A fault-tolerant control method for electric vehicle stability system actuator failure

Abstract

A fault-tolerant control method for actuator failure in an electric vehicle stability system includes the following steps: collecting vehicle driving state and actuator information; constructing a vehicle driving state feasible region using the tire force feasible region and the friction circle guidance range; calculating reference values ​​of state variables based on a vehicle reference model; using the reference values ​​and actual values ​​of the state variables as inputs to the upper-level vehicle motion controller; obtaining the resultant force and resultant torque of the current operating condition under the constraints of the vehicle driving state feasible region by model predictive control; allocating tire force under the constraints of the tire force feasible region; introducing tire force changes caused by actuator failure into the vehicle dynamic controller reconstruction; firstly obtaining the fault-tolerant feasible region of tire force and the fault-tolerant feasible region of the vehicle driving state after the failure; using the fault-tolerant feasible region as a new constraint for fault-tolerant control; finally, for situations where fault-tolerant control cannot control, reshaping the reference input to ensure the vehicle can drive safely to the maximum extent.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of vehicle active safety control, and provides an actuator fault tolerance control method for a four-wheel drive electric vehicle stability system, in particular to a fault tolerance control method for a four-wheel drive electric vehicle actuator fault causing the vehicle to lose stability due to limited vehicle driving space. BACKGROUND

[0002] With the continuous development of automatic driving, the complexity of the automobile system has greatly increased, which significantly increases the possibility of actuator and sensor failure. At the same time, due to the unexpected external disturbance such as load transfer and complex road conditions, the actuator or sensor may also fail. The occurrence of these faults greatly damages the stability of driving and seriously threatens the personal safety of passengers. Therefore, to ensure the driving safety of the automobile, the fault tolerance control of the automobile driving system must be performed.

[0003] The existing fault tolerance control algorithm or strategy lacks analysis of the vehicle driving space caused by the fault and the analysis of the change of the driving space is not comprehensive. At the same time, it is difficult to maintain the established state for the vehicle actuator fault, and the ability of the vehicle to change the established route or state according to the fault state and path information to make the vehicle safely drive or park is not clearly described. SUMMARY

[0004] The technical problem of the application is solved by calculating the tire force feasible region through the friction limit of the vehicle and the road surface and the actuator limit, and obtaining the whole vehicle driving state feasible region by using the tire force feasible region and the friction circle guidance range. When the actuator fails during the driving of the vehicle, the upper motion controller is constrained by the whole vehicle fault tolerance feasible region, and the lower distribution controller is reconstructed to distribute the tire force so that each tire force is within the feasible region. For the condition that the fault tolerance control cannot control, the vehicle state or driving route is adjusted by the reference input remodeling method to keep the vehicle safe operation.

[0005] The steps of the fault tolerance control method for the actuator fault of the electric vehicle stability system in the application are as follows:

[0006] Step 1: Real-time acquisition of the state quantity of the vehicle during driving by the vehicle-mounted sensor to obtain the longitudinal driving speed, lateral driving speed, yaw angular velocity, steering wheel angle, and state information including longitudinal acceleration, lateral acceleration, wheel speed, and road friction coefficient;

[0007] Step Two: Based on the information obtained in Step One, calculate the feasible region of tire force by determining the friction limit of the tire under vertical load changes and the actuator's own limitations; based on the state information, determine the friction circle guidance range; within the feasible region of tire force within the friction circle guidance angle range, calculate the virtual guidance angle to maximize and minimize the corresponding resultant force and resultant moment of the entire vehicle, i.e., obtain the virtual guidance angle. θ i Then, the resultant force and resultant torque at this point are the upper and lower bounds of the feasible region of the vehicle's driving state; the feasible region of the vehicle's driving state includes:

[0008] Longitudinal resultant force feasible region:

[0009]

[0010]

[0011]

[0012] Feasible region of lateral resultant force:

[0013]

[0014]

[0015]

[0016] Feasible region of yaw moment:

[0017]

[0018]

[0019]

[0020] Where F xi , F yi , θ i ,F xim max (i = 1, 2, 3, 4) represent the tire longitudinal force, the derivative of the tire longitudinal force, the tire lateral force, the derivative of the tire lateral force, the tire force angle, and the upper bound of the longitudinal force obtained from the motor torque, respectively (θ). yi ,θ xi ) represents the guiding range of the friction circle, μ represents the road surface friction coefficient, and F zi For tire load, F mxi , F myi These are the virtual steering angles at the upper and lower bounds of the feasible region of tire force, respectively. θi The corresponding extreme values ​​of longitudinal and lateral forces, δ i , θ i These are the tire steering angle, the upper virtual guide angle, and the lower virtual guide angle calculated from the guide range, respectively.

[0021] Step 3: Based on the state information obtained in Step 1, use the vehicle reference model to obtain the reference value of the vehicle's driving state quantity, namely the vehicle's longitudinal reference speed V. xre Vehicle lateral reference speed V yre and the vehicle's reference yaw rate ω rre ;

[0022]

[0023] In the formula V0,δ,a xre These represent the initial values ​​of the longitudinal velocity of the reference model, the front wheel steering angle, and the given values ​​of the longitudinal acceleration, respectively. δ is the front wheel steering angle, and l is the given value. f l is the distance from the vehicle's center of gravity to the front wheel. r L is the distance from the vehicle's center of gravity to the rear wheel, and K is the vehicle's wheelbase. f K represents the front wheel lateral stiffness. r Let ζ be the rear wheel lateral stiffness of the vehicle, and ζ be the neglected vehicle a. xre V yre The compensating dimensionless coefficient of the differential.

[0024] Step 4: Establish a vehicle model. The nonlinear vehicle model is transformed into a linear model using a direct discretization linearization method. The deviation between the reference and actual values ​​of the state variables is used as the input to the upper-level vehicle motion controller. The longitudinal resultant force, lateral resultant force, and yaw moment (i.e., virtual control variables) under the current operating condition are obtained through a model predictive control algorithm. The objective function is as follows:

[0025]

[0026] Where N p For prediction in the time domain, N c For the control time domain, Δu(k+i|k) is the virtual control quantity, Q, R, ρ are weighting factors, and η(k+i|k), η ref (k+i|k) represent the output state variable and the reference value, respectively, and ε is a relaxation factor that can prevent the problem from being unsolvable.

[0027] Control Incremental Constraint: ΔU min ≤Δu≤ΔU max The upper limit of the adjustable increment given to the actuator.

[0028] The lower-level distribution controller allocates the virtual control input from the upper-level controller to each wheel. The objective function is as follows:

[0029]

[0030]

[0031] Among them, W 4×4 C represents a four-dimensional symmetric matrix. i (i = 1, 2, 3, 4) are the diagonal elements in the matrix, where the numerator represents the weighting coefficient for each tire, expressed as: B 3×8 The control efficiency matrix can be obtained from the overall equilibrium equation of the vehicle model, expressed as:

[0032]

[0033] u d For the control vector, u dmax ,u dmin These represent the upper and lower bounds of the longitudinal and lateral forces of the tire within the feasible tire force domain, respectively. d =[F x1 F x2 F x3 F x4 F y1 F y2 F y3 F y4 ] T u is the virtual control vector, u = [∑F x ∑F y ∑M z ] T .

[0034] Step 5: Ignore the impact of the fault on lateral forces. Transform the impact of the actuator fault on the entire vehicle, i.e., its effect on the vehicle's electrical braking and driving, into the trend of tire force changes F on the wheels. xim =λ i F xi +q i (i = 1, 2, 3, 4, λ) i The system is introduced into the vehicle system (∈[0,1]). The feasible region of tire force is recalculated based on the influence of tire force changes on braking. This is called the tire force fault-tolerant feasible region, which is then used to obtain the tire force distribution constraints after a fault. The feasible region of the vehicle driving state is recalculated based on the tire force fault-tolerant feasible region. This is called the vehicle fault-tolerant feasible region. The vehicle fault-tolerant feasible region includes:

[0035] Longitudinal force fault-tolerant feasible region:

[0036]

[0037]

[0038]

[0039] Lateral resultant force fault-tolerant feasible region:

[0040]

[0041]

[0042]

[0043] The yaw moment tolerance feasible region:

[0044]

[0045]

[0046]

[0047] in F fxi , F fyi (i = 1, 2, 3, 4) represent the virtual steering angles at the upper and lower bounds of the feasible region for tire force tolerance, respectively. The corresponding extreme values ​​of longitudinal and lateral forces.

[0048] Step Six: Fault-Tolerant Controller Design. First, the control constraints are changed by adjusting the predictive control conditions of the upper-level model, i.e., from the feasible region of the vehicle's driving state to the fault-tolerant feasible region of the vehicle.

[0049]

[0050]

[0051]

[0052] Secondly, the allocation matrix of the lower-level control is reconstructed to optimize the distribution of tire force so that it meets the fault-tolerant feasible region of tire force. The virtual control quantity is adjusted as follows:

[0053] Its control efficiency matrix changes as follows:

[0054]

[0055] Step 7: When one or more of the virtual control quantities output by the upper-level controller, namely the longitudinal resultant force, lateral resultant force, and yaw resultant moment, are in n TIf the vehicle remains on the boundary of the fault-tolerant feasible region for ≥20 consecutive sampling periods T, it is considered a serious fault. For situations where the vehicle is determined to be a serious fault within the fault-tolerant feasible region, the relationship matrix between the control quantity and the state quantity obtained from model predictive control is used as the basis. Taking into account the vehicle's execution capability and tracking effect, a reference input reshaping scheme is designed based on the vehicle's fault-tolerant feasible region and tracking deviation. The tracking deviation is the difference between the reference input obtained from the reference model formula (1) and the actual state quantity of the vehicle collected in step one. The objective function of the reshaping scheme is as follows:

[0056]

[0057] Where Y d Reshaping reference input, Y re As the original reference input, U b =U tb -U l -Δu represents the degree of approximation of the virtual control variable to the feasible region, U l For the previous virtual control quantity, U tb To be with U l Let the feasible region boundary values ​​with the same sign be...

[0058] The prediction formula (37) Y(t)=ψ is derived from the model predictive controller. t ζ(t)+Θ t Δu(t)+ι t Ψ(t)(Let: Φ=ψ t ζ(t)+ι t The reshaping constraints are obtained by adjusting Ψ(t) and the control constraints (38) of the model predictive controller as follows:

[0059]

[0060]

[0061] U max U min Let ΔU be the upper and lower bounds of the fault-tolerant feasible region of the entire vehicle. max ΔU min Given the upper and lower limits of the control increment for the predictor controller, A I The expression (38) is the transformation matrix for converting control quantity constraints into control increment constraints. To reverse it.

[0062] The final calculated reference input after reshaping is:

[0063]

[0064] Where E is a matrix with the same number of rows and columns as Y, and all elements are one. t F For Θ t The reverse.

[0065] Based on the reshaping scheme obtained from the reference input, the reshaping state is obtained, namely the reshaping longitudinal velocity, the reshaping lateral velocity, and the reshaping yaw rate. The equation between the longitudinal reference velocity and the yaw rate in the reference model formula (1) is discussed. If the calculated longitudinal velocity and yaw rate do not conform to the reference model formula (1) and... Then, the steady-state steering model is used to calculate the steering angle to maintain the original route. The system adjusts the target input to the control system by calculating the steering angle based on the reference model, changing the original route, reducing speed, and altering the front wheel steering angle to allow the car to change its trajectory while reducing performance. Where ω... rc V xc To reshape the yaw rate and the longitudinal rate, ω rvx The yaw rate is calculated from the longitudinal velocity of the reshaped model using formula (1) of the reference model.

[0066] The feasible region of tire force calculated in step two is:

[0067]

[0068] Friction circle guide range (θ) yi ,θ xi ):

[0069]

[0070] Step four establishes the vehicle model as a three-degree-of-freedom model involving longitudinal, lateral, and yaw motions of the vehicle body. The model is based on the following assumptions:

[0071] Ⅰ: Ignore the influence of the steering system and use the front wheel angle as the system input.

[0072] II: Ignoring the role of the suspension, the vehicle body only performs planar motion parallel to the ground.

[0073] III: The vehicle's speed along the axial direction remains constant.

[0074] The model is expressed as shown in the equilibrium equations:

[0075]

[0076] In the above equation, m is the total mass of the car, and V... x V is the longitudinal velocity. y Let ω be the lateral velocity. r I is the yaw rate. zLet F be the moment of inertia of the entire vehicle about the Z-axis. x ,∑F y ,∑M z These are the longitudinal resultant force, lateral resultant force, and yaw moment acting on the vehicle, respectively.

[0077] Discretize and linearize the three-degree-of-freedom vehicle dynamics model:

[0078]

[0079]

[0080]

[0081] Where T is the sampling period.

[0082] The discrete state equations are obtained as follows:

[0083]

[0084] Where A, B, and C are respectively:

[0085]

[0086]

[0087] Further analysis revealed:

[0088]

[0089] Where d k Indicates the calculation deviation. These are the observed values.

[0090] Constructing the state vector:

[0091]

[0092] We obtain the new state-space expression:

[0093]

[0094] The state is predicted based on the state-space expression, and the system output is obtained from the predicted state. The future output Y(t) of the system is represented by the following matrix:

[0095] Y(t)=ψ t ζ(t)+Θ t Δu(t)+ι t Ψ(t) (37)

[0096] in:

[0097]

[0098]

[0099] N p To predict the step size, N c To control the step size;

[0100] Control constraints:

[0101]

[0102] in:

[0103]

[0104] ∑F xt max ,∑F yt max ,∑M zt max ∑F xt min ,∑F yt min ,∑M zt min Let ΔU be the upper and lower bounds of the feasible region corresponding to the upper and lower limits of the longitudinal resultant force, lateral resultant force, and yaw moment, respectively. max ΔU min Predict the upper and lower limits of the control increment for a given model.

[0105] In step five, the tire force change caused by actuator failure is added to the tire force feasible region to obtain the tire force fault-tolerant feasible region.

[0106] Tire force tolerance feasible region:

[0107]

[0108] Simultaneously, the guide range of the friction circle after the fault is obtained, the guide range of the fault friction circle (θ) fyi ,θ fxi ):

[0109]

[0110] In step seven, the reference input is reshaped based on the severe faults identified in the vehicle's fault-tolerant feasible region. Since the reference input reshaping can only provide the dynamic state value of the vehicle's stable driving (i.e., the vehicle speed will always decrease during the consideration of the feasible region), and cannot determine whether the vehicle can drive stably at a certain steady-state speed or constant speed, the following reshaping criteria should be considered when determining the final steady-state speed in consideration of the vehicle's driving capability:

[0111] (1) Maintain or adjust the trajectory during driving under fault conditions so that the vehicle speed and steering angle meet the requirements of the trajectory curvature:

[0112]

[0113] That is, the vehicle's longitudinal velocity and yaw rate meet the curvature change requirements.

[0114] (2) During the reshaping process, the longitudinal resultant force, lateral resultant force, and yaw moment of the vehicle must always be kept within the feasible range, i.e.:

[0115]

[0116]

[0117]

[0118] (3) Determination of vehicle stability: The longitudinal force, lateral force and yaw moment of the vehicle have a certain stability margin χ∈[0,1], that is, the vehicle has a certain driving ability under the new state to avoid instability under the continuous aggravation of the fault:

[0119]

[0120] The beneficial effects of this invention are:

[0121] This invention, based on the feasible domain of tire force, describes the vehicle's driving state through the friction circle guidance range, calculating the feasible domain of the overall vehicle driving state. For tire force changes caused by actuator failure, a new fault-tolerant feasible domain of tire force is calculated, and the corresponding fault-tolerant feasible domain of the entire vehicle is derived. This fault-tolerant feasible domain is then combined with the upper-level vehicle model predictive controller, along with the reconstruction of the lower-level distribution controller, to achieve fault-tolerant control, enabling the vehicle to drive stably even under actuator failure. For faults that are difficult to control with fault-tolerant control, a reference input reshaping scheme is employed to adjust the vehicle state or path, allowing the vehicle to drive or stop safely. In summary, this method is of significant importance to the driving safety of vehicles in assisted and autonomous driving systems. Attached Figure Description

[0122] Figure 1 This is a schematic diagram of the fault-tolerant control principle proposed in this invention;

[0123] Figure 2 For simulation (1) longitudinal velocity comparison diagram of fault-tolerant control;

[0124] Figure 3 For simulation (1) comparison of fault-tolerant control lateral velocity;

[0125] Figure 4 For simulation (1) comparison of fault-tolerant control yaw rate;

[0126] Figure 5 To simulate (1) the longitudinal resultant force and feasible domain diagram of fault-tolerant control;

[0127] Figure 6To simulate (1) the resultant lateral force and feasible domain diagram of fault-tolerant control;

[0128] Figure 7 To simulate (1) the resultant moment of yaw and feasible domain diagram of fault-tolerant control;

[0129] Figure 8 To simulate (1) the longitudinal resultant force and feasible domain diagram of fault-tolerant control;

[0130] Figure 9 To simulate (1) the resultant lateral force and feasible domain diagram of fault-tolerant control;

[0131] Figure 10 To simulate (1) the resultant moment of yaw under fault-tolerant control and the feasible domain diagram;

[0132] Figure 11 For simulation (2) longitudinal velocity comparison diagram of fault-tolerant control;

[0133] Figure 12 For simulation (2) comparison of fault-tolerant control lateral velocity;

[0134] Figure 13 For simulation (2) comparison of fault-tolerant control yaw rate;

[0135] Figure 14 To simulate (2) the longitudinal resultant force and feasible domain diagram of fault-tolerant control;

[0136] Figure 15 To simulate (2) the resultant lateral force and feasible domain diagram of fault-tolerant control;

[0137] Figure 16 To simulate (2) the resultant moment of yaw and feasible domain diagram of fault-tolerant control;

[0138] Figure 17 To simulate (2) the longitudinal resultant force and feasible domain diagram of fault-tolerant control;

[0139] Figure 18 To simulate (2) the resultant lateral force and feasible domain diagram of fault-tolerant control;

[0140] Figure 19 To simulate (2) the resultant moment of yaw under fault-tolerant control and the feasible domain diagram;

[0141] Figure 20 For simulation (2), refer to the longitudinal velocity comparison diagram after reshaping;

[0142] Figure 21 The simulation (2) references the comparison of lateral velocities after reshaping.

[0143] Figure 22 For simulation (2), the comparison diagram of yaw rate after reshaping is used as the reference input;

[0144] Figure 23 For simulation (2), the longitudinal resultant force and feasible domain diagram after reshaping are referenced inputs;

[0145] Figure 24 For simulation (2), the lateral resultant force and feasible domain diagram after reshaping are used as reference inputs;

[0146] Figure 25 For simulation (2), the yaw resultant moment and feasible domain diagram after reshaping are referenced inputs; Detailed Implementation

[0147] The proposed timely solution will be further elaborated and explained below with reference to the accompanying drawings.

[0148] Fault-tolerant control principle diagram as follows Figure 1 As shown, this invention proposes a fault-tolerant control method for actuators in electric vehicle stability systems, implemented according to the following steps:

[0149] Step 1: Collect real-time information on vehicle driving status and identify the friction coefficient of the road surface.

[0150] The system collects vehicle driving status information fed back from onboard sensors, including steering wheel angle signals from a steering wheel angle sensor, yaw rate signals from a yaw rate sensor, acceleration from a accelerometer, wheel speed information from magnetoelectric or Hall effect sensors, and transmission output shaft speed from an electromagnetic induction speed sensor. Vehicle speed is then calculated from the vehicle speed sensor signal. After specific filtering, the system obtains vehicle status information such as steering wheel angle, yaw rate, and acceleration. Based on the vehicle speed and wheel speed, the system calculates the tire longitudinal slip ratio and center-of-gravity sideslip angle, and uses existing estimation algorithms to identify the road surface friction coefficient, i.e., the road adhesion coefficient μ.

[0151] Step Two: Based on the information obtained in Step One, calculate the tire force feasible region by determining the tire's friction limit under varying vertical loads and the actuator's own limitations; determine the friction circle guiding range based on the state information.

[0152] Tire force feasible range:

[0153]

[0154] Friction circle guide range (θ) yi ,θ xi ):

[0155]

[0156] Within the feasible region of tire force in the friction circle guide angle range, calculate the virtual guide angle to maximize and minimize the corresponding resultant force and resultant moment of the entire vehicle, i.e., obtain the virtual guide angle. θ i Then the resultant force and resultant torque at this time are the upper and lower bounds of the feasible domain of the overall driving state.

[0157] The feasible domain for the overall vehicle driving status includes:

[0158] Longitudinal resultant force feasible region:

[0159]

[0160]

[0161]

[0162] Feasible region of lateral resultant force:

[0163]

[0164]

[0165]

[0166] Feasible region of yaw moment:

[0167]

[0168]

[0169]

[0170] Where F xi , F yi , θ i ,F ximmax (i = 1, 2, 3, 4) represent the tire longitudinal force, the derivative of the tire longitudinal force, the tire lateral force, the derivative of the tire lateral force, the tire force angle, and the upper bound of the longitudinal force obtained from the motor torque, respectively (θ). yi ,θ xi ) represents the guiding range of the friction circle, μ represents the road surface friction coefficient, and F zi For tire load, F mxi , F myi These are the virtual steering angles at the upper and lower bounds of the feasible region of tire force, respectively. θ i The corresponding extreme values ​​of longitudinal and lateral forces, δ i , θ iThese are the tire steering angle, the upper virtual guide angle, and the lower virtual guide angle calculated from the guide range, respectively.

[0171] Step 3: Based on the state variables obtained in real time in Step 1, use the vehicle reference model to determine the reference values ​​of the control state variables for the vehicle's planar motion, including: the vehicle's longitudinal reference velocity V. xre Vehicle lateral reference speed V yre and the vehicle's reference yaw rate ω rre .

[0172]

[0173] In the formula V0,δ,a xre These represent the initial values ​​of the longitudinal velocity of the reference model, the front wheel steering angle, and the given values ​​of the longitudinal acceleration, respectively. δ is the front wheel steering angle, and l is the given value. f l is the distance from the vehicle's center of gravity to the front wheel. r L is the distance from the vehicle's center of gravity to the rear wheel, and K is the vehicle's wheelbase. f K represents the front wheel lateral stiffness. r Let ζ be the rear wheel lateral stiffness of the vehicle, and ζ be the neglected vehicle a. xre V yre The compensating dimensionless coefficient of the differential.

[0174] Step 4: Establish a whole vehicle model. Use discrete linearization to obtain a linear vehicle model. Then, use the deviation between the expected value and the actual value of the state variables as the input of the controller. The upper-level motion controller uses a model predictive controller to obtain virtual control variables, namely longitudinal resultant force, lateral resultant force and yaw moment.

[0175] A three-degree-of-freedom model involving the longitudinal, lateral, and yaw motions of the vehicle body is established. The model is based on the following assumptions:

[0176] Ⅰ: Ignore the influence of the steering system and use the front wheel angle as the system input.

[0177] II: Ignoring the role of the suspension, the vehicle body only performs planar motion parallel to the ground.

[0178] III: The vehicle's speed along the axial direction remains constant.

[0179] The model is expressed as shown in the equilibrium equations:

[0180]

[0181] In the above equation, m is the total mass of the car, and V... x V is the longitudinal velocity. y Let ω be the lateral velocity. r I is the yaw rate.z Let F be the moment of inertia of the entire vehicle about the Z-axis. x ,∑F y ,∑M z These are the longitudinal resultant force, lateral resultant force, and yaw moment acting on the vehicle, respectively.

[0182] Discretize and linearize the three-degree-of-freedom vehicle dynamics model:

[0183]

[0184]

[0185]

[0186] The discrete state equations are obtained as follows:

[0187]

[0188] Where A, B, and C are respectively:

[0189]

[0190]

[0191] Further analysis revealed:

[0192]

[0193] Where d k Indicates the calculation deviation. These are the observed values.

[0194] Constructing the state vector:

[0195]

[0196] We obtain the new state-space expression:

[0197]

[0198] The state is predicted based on the state-space expression, and the system output is obtained from the predicted state. The future output Y(t) of the system is represented by the following matrix:

[0199] Y(t)=ψ t ζ(t)+Θ t Δu(t)+ι t Ψ(t) (37)

[0200] in:

[0201]

[0202]

[0203] Prediction step size N p =50, control step size N c =5.

[0204] Control constraints:

[0205]

[0206] in:

[0207]

[0208] ∑F xt max ,∑F yt max ,∑M zt max ∑F xt min ,∑F yt min ,∑M zt min Let ΔU be the upper and lower bounds of the feasible region corresponding to the upper and lower limits of the longitudinal resultant force, lateral resultant force, and yaw moment, respectively. max ΔU min Predict the upper and lower limits of the control increment for a given model.

[0209] Objective function:

[0210]

[0211] Transform it into a standard quadratic form:

[0212] J=[Δu Τ ε] Τ H t [Δu Τ ε]+G t [Δu Τ ε] (39)

[0213] in:

[0214]

[0215] The controller outputs the control quantity u(k) as follows:

[0216] u(k)=u(k-1)+Δu (40)

[0217] The objective function for distributing virtual control quantities to each wheel using a lower-level distribution controller is as follows:

[0218]

[0219]

[0220] Among them, W4×4 C represents a four-dimensional symmetric matrix. i (i = 1, 2, 3, 4) are the diagonal elements in the matrix, where the numerator represents the weighting coefficient for each tire, expressed as: B 3×8 The expression for the control efficiency matrix is ​​as follows:

[0221]

[0222] u d For the control vector, u dmax ,u dmin These represent the upper and lower bounds of the longitudinal and lateral forces of the tire within the feasible tire force domain, respectively. d =[F x1 F x2 F x3 F x4 F y1 F y2 F y3 F y4 ] T u is the virtual control vector, u = [∑F x ∑F y ∑M z ] T .

[0223] Step 5: Add the tire force change caused by actuator failure to the tire force feasible region to obtain the tire force fault-tolerant feasible region. At the same time, obtain the friction circle guidance range after the failure and calculate the vehicle fault-tolerant feasible region.

[0224] Tire force tolerance feasible region:

[0225]

[0226] Fault friction circle guide range (θ) fyi ,θ fxi ):

[0227]

[0228] Vehicle fault tolerance feasible domain:

[0229] Longitudinal force fault-tolerant feasible region:

[0230]

[0231]

[0232]

[0233] Lateral resultant force fault-tolerant feasible region:

[0234]

[0235]

[0236]

[0237] The yaw moment tolerance feasible region:

[0238]

[0239]

[0240]

[0241] in F fxi , F fyi (i = 1, 2, 3, 4) represent the virtual steering angles at the upper and lower bounds of the feasible region for tire force tolerance, respectively. The corresponding extreme values ​​of longitudinal and lateral forces.

[0242] Step Six: Design fault-tolerant control for actuator failure in a four-wheel drive electric vehicle. Based on the current road conditions and vehicle driving status information, determine the relationship between tire force and feasible region after a failure occurs, and select an optimization method according to the following steps:

[0243] (1) Apply the fault-tolerant feasible domain of the whole vehicle to the upper-level motion controller of the vehicle, that is, achieve the effect of fault tolerance by changing the control constraints predicted by the model;

[0244] The control constraints for the fault-tolerant feasible region of the whole vehicle are as follows:

[0245]

[0246]

[0247]

[0248] (2) Reconstruct the allocation matrix of the lower-level control to optimize the tire force allocation so that it meets the tire force fault-tolerant feasible region. The virtual control quantity is adjusted as follows:

[0249] Its control efficiency matrix changes as follows:

[0250]

[0251] Step 7: Based on the serious faults identified in the vehicle's fault tolerance feasible domain, perform reference input reshaping to adjust the vehicle status or driving route to ensure safe driving or parking.

[0252] For situations where the vehicle's fault-tolerant feasible region is determined to be a severe fault, based on the relationship matrix between the control quantity and the state quantity obtained from model predictive control, and taking into account the vehicle's execution capability and tracking performance, a reference output reshaping scheme is designed with the fault-tolerant feasible region and tracking deviation as the objective function:

[0253]

[0254] Where Y d Reshaping reference input, Y re As the original reference input, U b =U tb -U l -Δu represents the degree of approximation of the virtual control variable to the feasible region, U l For the previous virtual control quantity, U tb To be with U l Let the feasible region boundary values ​​with the same sign be...

[0255] The prediction formula (37) Y(t)=ψ is derived from the model predictive controller. t ζ(t)+Θ t Δu(t)+ι t Ψ(t)(Let: Φ=ψ t ζ(t)+ι t The reshaping constraints are obtained by adjusting Ψ(t) and the control constraints (38) of the model predictive controller as follows:

[0256]

[0257]

[0258] U max U min Let ΔU be the upper and lower bounds of the fault-tolerant feasible region of the entire vehicle. max ΔU min Given the upper and lower limits of the control increment for the predictor controller, A I The expression (38) is the transformation matrix for converting control quantity constraints into control increment constraints. To reverse it.

[0259] The final calculated reference input after reshaping is:

[0260]

[0261] Where E is a matrix with the same number of rows and columns as Y, and all elements are one. t F For Θ t The reverse.

[0262] Based on the reshaping scheme obtained from the reference input, the reshaping state is obtained, namely the reshaping longitudinal velocity, the reshaping lateral velocity, and the reshaping yaw rate. The equation between the longitudinal reference velocity and the yaw rate in the reference model formula (1) is discussed. If the calculated longitudinal velocity and yaw rate do not conform to the reference model formula (1) and... Then, the steady-state steering model is used to calculate the steering angle to maintain the original route. The system adjusts the target input to the control system by calculating the steering angle based on the reference model, changing the original route, reducing speed, and altering the front wheel steering angle to allow the car to change its trajectory while reducing performance. Where ω... rc V xc To reshape the yaw rate and the longitudinal rate, ω rvx The yaw rate is calculated from the longitudinal velocity of the reshaped model using formula (1) of the reference model.

[0263] When reshaping the reference input based on the severe faults identified in the vehicle's fault-tolerant feasible region, the reference input reshaping can only provide the dynamic state value of the vehicle's stable driving, i.e., the vehicle speed will always decrease during the consideration of the feasible region, and cannot determine whether the vehicle can drive stably at a certain steady-state speed or constant speed. Therefore, when considering the vehicle's driving capability, the determination of the final steady-state speed should consider the following reshaping criteria:

[0264] (1) Maintain or adjust the trajectory during driving under fault conditions so that the vehicle speed and steering angle meet the requirements of the trajectory curvature:

[0265]

[0266] That is, the vehicle's longitudinal velocity and yaw rate meet the curvature change requirements.

[0267] (2) During the reshaping process, the longitudinal resultant force, lateral resultant force, and yaw moment of the vehicle must always be kept within the feasible range, i.e.:

[0268]

[0269]

[0270]

[0271] (3) Determination of vehicle stability: The longitudinal force, lateral force and yaw moment of the vehicle have a certain stability margin χ∈[0,1], that is, the vehicle has a certain driving ability under the new state to avoid instability under the continuous aggravation of the fault:

[0272]

[0273] The simulation experimental data of the technical solution provided by this invention are given below.

[0274] (1) Under the single lane change driving condition of the vehicle, given the desired speed of the vehicle of 22.2 m / s and the front wheel steering angle of the sinusoidal input. With a friction coefficient μ = 0.8, the vehicle maintains safe and stable operation. When the actuators of the right front and right rear wheels of the vehicle stability control system simultaneously fail within 2.5 seconds, the failure coefficient of the failed tires is λ. i =0.3, the fault value for tire force jamming is q i =320N. At this point, due to the fault, the overall resultant force and resultant torque of the vehicle exceed the vehicle's fault-tolerant feasible range, and the system cannot continue to operate. The vehicle loses stability, but after fault-tolerant control, the vehicle can continue to drive. Simulation verification shows that the vehicle's instability state variables at this point are as follows: Figures 2-4 And the resultant force and resultant torque of the whole vehicle, such as Figures 7-10 .

[0275] (2) Under the single lane change driving condition of the vehicle, given the desired speed of the vehicle of 22.2 m / s and the sinusoidal input front wheel angle With a friction coefficient μ = 0.8, the vehicle maintains safe and stable operation. However, if the actuators of the right front and right rear wheels of the vehicle stability control system simultaneously fail within 2.5 seconds, the failed wheels will completely fail, meaning their failure coefficients will both be λ. i =0, the fault value for tire force jamming is q i =320N. At this point, due to the fault, the overall resultant force and resultant torque of the vehicle exceed the vehicle's fault tolerance range, the system cannot continue to operate, and the vehicle loses stability. Although the vehicle's operating condition improves somewhat after fault-tolerant control, it is still difficult to continue stable operation. The comparison between its fault-tolerant and non-fault-tolerant states is as follows: Figures 11-13 And the resultant force and resultant moment, such as Figures 14-9 The reference input is reshaped, and simulation verification shows that the vehicle state variables are now, for example... Figures 20-22 And the resultant force and resultant torque of the whole vehicle, such as Figures 23-25 .

[0276] This invention is based on the feasible domain of tire force. It describes the vehicle's driving state through the friction circle guidance range and calculates the feasible domain of the overall vehicle driving state. For tire force changes caused by actuator failure, a fault-tolerant feasible domain of tire force is recalculated, and the corresponding fault-tolerant feasible domain of the entire vehicle is derived. This fault-tolerant feasible domain is then combined with the upper-level vehicle model predictive controller and the reconstruction of the lower-level distribution controller to achieve fault-tolerant control. For faults that are difficult to control with fault-tolerant control, a reference input reshaping scheme is adopted to adjust the vehicle state or path so that the vehicle can drive or stop safely. In summary, this method has significant implications for vehicle driving safety in assisted driving and autonomous driving.

Claims

1. A fault-tolerant control method for actuator failure in an electric vehicle stability system, characterized in that, The steps of this method are as follows: Step 1: Collect the vehicle's state parameters in real time through onboard sensors to obtain the vehicle's longitudinal speed, lateral speed, yaw rate, and steering wheel angle, as well as state information including longitudinal acceleration, lateral acceleration, wheel speed, and road friction coefficient. Step Two: Based on the information obtained in Step One, calculate the feasible region of tire force by determining the friction limit of the tire under vertical load changes and the actuator's own limitations; based on the state information, determine the friction circle guidance range; within the feasible region of tire force within the friction circle guidance angle range, calculate the virtual guidance angle to maximize and minimize the corresponding resultant force and resultant moment of the entire vehicle, i.e., obtain the virtual guidance angle. θ i Then, the resultant force and resultant torque at this point are the upper and lower bounds of the feasible region of the vehicle's driving state; the feasible region of the vehicle's driving state includes: Longitudinal resultant force feasible region: Feasible region of lateral resultant force: Feasible region of yaw moment: Where F xi , F yi , θ i ,F ximmax (i = 1, 2, 3, 4) represent the tire longitudinal force, the derivative of the tire longitudinal force, the tire lateral force, the derivative of the tire lateral force, the tire force angle, and the upper bound of the longitudinal force obtained from the motor torque, respectively (θ). yi ,θ xi ) represents the guiding range of the friction circle, μ represents the road surface friction coefficient, and F zi For tire load, F mxi , F myi These are the virtual steering angles at the upper and lower bounds of the feasible region of tire force, respectively. θ i The corresponding extreme values ​​of longitudinal and lateral forces, δ i , θ i These are the tire steering angle, the upper virtual guide angle, and the lower virtual guide angle calculated from the guide range, respectively. Step 3: Based on the state information obtained in Step 1, use the vehicle reference model to obtain the reference value of the vehicle's driving state quantity, namely the vehicle's longitudinal reference speed V. xre Vehicle lateral reference speed V yre and the vehicle's reference yaw rate ω rre ; In the formula Where V0,δ,a xre These represent the initial values ​​of the longitudinal velocity of the reference model, the front wheel steering angle, and the given values ​​of the longitudinal acceleration, respectively. δ is the front wheel steering angle, and l is the given value. f l is the distance from the vehicle's center of gravity to the front wheel. r L is the distance from the vehicle's center of gravity to the rear wheel, and K is the vehicle's wheelbase. f K represents the front wheel lateral stiffness. r Let ζ be the rear wheel lateral stiffness of the vehicle, and ζ be the neglected vehicle a. xre V yre The compensating dimensionless coefficient of the differential; Step 4: Establish a vehicle model. The nonlinear vehicle model is transformed into a linear model using a direct discretization linearization method. The deviation between the reference and actual values ​​of the state variables is used as the input to the upper-level vehicle motion controller. The longitudinal resultant force, lateral resultant force, and yaw moment (i.e., virtual control variables) under the current operating condition are obtained through a model predictive control algorithm. The objective function is as follows: Where N p For prediction in the time domain, N c For the control time domain, Δu(k+i|k) is the virtual control quantity, Q, R, ρ are weighting factors, and η(k+i|k), η ref (k+i|k) represent the output state variable and the reference value, respectively, and ε is the relaxation factor that can avoid the problem being unsolvable. Control Incremental Constraint: ΔU min ≤Δu≤ΔU max The upper limit of the adjustable increment of the actuator; The lower-level distribution controller allocates the virtual control input from the upper-level controller to each wheel. The objective function is as follows: Among them, W 4×4 C represents a four-dimensional symmetric matrix. i (i = 1, 2, 3, 4) are the diagonal elements in the matrix, where the numerator represents the weighting coefficient for each tire, expressed as: B 3×8 The control efficiency matrix can be obtained from the overall equilibrium equation of the vehicle model, expressed as: u d For the control vector, u dmax ,u dmin These represent the upper and lower bounds of the longitudinal and lateral forces of the tire within the feasible tire force domain, respectively. d =[F x1 F x2 F x3 F x4 F y1 F y2 F y3 F y4 ] T u is the virtual control vector, u = [∑F x ∑F y ∑M z ] T ; Step 5: Ignore the impact of the fault on lateral forces. Transform the impact of the actuator fault on the entire vehicle, i.e., its effect on the vehicle's electrical braking and driving, into the trend of tire force changes F on the wheels. xim =λ i F xi +q i (i = 1, 2, 3, 4, λ) i The system is introduced into the vehicle system (∈[0,1]). The feasible region of tire force is recalculated based on the influence of tire force changes on braking. This is called the tire force fault-tolerant feasible region, which is then used to obtain the tire force distribution constraints after a fault. The feasible region of the vehicle driving state is recalculated based on the tire force fault-tolerant feasible region. This is called the vehicle fault-tolerant feasible region. The vehicle fault-tolerant feasible region includes: Longitudinal force fault-tolerant feasible region: Lateral resultant force fault-tolerant feasible region: The yaw moment tolerance feasible region: in F fxi , F fyi (i = 1, 2, 3, 4) represent the virtual steering angles at the upper and lower bounds of the feasible region for tire force tolerance, respectively. The corresponding extreme values ​​of longitudinal and lateral forces; Step Six: Fault-Tolerant Controller Design. First, the control constraints are changed by adjusting the predictive control conditions of the upper-level model, i.e., from the feasible region of the vehicle's driving state to the fault-tolerant feasible region of the vehicle. Secondly, the allocation matrix of the lower-level control is reconstructed to optimize the distribution of tire force so that it meets the fault-tolerant feasible region of tire force. The virtual control quantity is adjusted as follows: Its control efficiency matrix changes as follows: Step 7: When one or more of the virtual control quantities output by the upper-level controller, namely the longitudinal resultant force, lateral resultant force, and yaw resultant moment, are in n T If the vehicle remains on the boundary of the fault-tolerant feasible region for ≥20 consecutive sampling periods T, it is considered a serious fault. For situations where the vehicle is determined to be a serious fault within the fault-tolerant feasible region, the relationship matrix between the control quantity and the state quantity obtained from model predictive control is used as the basis. Taking into account the vehicle's execution capability and tracking effect, a reference input reshaping scheme is designed based on the vehicle's fault-tolerant feasible region and tracking deviation. The tracking deviation is the difference between the reference input obtained from the reference model formula (1) and the actual state quantity of the vehicle collected in step one. The objective function of the reshaping scheme is as follows: Where Y d Reshaping reference input, Y re As the original reference input, U b =U tb -U l -Δu represents the degree of approximation of the virtual control variable to the feasible region, U l For the previous virtual control quantity, U tb To be with U l Let the feasible region boundary values ​​with the same sign be... The prediction formula (37) Y(t)=ψ is derived from the model predictive controller. t ζ(t)+Θ t Δu(t)+ι t Ψ(t)(Let: Φ=ψ t ζ(t)+ι t The reshaping constraints are obtained by adjusting Ψ(t) and the control constraints (38) of the model predictive controller as follows: U max U min Let ΔU be the upper and lower bounds of the fault-tolerant feasible region of the entire vehicle. max ΔU min Given the upper and lower limits of the control increment for the predictor controller, A I The expression (38) is the transformation matrix for converting control quantity constraints into control increment constraints. Its opposite; The final calculated reference input after reshaping is: Where E is a matrix with the same number of rows and columns as Y, and all elements are one. t F For Θ t The reverse; Based on the reshaping scheme obtained from the reference input, the reshaping state is obtained, namely the reshaping longitudinal velocity, the reshaping lateral velocity, and the reshaping yaw rate. The equation between the longitudinal reference velocity and the yaw rate in the reference model formula (1) is discussed. If the calculated longitudinal velocity and yaw rate do not conform to the reference model formula (1) and... Then, the steady-state steering model is used to calculate the steering angle to maintain the original route. The steering angle is calculated based on the reference model to adjust the target input of the control system by changing the original route. This involves reducing speed and changing the front wheel steering angle, allowing the car to change its trajectory while reducing performance; where ω... rc V xc To reshape the yaw rate and the longitudinal rate, ω rvx The yaw rate is calculated from the longitudinal velocity of the reshaped model using formula (1) of the reference model.

2. The fault-tolerant control method for actuator failure in an electric vehicle stability system according to claim 1, characterized in that, The feasible region of tire force in step two is: Friction circle guide range (θ) yi ,θ xi ):

3. The fault-tolerant control method for actuator failure in an electric vehicle stability system according to claim 1, characterized in that, Step four establishes a three-degree-of-freedom model of the vehicle, involving longitudinal, lateral, and yaw motions. The model is based on the following assumptions: Ⅰ: Ignore the influence of the steering system and use the front wheel angle as the system input; II: Ignoring the role of the suspension, the vehicle body only performs planar motion parallel to the ground; III: The vehicle's speed along the axial direction remains constant; The model is expressed as shown in the equilibrium equations: In the above equation, m is the total mass of the car, and V... x V is the longitudinal velocity. y Let ω be the lateral velocity. r I is the yaw rate. z Let F be the moment of inertia of the entire vehicle about the Z-axis. x ,∑F y ,∑M z These are the longitudinal resultant force, lateral resultant force, and yaw moment acting on the vehicle, respectively. Discretize and linearize the three-degree-of-freedom vehicle dynamics model: T is the sampling period, resulting in the discrete state equation: Where A, B, and C are respectively: Further analysis revealed: Where d k Indicates the calculation deviation. These are the observed values; Constructing the state vector: We obtain the new state-space expression: The state is predicted based on the state-space expression, and the system output is obtained from the predicted state. The future output Y(t) of the system is represented by the following matrix: Y(t)=ψ t g(t)+Θ t Δu(t)+i t Ψ(t)(37) in: N p To predict the step size, N c To control the step size; Control constraints: in: ∑F xtmax ,∑F ytmax ,∑M ztmax ∑F xtmin ,∑F ytmin ,∑M ztmin Let ΔU be the upper and lower bounds of the feasible region corresponding to the upper and lower limits of the longitudinal resultant force, lateral resultant force, and yaw moment, respectively. max ΔU min Predict the upper and lower limits of the control increment for a given model.

4. The fault-tolerant control method for actuator failure in an electric vehicle stability system according to claim 1, characterized in that, In step five, the tire force change caused by actuator failure is added to the tire force feasible region to obtain the tire force fault-tolerant feasible region. Tire force tolerance feasible region: Simultaneously, the guide range of the friction circle after the fault is obtained, the guide range of the fault friction circle (θ) fyi ,θ fxi ):

5. The fault-tolerant control method for actuator failure in an electric vehicle stability system according to claim 1, characterized in that, In step seven, the reference input is reshaped based on the severe faults identified in the vehicle's fault-tolerant feasible region. Since the reference input reshaping can only provide the dynamic state value of the vehicle's stable driving (i.e., the vehicle speed will always decrease during the consideration of the feasible region), and cannot determine whether the vehicle can drive stably at a certain steady-state speed or constant speed, the following reshaping criteria should be considered when determining the final steady-state speed in consideration of the vehicle's driving capability: (1) Maintain or adjust the trajectory during driving under fault conditions so that the vehicle speed and steering angle meet the requirements of the trajectory curvature: That is, the vehicle's longitudinal velocity and yaw rate meet the curvature change requirements; (2) During the reshaping process, the longitudinal resultant force, lateral resultant force, and yaw moment of the vehicle must always be kept within the feasible range, i.e.: (3) Determination of vehicle stability: The longitudinal force, lateral force and yaw moment of the vehicle have a certain stability margin χ∈[0,1], that is, the vehicle has a certain driving ability under the new state to avoid instability under the continuous aggravation of the fault:

Images