of-freedom vehicle model as the reference model for the ideal motion of the vehicle body. Set the driver input to directly act on the steering wheel to generate the expectation of the driver's operation, and ignore the influence of the suspension system and load transfer during steering. Combine and simplify the left and right sides of the vehicle, and at the same time ignore the change of the vehicle body longitudinal speed. The state matrix of the linear two-degree-of-freedom vehicle model is as follows: wherein, are respectively the change rates of the expected values of the vehicle lateral speed and yaw rate; The expected value of the longitudinal speed is determined by using the given initial value of the initial speed and the acceleration curve. The expected value \(v\) of the longitudinal speed xd is the expected value \(a\) of the longitudinal acceleration xd integrated over time \(t\), as follows: where v x0 is the initial longitudinal speed of the vehicle, t is the integration time, and a xd (t) is the desired longitudinal acceleration; Take the vehicle pitch angle, pitch angular velocity, roll angle, and roll angular velocity expectation values as 0, as follows: where θ d , are the pitch angle and roll angle of the vehicle respectively, are the pitch angular velocity and roll angular velocity of the vehicle respectively; Write the longitudinal, lateral, and vertical desired vehicle speeds, as well as the yaw angular velocity, desired roll angular velocity, and desired pitch angular velocity of the vehicle in the form of state variables ; Write the vehicle roll dynamics model in the form of a continuous-time state space equation, which can be expressed as: y(t) = g(x(t), u(t)) In the formula, x(t) is the state vector of the system at time t, u(t) is the input vector of the system at time t, f(x(t), u(t)) is the state equation, g(x(t), u(t)) is the output equation, and y(t) is the output vector of the system at time t; Take the sampling time to discretize the continuous state space model. The discretized state space model is: x(k + 1) = F(x(k), u(k)) y(k) = G(x(k), u(k)) In the formula, x(k) represents the state vector at discrete time k, x(k + 1) represents the state vector at the next moment, u(k) is the input vector at discrete time k, F(x(k), u(k)) is the state equation at discrete time k, G(x(k), u(k)) is the output equation at discrete time k, and y(k) is the output vector at discrete time k; Define the prediction horizon as p and the control horizon as m, p > m. The state of the vehicle within the prediction horizon [k + 1, k + p] can be obtained based on the current state of the vehicle and the model predictive control process. At time k + p, the vehicle state is: x(k + p) = F(x(k), u(k), u(k + 1), …, u(k + m), …, u(k + p - 1)) where x(k) represents the state vector at discrete time k, u(k) is the input vector at discrete time k, u(k + 1) is the input vector at discrete time k + 1, u(k + m) is the input vector at discrete time k + m, u(k + p - 1) is the input vector at discrete time k + p - 1, and x(k + p) represents the state vector at the prediction horizon p; During the model predictive control process, the system's state and control inputs are optimized within a prediction horizon to minimize a certain performance index. When the prediction horizon p is greater than the control horizon m, i.e., when the sampling time is greater than the control horizon, the control input remains unchanged until the prediction horizon: u(k + m - 1) = u(k + m) = u(k + m + 1) = … = u(k + p - 1) where u(k + m - 1) is the input vector at discrete time k + m - 1, u(k + m) is the input vector at discrete time k + m, u(k + m + 1) is the input vector at discrete time k + m + 1, and u(k + p - 1) is the input vector at discrete time k + p - 1; Let the optimal control input and predicted output at time k be: where U(k) is the optimal control input vector at discrete time k, which contains the optimal control input sequence from the current time k to k + m - 1, u(k|k) is the optimal control input based on the information at the current time k at discrete time k, u(k + 1|k) is the optimal control input based on the information at the current time k at discrete time k + 1, u(k + m - 1|k) is the optimal control input based on the information at the current time k at discrete time k + m - 1, Y(k + 1|k) is the predicted output vector at discrete time k, which contains the predicted output sequence from k + 1 to k + p, r(k + 1|k) is the predicted output based on the information at the current time k at discrete time k + 1, r(k + 2|k) is the predicted output based on the information at the current time k at discrete time k + 2, and r(k + p|k) is the predicted output based on the information at the current time k at discrete time k + p; Considering the vehicle body attitude and suspension response, while realizing the real-time monitoring and optimization of the vehicle vertical load, the corresponding formula for the tire vertical load is: where a xs is the longitudinal acceleration of the sprung mass, h sp is the height from the pitch center to the ground.

5. The roll risk assessment method for the steer-by-wire four-wheel independent steering system according to claim 4, characterized in that, The specific steps of step 2) include: 21) Based on the vertical load calculation result in step 1), fuse the longitudinal acceleration, lateral acceleration, suspension dynamic response, and tire lateral force characteristics to establish a real-time roll risk assessment model; The vehicle lateral load transfer ratio LTR is used as the vehicle roll risk assessment index. The vehicle lateral load transfer ratio is defined as the ratio of the absolute value of the difference between the vertical loads of the left and right tires of the vehicle to the total tire vertical load, as follows: where F zl and F zr respectively represent the total vertical load of the left tires and the total vertical load of the right tires of the vehicle; 22) Based on the real-time roll risk assessment model in step 21), for the non-linear changes under high-speed driving, sharp turning and extreme driving conditions, use the unscented Kalman filter to estimate the vehicle body roll angle, roll angle speed and lateral acceleration, and design an adaptive dynamic compensation algorithm for the vehicle lateral load transfer ratio to correct the model; Based on the established vehicle roll dynamics model, an estimation model of the vehicle body roll angle based on the unscented Kalman filter is constructed. Let the state vector be The input vector is u1 = [δ f δ r T , and the output vector is ​ 23) Based on the corrected roll risk assessment model in step 22), use the long short-term memory model for time series analysis and prediction to generate a sequence of roll risk prediction values for future working conditions; The long short-term memory model consists of an input layer, a fully connected layer, an output layer, and two LSTM layers. The input layer is X = (X t-i+1 ,..., X t-2 , X t-1 , X t ), where X is the state vector containing the states from i moments before t. represents the vehicle body roll angle at time t vehicle body roll angle speed lateral acceleration a y , and the lateral acceleration a of the sprung mass ys four state variables; The output layer is which represents the predicted output sequence containing the LTR from t to i time steps ago; Each node in the LSTM layer is selected as an LSTM cell, and the structure of the LSTM cell includes four components: a forget gate, an input gate, a cell state, and an output gate; When the X vector is used as the unit input at time t, the formula for the forget gate is as follows: f t = σ(W f [h t-1 , X t + b f ) where f t is the output, W f is the input weight, b f is the bias, and σ(·) represents the sigmoid activation function; When the X vector is used as the unit input at time t, the formula for the input gate is as follows: i t = σ(W i [h t-1 , X t + b i ) where \(i\) t is the output, \(W\) i is the input weight, \(b\) i is the bias, and \(\sigma(\cdot)\) represents the sigmoid activation function; When the X vector is used as the unit input at time t, the formula for the cell state is as follows: In the formula, is the candidate cell state, W c is the input weight of the cell state, b c is the bias of the cell state, c t is the cell state at time t, c t-1 is the cell state at time t - 1, and tanh(·) represents the hyperbolic tangent function; When the X vector is used as the unit input at time t, the formula for the output gate is as follows: o t = σ(W o [h t-1 , X t + b o ) h t = o t tanh(c t ) where, o t is the output, W o is the input weight, b o is the bias; h t and h t-1 represent the hidden states at the current time and the previous time respectively; c t and c t-1 represent the cell states at the current time and the previous time respectively; σ(·) represents the sigmoid activation function, and tanh(·) represents the hyperbolic tangent function; The regression output layer needs to calculate the root mean square error loss of the regression. For a single observation, the root mean square error e RMSE The calculation formula is as follows: where η is the number of responses, and t i is the target output, and Y i is the predicted output.

6. The roll risk assessment method for the steer-by-wire four-wheel independent steering system according to claim 5, wherein The unscented Kalman filtering process in step 22) is as follows: 221) Generate the importance density function for each particle using the unscented Kalman filter, and construct the 2n+1 Sigma point sampling distributions χ of each particle i and the weight assignment formula is as follows: In the formula, n is the dimension of the state quantity; λ is the scaling factor, λ = α 2 (n + κ) - n, where α is a small positive value that determines the distribution state of the Sigma points; δ is a non - negative weight coefficient that can combine the dynamic errors of the high - order terms in the equation; is the state quantity of the i - th Sigma point at the (k - 1) - th moment; is the covariance of the i - th Sigma at the (k - 1) - th moment; is the weight distribution coefficient for approximating the mean; is the weight distribution coefficient for approximating the covariance; κ ∈ R, which characterizes the dispersion degree of the Sigma points relative to the mean. R is the set of real numbers. The larger κ is, the farther the Sigma points away from the mean are, and the smaller their weights are. Take κ = 3 - n; 222) Calculate the state quantity at time k with weighting The prior distribution is as follows: Wherein, is the prior estimate of the state quantity at time k, is the state quantity at time k, is the prior estimate of the covariance of the state quantity at time k, is the weight of the mean value of the state quantity of each Sigma point, is the weight of the covariance of the state quantity of each Sigma point, is the variance of the system noise at time k; 223) Perform Sigma point sampling on the prior distribution of the state quantity at time k and the probability distribution of the observed quantity are as follows: In the formula, is Gaussian white noise with zero mean and covariance R of the observed quantity, is the weight distribution coefficient of the approximate mean, and 2n is the number of Sigma points, is the observed quantity at time k, is the weighted mean of the observed quantity at time k; 224) Calculate the cross-covariance of the state quantity and the observed quantity and the autocovariance of the observed quantity as follows: Wherein, is the cross-covariance of the observed quantities, is the auto-covariance of the observed quantities, is the weight distribution coefficient of the approximate mean square error, is the observed quantity at time k, is the weighted mean of the observed quantity at time k, is the variance of the observation noise; 225) Calculate the Kalman gain of the state quantity at time k and update the state quantity as follows: where K i is the Kalman gain of the i-th Sigma point, is the mean of the state variables at time k, z k is the value of the observed quantity at time k, is the weighted mean of the observed quantity at time k, is the covariance estimate of the state variables at time k; Based on the vehicle body roll angle, roll angle speed and lateral acceleration estimated by the unscented Kalman filter according to steps 221)-225), the lateral load transfer ratio LTR is expressed as: Select the vehicle lateral acceleration and the vehicle body roll angle as the threshold parameters of the adaptive compensation strategy, and the formula is as follows: LTR * = LTR + ΔLTR where ΔLTR is the lateral load transfer rate compensation amount, and k ay , are the lateral acceleration compensation coefficient and the vehicle body roll angle compensation coefficient respectively, is the vehicle lateral acceleration threshold, is the vehicle body roll angle threshold.

7. The roll risk assessment method for the steer-by-wire four-wheel independent steering system according to claim 6, wherein The specific content of step 3) includes: 31) Based on the sequence of roll risk prediction values generated in step 23), construct a roll risk warning model based on the rollover warning time TTR. The rollover warning time TTR is defined as the time when the vehicle runs to roll over with the past period of time and the current moment state as the initial conditions. The roll risk warning model can be expressed as: TTR = t r - t0 where t r is the vehicle rollover time and t0 is the current time; 32) Based on the dynamic adjustment of the roll risk warning model constructed in step 31), the step size of the roll risk warning model is set to T s = the workload of a0, and the upper limit value of the rollover warning time TTR is T up = a. When the vehicle rollover warning time is within the upper limit value T up and the lateral load transfer ratio is not higher than the maximum value LTR max or lower than the minimum value LTR min , the vehicle is driving safely and will not roll over, and the next prediction is carried out; According to the step size T of the roll risk warning model s and the vehicle rollover dynamics model in step 1), calculate the lateral load transfer ratio of the vehicle. When the lateral load transfer ratio LTR reaches the rollover condition, the number of running steps N is obtained, and the rollover warning time is TTR = N * T s , where N is the number of running steps and T s is the warning step size; Use the lateral load transfer ratio LTR to dynamically adjust the threshold of the rollover warning time TTR. The specific adjustment method is as follows: When the lateral load transfer rate LTR exceeds the maximum value LTR max it indicates that the vehicle load transfer is severe, the roll angle speed increases, and the vehicle rollover risk is high. It is necessary to adjust the TTR threshold downward from a seconds to a - a h seconds to trigger the rollover warning earlier, where a h is the downward adjustment amount of the TTR threshold when the vehicle rollover risk increases; when the lateral load transfer rate LTR of the vehicle is lower than the minimum value LTR min the vehicle rollover risk is small, and the TTR threshold is adjusted upward to a + a l seconds to avoid the possibility of false alarms for the vehicle, where a l is the upward adjustment amount of the TTR threshold when the vehicle rollover risk decreases; when the lateral load transfer rate LTR of the vehicle is between the maximum value LTR max and the minimum value LTR min the TTR threshold remains unchanged at a; 33) Use the risk level determination mechanism that fuses multiple indicators established based on the lateral load transfer ratio LTR and the rollover warning time TTR to generate safety warning signals of different levels; According to the combined values of LTR and TTR, set the following different warning level mechanisms: Low risk: LTR < LTR min and TTR > a + a l ; Medium risk: LTR min ≤LTR≤LTR max or TTR < a; High risk: LTR > LTR max and TTR < a - a h ; Send different risk warning signals according to the risk warning results.

Images